FIFTH

EDITION

IRVING M. COPI University of Hawaii

9
OXFORD & IBH PUBLISHING Ca
New Delhi Bombay Calcutta
First Published January 1960

This book has been published with a subsidy under the Indo-American
Text Book Programme operated by National Book Trust, India.

Published by and IBH Publishing Co., 66 Janpath,

New DelM-lf5j|#^ted at Everest Press, 4 Chamelian Road* Delhi-6,
Introduction: Logic and Language
1.1 What Is Logic? 1
1.2 The Nature of Argument 2
1.3 Truth and Validity 4
1.4 Symbolic Logic 5

Arguments Containing Compound Statements

2.1 Simple and Compound Statements S
2.2 Conditional Statements 16
2.3 Argument Forms and Truth Tables 19
2.4 Statement Forms 27

The Method of Deduction

3.1 Formal Proof of Validity 32

3.2 The Rule of Replacement 39
3.3 Proving Invalidity 48
3.4 The Rule of Conditional Proof 49
3.5 The Rule of Indirect Proof 52
3.6 Proofs of Tautologies 54
3.7 The Strengthened Rule of Conditional Proof
3.8 Shorter Truth Tahiti Technique—Reductio ad
Absurd-urn Method 61

Quantification Theory
4.1 Singular Propositions and General
Propositions 63
4.2 Proving Validity: Preliminary Quantification
Rules 71
4.3 Proving Invalidity 78
4.4 Multiply-Ceneral Propositions 83
xil Contents

4.5 Quantification Rules 89

4.6 More on Proving Invalidity 105
4.7 Logical Truths Involving Quantifiers 108

5 The Logic of Relations 116

5.1 Symbolizing Relations 116
5.2 Arguments Involving Relations 130
5.3 Some Attributes of Relations 134
5.4 Identity and the Definite Description 140
5.5 Predicate Variables and Attributes of
Attributes 150

6 Deductive Systems 157

6.1 Definition and Deduction 157
6.2 Euclidean Geometry 158
6.3 Formal Deductive Systems 162
6.4 Attributes of Formal Deductive Systems 164
6.5 Logistic Systems 166

7 Set Theory 170

7.1 The Algebra of Classes 170
7.2 Axioms for Class Algebra 175
7.3 Zermelo-Fraenkel Set Theory (ZF)—The First Six
Axioms 176 4
7.4 Relations and Functions 185
7.5 Natural Numbers and the Axiom of Infinity 190
7.6 Cardinal Numbers and the Choice Axiom 195
7.7 Ordinal Numbers and the Axioms of Replacement
and Regularity 202

8 A Propositional Calculus 213

8.1 Object Language and Metalanguage 213
8.2 Primitive Symbols and Well Formed
Formulas 215
8.3 Axioms and Demonstrations 227
8.4 Independence of the Axioms 231
8.5 Development of the Calculus 237
8.6 Deductive Completeness 250
 Contents xili

9 Alternative Systems and Notations 259

9.1 Alternative Systems of Logic 259
9.2 The Hilbert-Ackermann System 260
9.3 The Use of Dots as Brackets 276
9.4 A Parenthesis-Free Notation 279
9.5 The Stroke and Dagger Operators 281
9.6 The Nicod System 282

A First-Order Function Calculus 290

io 10.1 The New Logistic System RS, 290
10.2 The Development of RS, 296
10.3 Duality 302
10.4 RS, and the ‘Natural Deduction’
Techniques 307
10.5 Normal Forms 311
10.6 Completeness of RS, 318
10.7 RS, with Identity 328
10.8 First-Order Logic Including ZF Set Theory 331

Appendix A: Incompleteness of the Nineteen Rules 333

Appendix B: Normal Forms and Boolean Expansions 337
Appendix C: The Ramified Theory of Types 344
Solutions to Selected Exercises 355
Special Symbols 381
Index 385
 1
Introduction:

1.1 What Is Logic?

It is easy to find answers to the question ‘What is Logic?' According to

Charles Peirce, ‘Nearly a hundred definitions of it have been given’.1 But
Peirce goes on to write: ‘It will, however, generally be conceded that its
central problem is the classification of arguments, so that all those that are bad
are thrown into one division, and those which are good into another.. .\
The study of logic, then, is the study of the methods and principles used in
distinguishing correct (good) from incorrect (bad) arguments. This definition is
not intended to imply, of course, that only the student of logic can make the
distinction. But the study of logic will help one to distinguish between correct
and incorrect arguments, and it will do so in several ways. First of all, the
proper study of logic will approach it as an art as well as a science, and the
student will do exercises in all parts of the theory being learned. Here, as
anywhere else, practice will help to make perfect. In the second place, the
study of logic, especially symbolic logic, like the study of any other exact
science, will tend to increase proficiency in reasoning. And finally, the study of
logic will give students certain techniques for testing the validity of all
arguments, including their own. This knowledge is of value because when
mistakes arc easily detected, they are less likely to be made.
Logic has frequently been defined as the science of reasoning. That defini¬
tion, although it gives a clue to the nature of logic, is not quite accurate.
Reasoning is that special kind of thinking called inferring, in which conclusions
are drawn from premisses. As thinking, however, reasoning is not the special
province of logic, but part of the psychologist’s subject matter as well. Psy¬
chologists who examine the reasoning process find it to be extremely complex
and highly emotional, consisting of awkward trial and error procedures that are
illuminated by sudden—and sometimes apparently irrelevant—flashes of
insight. These are all of importance to psychology. Logicians, however, are not
interested in the actual process of reasoning, but rather with the correctness of

1 Logic, in Dictionary of Phibsophy and Psychology, James Mark Baldwin, ed., Macmillan
Publishing Co., Inc., New York, 1925.
2 Introduction: Logic and Language [Ch. 1
the completed reasoning process. Their question is always: Does the conclusion
that is reached follow from the premisses used or assumed? If the premisses
provide adequate grounds for accepting the conclusion, if asserting the prem¬
isses to be true warrants asserting the conclusion to be true also, then the
reasoning is correct. Otherwise, the reasoning is incorrect. Logicians’ methods
and techniques have been developed primarily for the purpose of making this
distinction clear. Logicians are interested in all reasoning, regardless of its
subject matter, but only from this special point of view.

1.2 The Nature of Argument

Inferring is an activity in which one proposition is affirmed on the basis of

one or more other propositions that are accepted as the starting point of the
process. The logician is not concerned with the process of inference, but with
the propositions that are the initial and end points of that process, and the
relationships between them.
Propositions are either true or false, and in this they differ from questions,
commands, and exclamations. Grammarians classify the linguistic formulations
of propositions, questions, commands, and exclamations as declarative, inter¬
rogative, imperative, and exclamatory sentences, respectively. These are
familiar notions. It is important to distinguish between declarative sentences
and the propositions they may be uttered to assert. A declarative sentence is
always part of a language, the language in which it is spoken or written,
whereas propositions are not peculiar to any of the languages in which they
may be expressed. Another difference between declarative sentences and
propositions is that the same sentence may be uttered in different contexts to
assert different propositions. (For example, the sentence ‘I am hungry’ may be
uttered by different persons to make different assertions.)
The same sort of distinction can be drawn between sentences and statements.
The same statement can be made using different words, and the same sentence
can be uttered in different contexts to make different statements. The terms
‘proposition’ and ‘statement’ are not exact synonyms, but in the writings of
logicians they are used in much the same sense. In this book, both terms will be
used. In the following chapters, we shall also use the term ‘statement’ (espe¬
cially in Chapters 2 and 3) and the term ‘proposition’ (especially in Chapters 4
and 5) to refer to the sentences in which statements and propositions are
expressed. In each case the c ontext should make clear what is meant.
Corresponding to every possible inference is an argument, and it is with
these arguments that logic is chiefly concerned. An argument maybe defined as
any group of propositions or statements, of which one is claimed to follow from
the others, which are alleged to provide grounds for the truth of that one. In
ordinary usage, the word ‘argument’ also has other meanings, but in logic it has
the technical sense explained. In the following chapters we shall use the word
‘argument’ also in a derivative sense'to refer to any sentence or collection of
Sec. 1.2] The Nature of Argument 3
sentences in which an argument is formulated or expressed. When we do, we
shall be presupposing that the context is sufficiently clear to ensure that unique
statements are made or unique propositions are asserted by the utterance of
those sentences.
Every argument has a structure, in the analysis of which the terms ‘premiss’
and ‘conclusion’ are usually employed. The conclusion of an argument is that
proposition which is affirmed on the basis of the other propositions of the
argument. These other propositions, which are affirmed as providing grounds
or reasons for accepting the conclusion, are the premisses of that argument.
We note that ‘premiss’ and ‘conclusion’ are relative terms, in the sense that
the same proposition can be a premiss in one argument and a conclusion in
another. Thus the proposition All humans are mortal is premiss in the argu¬
ment

All humans are mortal.

Socrates is human.
Therefore, Socrates is mortal.

and conclusion in the argument

All animals are mortal.

All humans are animals.
Therefore, all humans are mortal.

Any proposition can be either a premiss or a conclusion, depending upon its

context. It is a premiss when it occurs in an argument in which it is assumed for
the sake of proving some other proposition. And it is a conclusion when it
occurs in an argument that is claimed to prove it on the basis of other
propositions that arc assumed.
It is customary to distinguish between deductive and inductive arguments.
All arguments involve the claim that their premisses provide some grounds for
the truth of their conclusions, but only a deductive argument involves the claim
that its premisses provide absolutely conclusive grounds. The technical terms
‘valid’ and ‘invalid’ are used in place of ‘correct’ and ‘incorrect’ in character¬
izing deductive arguments. A deductive argument is valid when its premisses
and conclusion are so related that it is absolutely impossible for the premisses
to be true unless the conclusion is true also. The task of deductive logic is to
clarify the nature of the relationship that holds between premisses and con¬
clusion in a valid argument, and to provide techniques for discriminating valid
from invalid arguments.
Inductive arguments involve the claim only that their premisses provide
some grounds for their conclusions. Neither the term ‘valid’ nor its opposite
‘invalid’ is properly applied to inductive arguments. Inductive arguments differ
among themselves in the degree of likelihood or probability that their prem¬
isses confer upon their conclusions. Inductive arguments are studied in indue-
4 Introduction: Logic and Language [Ch. 1

tive logic. In this book, however, we shall be concerned only with deductive
arguments, and shall use the word ‘argument’ to refer to deductive arguments
exclusively.

1.3 Truth and Validity

Truth and falsehood characterize propositions or statements and may also be

said to characterize the declarative sentences in which they are formulated.
Arguments, however, are not properly characterized as being either true or
false but as valid or invalid.2 There is a connection between the validity or
invalidity of an argument and the truth or falsehood of its premisses and
conclusion, but the connection is by no means a simple one.
Some valid arguments contain true propositions only, as, for example.

All bats are mammals.

All mammals have lungs.
Therefore, all bats have lungs.

An argument may contain false propositions exclusively and still be valid, as,
for example,

All trout are mammals.

All mammals have wings.
Therefore, all trout have wings.

This argument is valid because if its premisses were true, its conclusion would
have to be true also, even though, in fact, they are all false. These two examples
show that although some valid arguments have true conclusions, not all of them
do. The validity of an argument does not, therefore, guarantee the truth of its
conclusion.
When we consider the argument

If I am President, then I am famous.

I am not President.
Therefore, I am not famous.

we can see that although both premisses and conclusion are true, the argument
is invalid. Its invalidity becomes obvious when it is compared with another
argument of the same form:

2Some logicians use the term ‘valid’ to characterize statements that are logically true. We shall
adopt that usage in Chapter 10, Section 10.6. Until then, however, we apply the terms ‘valid’ and
‘invalid’ to arguments exclusively.
Sec. 1.4] Symbolic Logic 5
If Rockefeller is President, then Rockefeller is famous.
Rockefeller is not President.
Therefore, Rockefeller is not famous.

This argument is clearly invalid because its premisses are true but its conclu¬
sion is false. The two latter examples show that although some invalid argu¬
ments have false conclusions, not all of them do. The falsehood of its conclusion
does not guarantee the invalidity of an argument. But the falsehood of its
conclusion does guarantee that either the argument is invalid or at least one of
its premisses is false.
An argument must satisfy two conditions to establish the truth of its conclu¬
sion. It must be valid, and all of its premisses must be true. Such an argument is
termed ‘sound’. To determine the truth or falsehood of premisses is the task of
scientific inquiry in general, since premisses may deal with any subject matter
at all. But determining the validity or invalidity of arguments is the special
province of deductive logic. The logician is interested in the question of
validity even for arguments that might be unsound because their premisses
might happen to be false.
A question might be raised about the legitimacy of that interest. It might be
suggested that logicians should confine their attention to arguments that have
true premisses only. It is often necessary, however, to depend upon the validity
of arguments whose premisses are not known to be true. Scientists test their
theories by deducing from them conclusions that predict the behavior of
observable phenomena in the laboratory or observatory. The conclusion is then
tested directly by observation of experimental data and if it is true, the results
confirm the theory from which the conclusion was deduced. If the conclusion is
false, the results disconfirm or refute the theory. In either case, the scientist is
vitally interested in the validity of the argument by which the testable con¬
clusion is deduced from the theory being investigated, for if that argument is
invalid, his whole procedure is without point. Although an oversimplification
of scientific method, our example shows that questions of validity are important
even for arguments whose premisses are not true.

1.4 Symbolic Logic

It has been explained that logic is concerned with arguments and that
arguments contain propositions or statements as their premisses and conclu¬
sions. These premisses and conclusions are not linguistic entities, such as
declarative sentences, but are, rather, what declarative sentences are typically
uttered to assert. The communication of propositions and arguments, however,
requires the use of language, and this complicates our problem. Arguments
formulated in English or any other natural language are often difficult to
appraise because of the vague and equivocal nature of the w'ords in which they
6 Introduction: Logic and Language [Ch. 1

are expressed, the ambiguity of their construction, the misleading idioms they
may contain, and their pleasing but deceptive metaphorical style. The resolu¬
tion of these difficulties is not the central problem for the logician, however, for
even when they are resolved, the problem of deciding the validity or invalidity
of the argument still remains.
To avoid the peripheral difficulties connected with ordinary language,
workers in the various sciences have developed specialized technical vocabu¬
laries. The scientist economizes the space and time required for writing his
reports and theories by adopting special symbols to express ideas that would
otherwise require a long sequence of familiar words to formulate. This has the
further advantage of reducing the amount of attention needed, for when a
sentence or equation grows too long, its meaning is more difficult to grasp. The
introduction of the exponent symbol in mathematics permits the expression of
the equation

AxAxAxAxAxAxAxAxAxAxAxA
= BXBXBXBXBXB XB

more briefly and intelligibly as

A12 = B7

A like advantage has been obtained by the use of graphic formulas in organic
chemistry. The language of every advanced science has been enriched by
similar symbolic innovations.
A special technical notation has been developed for logic as well. Aristotle
made use of certain abbreviations to facilitate his own investigations. Modern
symbolic logic augmented this base by the introduction of many more special
symbols. The difference between the old and the new logic is one of degree
rather than of kind, but the difference in degree is tremendous. Modern
symbolic logic has become immeasurably more powerful a tool for analysis and
deduction through the development of its own technical language. The special
symbols of modern logic permit us to exhibit with greater clarity the logical
structures of arguments that may be obscured by formulation in ordinary
language. It is easier to divide arguments into the valid and the invalid when
they are expressed in a special symbolic language, for with symbols the
peripheral problems of vagueness, ambiguity, idiom, metaphor, and amphiboly
do not arise. The introduction and use of special symbols serve not only to
facilitate the appraisal of arguments, but also to clarify the nature of deductive
inference.
The logician’s special symbols are much better adapted to the actual
drawing of inferences than is ordinary language. Their superiority in this
respect is comparable to that of Arabic over the older Roman numerals for
purposes of computation. It is easy to multiply 148 by 47, but very difficult to
compute the product of CXLVIII and XLVII. Similarly, the drawing of
Sec. 1.4] Symbolic Logic 7
inferences and the evaluation of arguments is greatly facilitated by the adop¬
tion of a special logical notation. To quote Alfred North Whitehead, an
important contributor to the advance of symbolic logic:

... by the aid of symbolism, we can make transitions in ^reasoning almost

mechanically by the eye, which otherwise would call into play the
higher faculties of the brain.3

Although this book treats symbolic logic systematically rather than histori¬
cally, a few historical remarks may be appropriate at this point. Since the
1840s, symbolic logic has developed along two different historical paths. One
of them began with the English mathematician George Boole (1815-1864).
Boole applied algebraic notations and methods first to symbolize and then to
validate arguments of the kind studied by Aristotle in the fourth century b.c.
This route may be characterized as an effort to apply mathematical notations
and methods to traditional, nonmathematical kinds of arguments. The other
path began with the independent efforts of the English mathematician Augus¬
tus De Morgan (1806-1871) and the American scientist and philosopher
Charles Peirce (1839-1914) to devise a very precise notation for relational
arguments. The earlier logic had largely ignored this type of argument, which,
nevertheless, plays a central role in mathematics. This historical path may be
characterized as an effort to create a new quasi-mathcmatical kind of logical
notation and analytical technique for use in mathematical derivations and
demonstrations.
These two historical paths coalesced in the brilliant works of the German
mathematician and philosopher Gottlob Frege (1848-1925), the Italian math¬
ematician Guiseppe Peano (1858-1932), and the English philosophers Alfred
North Whitehead (1861-1947) and Bertrand Bussell (1872-1970), whose
Principia Mathematica was an important landmark in the history of symbolic
logic. Some of Boole’s contributions are reported in the first two sections of
Chapter 7 and in Appendix B. The contributions of the others have become so
thoroughly incorporated into modern symbolic logic that only occasional
references to their more distinctive ideas are appropriate.

3 A. N. Whitehead, An Introduction to Mathematics, Oxford University Press, Oxford, England,

1911.
Arguments Containing
Compound Statements
2.1 Simple and Compound Statements

All statements can he divided into two kinds, simple and compound. A
simple statement is one that does not contain any other statement as a compo¬
nent part, whereas every compound statement does contain another statement
as a component part. For example, ‘Atmospheric testing of nuclear weapons
will be discontinued or this planet will become uninhabitable’ is a compound
statement that contains, as its components, the two simple statements ‘Atmos¬
pheric testing of nuclear weapons will be discontinued’ and ‘this planet will
become uninhabitable’. The component parts of a compound statement may
themselves be compound, of course.
As the term ‘component’ is used in logic, not every statement that is part of
a larger statement is a component of the larger statement. For example, the last
six words of the statement, The third wife of Bertrand Russell was a beautiful
girl’, can be regarded as a statement in its own right. But it is not a component
part or a component of the larger statement containing it. For a part of a
statement to be a component of a larger statement, two conditions must be
satisfied. First, the part must be a statement in its own right; and second, if the
part is replaced in the larger statement by any other statement, the result of
that replacement must be meaningful. Although the first condition is satisfied in
the example given, the second is not. For if the part ‘Bertrand Russell was a
beautiful girl’ is replaced by the statement ‘Where there’s smoke, there’s fire’,
the result of that replacement is arrant nonsense.1
The statement ‘Roses are red and violets are blue’ is a conjunction, a
compound statement formed by inserting the word ‘and’ between two state¬
ments. Two statements so combined are called conjuncts. The word ‘and’ has
other uses, however, as in the statement ‘Castor and Pollux were twins', which
is not compound, but a simple statement asserting a relationship. We introduce
the dot *•’ as a special symbol for combining statements conjunctively. Using

lrThis account of compound and component statements has been suggested by Proiessor
('. Mason Myers of Northern Illinois University, Professor Alex Blum of Bar-llan University, and
Professor James A. Martin of the University" yf Wyoming. See Martin’s ‘How Not to Define
Truth-Functionality,' Logique et Analyse, 1970, 14' Anncc, N° 52, pp. 476-482.
Sec. 2.1] Simple and Compound Statements 9

this notation, the preceding conjunction is written ‘Roses are red-violets are
blue’. Where p and q are any two statements whatever, their conjunction is
written p*q.
Every statement is either true or false, so we can speak of the truth value of
a statement, where the truth value of a true statement is true and the truth
value of a false statement is false. There are two broad categories into which
compound statements can be divided, according to whether or not there is
anything other than the truth values of its component statements that deter¬
mines the truth value of the compound statement. The truth value of the
conjunction of two statements is completely determined by the truth value of
its conjuncts. A conjunction is true if both its conjunct!* are true, but false
otherwise. For this reason a conjunction is a truth-functional compound
statement, and its conjuncts are truth-functional components of it.
Not every compound statement is truth-functional. For example, the
truth-value of the compound statement ‘Percival Lowell believed that Mars is
inhabited’ is not in any way determined by the truth value of its component
simple statement ‘Mars is inhabited’. This is so because it could be true that
Percival Lowell believed that Mars is inhabited, regardless of whether it is
inhabited or not. A person can believe one truth without believing another, and
a person can believe one falsehood without believing all of them. So the
component ‘Mars is inhabited’ is not a truth-functional component of the
compound statement ‘Percival Lowell believed that Mars is inhabited,’ and
the latter statement is not a truth-functional compound statement.
We define an occurrence of a component of a compound statement to be a
truth-functional component of that compound statement provided that: If
that occurrence of the component is replaced in the compound (or in any
component of the compound statement which contains the component in
question) by different statements that have the same truth value as each other,
the different compound statements produced by these replacements will also
have the same truth values as each other. And now a compound statement is
defined to be a truth-functional compound statement, if all of its components
are truth-functional components of the statement.2
The only compound statements we shall consider here will be truth-func-
tionally compound statements. Therefore, in the rest of this book we shall use
the term “simple statement” to refer to any statement that is not truth-func-
tionally compound.
Since conjunctions are truth-functionally compound statements, our dot
symbol is a truth-functional connective. Given any two statements p and q
there are just four possible sets of truth values they can have, and in every case
the truth value of their conjunction pmq is uniquely determined. The four
possible cases can be exhibited as follows:

2Compare this definition with the one proposed by Professor David II. Sanford in his ‘What Is
Truth Functional Component?’ Lo^iquc ct Analyse. 1970, I t*' Annce, N° 52, pp. 483-486.
10 Arguments Containing Compound Statements [Ch. 2

in case p is true and q is true, p*q is true;

in case p is true and q is false, puj is false;
in case p is false and q is true, p*q is false;
in case p is false and q is false, puq is false. '

Representing the truth values true and false by the capital letters T and ‘F\
respectively, the way in which the truth value of a conjunction is determined
by the truth values of its conjuncts can be displayed more briefly by means of
a truth table as follows:

V_P'cl
T T T
T F F
F T F
F F F

Since it specifics the truth value of p*q in every possible case, this truth table
can be taken as defining the dot symbol. Other English words such as ‘more¬
over’, ‘furthermore’, ‘but’, ‘yet’, ‘still’, ‘however’, ‘also’, ‘nevertheless’, ‘al¬
though’, and so forth, and even the comma and the semicolon, are also used to
conjoin two statements into a single compound one, and all of them can be
indifferently translated into the dot symbol, so far as truth values are con¬
cerned.
The statement ‘It is not the case that lead is heavier than gold’ is also
compound, being the negation (or denial or contradictory) of its single com¬
ponent statement ‘lead is heavier than gold’. We introduce the symbol
called a curl (or a tilde) to symbolize negation. There are often alternative
formulations in English of the negation of a given statement. Thus where L
symbolizes the statement ‘lead is heavier than gold’, the different statements ‘it
is not the case that lead is heavier than gold’, ‘it is false that lead is heavier than
gold’, ‘it is not true that lead is heavier than gold’, ‘lead is not heavier than
gold’ are all indifferently symbolized as — L. More generally, where p is any
statement whatever, its negation is written ~~p. Since the negation of a true
statement is false and the negation of a false statement is true, we can take the
following truth table as defining the curl symbol:

P_
T “ F
F T

When two statements are combined disjunctively by inserting the word ‘or’
between them, the resulting compound statement is a disjunction (or alterna¬
tion), and the two statements so combined are called disjuncts (or alternatives).
The word ‘or’ has two different senses’, one of which is clearly intended in the
Sec. 2.1] Simple and Compound Statements 11

statement ‘Premiums will be waived in the event of sickness or unemploy¬

ment*. The intention here is obviously that premiums are waived not only for
sick persons and for unemployed persons, but also for persons who are both sick
and unemployed. This sense of the word or* is called weak or inclusive. Where
precision is at a premium, as in contracts and other legal documents, this sense
is made explicit by use of the phrase ‘and/or’.
A different sense of ‘or’ is intended when a restaurant lists ‘tea or coffee* on
its table d’hote menu, meaning that for the stated price of the meal the
customer can have one or the other, but not both. This second sense of ‘or* is
called strong or exclusive. Where precision is at a premium and the exclusive
sense of ‘or’ is intended, the phrase ‘but not both’ is often added.
A disjunction which uses the inclusive ‘or* asserts that at least one disjunct is
true, whereas one which uses the exclusive ‘or’ asserts that at least one disjunct
is true but not both are true. The partial common meaning that at least one
disjunct is true, is the whole meaning of an inclusive disjunction and a part of
the meaning of an exclusive disjunction.
In l iatin, the word ‘vel’ expresses the inclusive sense of the word ‘or’ and the
word ‘aut* expresses the exclusive sense. It is customary to use the first letter of
‘vel’ to symbolize ‘or’ in its inclusive sense. Where p and (/ are any two
statements whatever, their weak or inclusive disjunction is written p v q. The
symbol ‘v*, called a wedge (or a vee), is a truth-functional connective, and is
defined by the following truth table:

p_(j_pyji
f T T
TFT
F T T
F F F

An obviously valid argument containing a disjunction is the following

Disjunctive Syllogism:

The United Nations will become more responsible or there will be a third
world war.
The United Nations will not become more responsible.
Therefore, there will be a third world war.

It is evident that a Disjunctive Syllogism is valid on either interpretation of the

word ‘or*, that is, regardless of whether its first premiss asserts an inclusive or
exclusive disjunction. It is usually difficult, and sometimes impossible, to
discover which sense of the word ‘or’ is intended in a disjunction. But the
typical valid argument that has a disjunction for a premiss is, like the Dis¬
junctive Syllogism, valid on either interpretation of the word ‘or*. Hence we
effect a simplification by translating any occurrence of the word ‘or’ into the
logical symbol V—regardless of which sense of or' is intended. Of course.
12 Arguments Containing Compound Statements [Ch. 2

where it is explicitly stated that the disjunction is exclusive, by use of the added
phrase ‘but not both’, for example, we do have the symbolic apparatus for
expressing that sense, as will be explained below.
The use of parentheses, brackets, and braces for punctuating mathematical
expressions is familiar. In the absence of a special convention, no number is
uniquely denoted by the expression ‘6 + 9 -f- 3\ although when punctuation
makes clear how its constituents are to be grouped, it denotes either 5 or 9.
Punctuation is needed to resolve ambiguity in the language of symbolic logic
too, since compound statements may themselves be combined to produce more
complicated compounds. Ambiguity is present in paq v r, which could be
either the conjunction of p with q v r, or else the disjunction of paq with r.
These two different senses are unambiguously given by different punctuations:
pa(q v r) and (paq) v r. In case p and q are both false and r is true, the first
punctuated expression is false (since its first conjunct is false), but the second
punctuated expression is true (since its second disjunct is true). Here, a differ¬
ence in punctuation makes all the difference between truth and falsehood. In
symbolic logic, as in mathematics, we use parentheses, brackets, and braces for
punctuation. To cut down on the number of punctuation marks required,
however, we establish the symbolic convention that in any expression the curl
will apply to the smallest component that the punctuation permits. Thus the
ambiguity of — p v </, which might mean either (~~p) v q or ~~(p v q), is
resolved, by our convention, to mean the first of these. The curl can, and
therefore by our convention does, apply to the first component p rather than lo
the larger expression p v (/.
The word ‘either’ has a variety of different uses in English. It has conjunctive
force in The Disjunctive Syllogism is valid on either interpretation of the word
‘or’/ It frequently serves merely to introduce the first disjunct in a.disjunction,
as in ‘Either the United Nations will become more responsible or there will be
a third world war’. Perhaps the most useful function of the word ‘either’ is to
punctuate some compound statements. Thus the sentence

More stringent antipollution measures will be enacted and the laws will
be strictly enforced or thejjuality of life will be degraded still further.

can have its ambiguity resolved in one direction by placing the word ‘cither’ at
its beginning, or in the other direction by inserting the word ‘either’ right after
the word ‘and’. Such punctuation is effected in our symbolic language by
parentheses. The ambiguous formula paqv r discussed in the preceding para¬
graph corresponds to the ambiguous sentence considered in this one. The two
different punctuations of the formula correspond to the two different punctu¬
ations of the sentence effected by the two different insertions of the word
‘either’.
Not all conjunctions are formulated by explicitly placing the word ‘and’
between complete sentences, as in ‘Charlie’s neat and Charlie’s sweet’. Indeed,
the latter would more naturally be expressed as ‘Charlie’s neat and sweet’. The
Sec. 2.1] Simple and Compound Statements 13

familiar ‘Jack and Jill went up the hill’ is the more natural way of expressing the
conjunction ‘Jack went up the hill and Jill went up the hill’. It is the same with
disjunctions: ‘Either Alice or Betty will be elected’ expresses more briefly the
proposition alternatively formulated as ‘Either Alice will be elected or Betty
will be elected’; and ‘Charlene will be either secretary or treasurer’ expresses
somewhat more briefly the same proposition as ‘Either Charlene will be
secretary or Charlene will be treasurer’.
The negation of a disjunction is often expressed by using the phrase ‘nei-
ther-nor’. Thus the disjunction ‘Either Alice or Betty will be elected’ is denied
by the statement ‘Neither Alice nor Betty will be elected’. The disjunction
would be symbolized as A v B and its negation as either —(A v B) or as
(— A)'(^B). (The logical equivalence of these two formulas will be discussed
in Section 2.4.) To deny that at least one of two statements is true is to assert
that both of the two statements are false.
The word ‘both’ serves various functions. One is simply a matter of emphasis.
To say ‘Both Jack and Jill went up the hill’ is only to emphasize that the two of
them did what they are asserted to have done by saying ‘Jack and Jill went up
the hill'. A more useful function of the word ‘both’ is punctuational, like that of
‘either’. ‘Both .. . and -• are not-’ is used to make the same statement as
‘Neither . . . nor • - is.In such sentences, the order of the words ‘both' and
‘not’ is very significant. There is a great difference between

Alice and Betty will not both be elected.

and

Alice and Betty will both not be elected.

The former would be symbolized as —(A*B), the latter as (—-AJ•(—-7^).

Finally, it should be remarked that the word ‘unless’ can also be used in
expressing the disjunction of two statements. Thus ‘Our resources will soon be
exhausted unless more recycling of materials is effected’ and ‘Unless more
recycling of materials is effected our resources will soon be exhausted’ can
equally well be expressed as ‘Either more recycling of materials is effected or
our resources will soon be exhausted’ and symbolized as M v E.
Since an exclusive disjunction asserts that at least one of its disjuncts is true
but they are not both true, we can symbolize the exclusive disjunction of any
two statements p and q quite simply as (p v q)m~~(pmq). Thus we are able to
symbolize conjunctions, negations, and both inclusive and exclusive disjunc¬
tions. Any compound statement that is built up out of simple statements by
repeated use of truth-functional connectives will have its truth value com¬
pletely determined by the truth values of those simple statements. For exam¬
ple, if A and R are true statements and X and Y are false, the truth value of the
compound statement ^[(^A vX) v ~~(B'mY)] can be discovered as follows.
Since A is true, —A is false, and since X is false also, the disjunction (^A v X)
is false. Since Y is false, the conjunction (B* Y) is false, and so its negation
~(B • Y) is true. Hence the disjunction (~A v X) v B • Y) is true, and its nega¬
tion, which is the original statement, is false. Such' a stepwise procedure,
beginning with the inmost components, always permits us to determine the
truth value of a truth-functionally compound statement from the truth values
of its component simple statements.

EXERCISES3

I. If A and B are true statements and X and Y are false statements, which of the
following compound statements are true?

*1. ~(A v X) 11. A vfX-(Bv Y)]

2. ~A v ~X 12. X v [A*(Y v 8)]
3. ~B—Y 13. ~ {~[~(A • —X) • ~A] —X}
4.. ~(B*Y) 14. ~{~[~(A-~B)-~A]-~A)
*5. Av(X-Y) *15. [(A-X)v~B]-~[(A-X)v~B]
6. (AvX)-Y 16. f(X-A) v ~Y] v ~[(X*A) v ~~Y]
7. (AvB)-(XvY) 17. [A-(Xv Y)]v~[(A-X)v(A-Y))
8. (A-B)v(X-Y) 18. [Xv(A-Y)]v~[(XvA)-(Xv Y)]
9. (A-X)v(B-Y) 19. [X-(AvB)]v~[(XvA)-(XvB)]
*10. A-[Xv(B-Y)j 20. [X v (A • Y)] v ~[X v A) v (X v Y)]

II. If A and B are known to he true, and X and Y are known to be false, but the
truth values of P and Q are not known, of which of the following statements can
you determine the truth value?

*1. P-X 6. P-~P

2. BvQ 7. ~Pv (~A v P)
3. ~A*P 8. ~P v (—X v P)
4. t>v~p 9. ~Bv(~pvP)
*5. Pv ~Y *10. Qv~~(P-Q)
11. (P-0v(~()v~P)
12. (PvQ)v(-P—Q)
13. (Pv0-(~Bv Y)
14. (P-())v~(p-P)
*15. (P-p)-(-AvX)
16. -(P-Q)vP
17. (~Pv0v(Pv~p)
18. [Pv(pvX)]-~[(Pvp)vX]
19. [Pv(^-A)]-~[(Pv9)-(PvA)]
20. [P-(Qv X)]v ~[(P-Q)v (P-X)] ■

3 Solutions to starred exercises will be found on pages 355-380.

Sec. 2.1] Simple and Compound Statements 15

III. Using the letters A, B, C, and D to abbreviate the simple statements: ‘Atlanta
wins their conference championship’, ‘Baltimore wins their conference champi¬
onship’, ‘Chicago wins the superbowl’, and ‘Dallas wins the superbowl’, symbol¬
ize the following:

*1. Either Atlanta wins their conference championship and Baltimore wins their
conference championship or Chicago wins the superbowl.
2. Atlanta wins their conference championship and^either Baltimore wins theif
conference championship or Dallas does not win the superbowl.
3. Atlanta and Baltimore will hotboth win their conference championships but
Chicago and Dallas will both not win the superbowl.
4. Either Atlanta or Baltimore will win their conference championships but
neither Chicago nor Dallas will win the superbowl.
*5. Either Chicago or Dallas will win the superbowl but they will not both win
the superbowl.
6. Chicago will win the superbowl unless Atlanta wins their conference cham¬
pionship.
7. It is not the case that neither Atlanta nor Baltimore wins their Conference
championship.
8. Either Chicago or Dallas will fail to win the superbowl.
9. Either Chicago or Dallas will win the superbowl unless both Atlanta and
Baltimore win their conference championships.
10. Either Chicago will win the superbowl and Dallas will not win the superbowl
or both Atlanta and Baltimore will win their conference championships.

IV. Using capital letters to abbreviate simple statements, symbolize the following:
*1. The words of his mouth were smoother than butter, but war was in his heart.
(Psalm 55:21)
2. Promotion cometh neither from the east, nor from the west, nor yet from the
south. (Psalm 75:6)
3. As for man, his days are as grass: as a flower of the field, so he Hourisheth.
(Psalm 103:15)
4. Wine is a mocker, strong drink is raging. (Proverbs 20:1)
*5. God hath made man upright; but they have sought out many inventions
(Ecclesiastes 7:29)
6. The race is not to the swift, nor the battle to the strong... (Ecclesiastes 9:11)
7. Love is strong as death; jealousy is cruel as the grave. (The Song of Solomon
8:6)
8. A bruised reed shall he not break, and the smoking flax shall he not quench.
(Isaiah 42:3)
9. Saul and Jonathan were lovely and pleasant in their lives ... (2 Samuel 1:23)
10. His eye was not dim, nor his natural force abated. (Deuteronomy 34:7)
11. The voice is Jacob’s voice, but the hands are the hands of Esau. (Genesis
27:22)
12. He shall return no more to his house, neither shall his place know him any
more. (Job 7:10)
16 Arguments Containing Compound Statements [Ch. 2

2.2 Conditional Statements

The compound statement Tf the train is late, then we shall miss our con¬
nection* is a conditional (or a hypothetical, an implication, or an implicative
statement). The component between the ‘if’ and the ‘then* is called the
antecedent (or the implicans or protasis), and the component that follows the
Then* is the consequent (or the implicate or apodosis). A conditional does not
state either that its antecedent is true or that its consequent is true; it says only
that if its antecedent is true, then its consequent is true also, that is, its
antecedent implies its consequent. The key to the meaning of a conditional is
the relation of implication that is asserted to hold between its antecedent and
its consequent, in that order.
If we examine a number of different conditionals, we can see that there are
different kinds of implications which they may express. In the conditional ‘If all
cats like liver and Dinah is a cat, then Dinah likes liver’, the consequent follows
logically from its antecedent. On the other hand, in the conditional ‘if the
figure is a triangle, then the figure has three sides’, the consequent follows from
the antecedent by the very definition of the word ‘triangle’. The truth of the
conditional ‘If this piece of gold is placed in aqua regia, then this piece of gold
dissolves’ is not a matter of either logic or definition. The connection asserted
here is causal and must be discovered empirically. These examples show that
there are different kinds of implications that constitute different senses of the
‘if-then’ phrase. Having noted these differences, we now seek to find some
identifiable common meaning, some partial meaning that is common to these
different types of conditionals.
Our discussion of ‘if-then’ will parallel our previous discussion of the word
‘or’. First, we pointed out two different senses of that word. Second, we noted
that there was a common partial meaning: That at least one disjunct is true was
seen to be involved in both the inclusive and the exclusive ‘or’. Third, we
introduced the special symbol V to represent this common partial meaning
(which was the whole meaning of ‘or’ in its inclusive sease). Fourth, we
observed that because arguments like the Disjunctive Syllogism are valid on
either interpretation of the word ‘oi^, symbolizing any occurrence of the word
‘or’ by the wedge symbol, preserves the validity of such arguments. And
because we are interested in arguments from the point of view of determining
their validity, this translation of the word ‘or’ into V, which may abstract from
or ignore part of its meaning in some cases, is wholly adequate for our present
purposes.
A common partial meaning of these different kinds of conditional statements
emerges when we ask what circumstances would suffice to establish the
falsehood of a conditional. Under what circumstances would we agree that the
conditional Tf this piece of gold is placed in this solution, then this piece of
gold dissolves’ is false? Clearly, the statement is false in the case that this piece
of gold is actually placed in this solution and does not dissolve. Any conditional
Sec. 2.2] Conditional Statements 17

with a true antecedent and a false consequent must be false. Hence any
conditional if p, then q is known to be false in case the conjunction pa~~q is
known to be true, that is, in case the antecedent of the conditional is true and
its consequent false. For the conditional to be true, the indicated conjunction
must be false, which means that the negation of that conjunction must be true.
In other words, for any conditional if p, then q to be true, ~~(p'~~q), the
negation of the conjunction of its antecedent with the negation of its conse¬
quent, must be true also. We may, then, regard the latter as a part of the
meaning of the former.
We introduce the new symbol ‘D’, called a horseshoe, to represent the
partial meaning that is common to all conditional statements. We define
‘p D */’ as an abbreviation for (p* ~~q)\ The horseshoe is a truth-functional
connective, whose exact significance is indicated by the following truth table:

P_ q ~q p—q p2JI
T T F F T T
T F T T F F
F T F F T T
F F T F T T

Here, the first two columns represent all possible truth values for the compo¬
nent statements p and q, and the third, fourth, and fifth represent successive
stages in determining the truth value of the compound statement — (p* — q) in
each case. The sixth column is identically the same as the fifth, since the
formulas that head them are defined to express the same proposition. The
horseshoe symbol must not be thought of as representing the meaning of
‘if-then’, or the relation of implication, but rather symbolizes a common partial
factor of the various different kinds of implications signified by the ‘if-then’
phrase.
We can regard the horseshoe as symbolizing a special, extremely weak kind
of implication. It is expedient for us to do so, since convenient ways to read
‘p D q’ are ‘if p, then q\ ‘p implies q\ or ‘p only if q. The weak implication
symbolized by ‘ D’ is called a material implication. Its special name indicates
that it is a special concept and not to be confused with the other, more usual
kinds of implication. Some conditional statements in English do express merely
material implications, as for example ‘If Russia is a democracy, then I’m a
Dutchman’. It is clear that the implication expressed here is neither logical,
definitional, nor causal. No ‘real connection’ is alleged to hold between what
the antecedent states and what is stated by the consequent. This sort of
conditional is ordinarily intended as an emphatic or humorous method of
denying the truth of its antecedent, for it typically contains a notoriously or
ridiculously false statement as its consequent. Any such assertion about truth
values is adequately symbolized by using the truth-functional connective 4 D *.
Although most conditional statements express more than a merely material
18 Arguments Containing Compound Statements [Ch. 2

implication between antecedent and consequent, we now propose to symbolize

any occurrence of ‘if-then* by the truth-functional connective ‘3\ Such
symbolizing abstracts from or ignores part of the meaning of most conditional
statements. But the proposal can be justified on the grounds that the validity of
valid arguments involving conditionals is preserved when the conditionals are
regarded as expressing material implications only, as will be established in the
following section.
Conditional statements can be expressed in a variety of ways. A statement of
the form ‘if p, then q’ could equally well be expressed as ‘if p, q\ *q if p\ ‘that p
implies that q\ ‘that p entails that q ‘p only if q\ ‘that p is a sufficient condition
that q\ or as ‘that q is a necessary condition that p\ and any of these formula¬
tions will be symbolized as p 3 q.

EXERCISES

I. If A and B are true statements and X and Y are false statements, which of the
following compound statements are true?

*1. X 3 (X 3 Y) 11. (X 3 A) 3 (~~X 3 -A)

2. (X DX)DY 12. (X 3 -Y) 3 (-X 3 Y)
3. (A 3 X) 3 Y 13. [(A-X) 3 Y] 3 (A 3 Y)
4. (X 3 A) 3 Y 14. [(A-B) 3 X] 3 [A 3 (B 3 X)]
*5. AD(BDY) *15. [(X-Y) 3 A] 3 [X 3 (Y 3 A))
6. AD (X 3 B) 16. [(A-X) 3 B] 3 [A 3 (B 3 X)]
7. (X DA)D(BDY) 17. [X 3 (A 3 Y)] 3 [(X 3 A) 3 Y]
8. (A 3 X) 3 (Y 3 B) 18. [X 3 (X 3 Y)] 3 [(X 3 X) 3 X]
9. (A 3 B) 3 (~~A 3 ~~B) 19. [(A 3«)3A]3A
MO. (X 3 Y) 3 (~~X 3 ~~Y) 20. [(X 3 Y) 3 XI 3 X

II. If A and B are known to be true, and X and Y are known to be false, but the truth
values of P and are not known, of which of the following statements can you
determine the truth values?

*1. X 3 (-Y 3 {)) 11. [(P D A) DP]D P

2. -Y 3 (P 3 -X) 12. [(P DX)DP\DP
3. (PDB)DY ~13. [(P 3 t>) 3 P] 3 P
4. (PvA) 3 (Q-X) ' 14. (F3X) 3 (X 3 P)
*5. (P 3 A) 3 (B 3 Y) *15. (<;) 3 B) 3 (A 3 Y)
6. (B 3 P) 3 (Q 3 A) 16. (PDQ)D (^Q 3 ~~P)
7. AD (PDA) 17. (P 3 ()) 3 {[P 3(()3 A)] 3 i[P 3 A)}
8. (X 3 Q) 3 X 18. (P D Q) 3 {[P D(QD X)] 3 (P 3 X)}
9. -~(P-X) 3 Y 19. (A 3 P) 3 {[A 3 (P 3 Q)\ 3 (A 3 Q)}
*10. (QvB)D X 20. (P 3 X) 3 {[P 3 (X 3 Q)] 3 (P 3 Q))

III. Symbolizing ‘Amherst wins its first game’ as A, ‘Colgate wins its first game* as C,
and ‘Dartmouth wins its first game’ as D, symbolize the following compound
statements:
Sfi£u24L Argument Forms and Truth Tables 19

*1. Both Amherst and Colgate win their first games only if Dartmouth does not
win its first game.
2. Amherst wins its first game if either Colgate wins its first game or Dartmouth
wins its first game.
3. If Amherst wins its first game, then both Colgate and Dartmouth win their
first games.
4. If Amherst wins its first game, then either Colgate or Dartmouth wins its first
game.
*5. If Amherst does not win its first game, then it is not the case that either
Colgate or Dartmouth wins its first game.
6. If it is not the case that both Amherst and Colgate win their first game, then
both Colgate and Dartmouth win their first games. *
7. If Amherst wins its first game, then not both Colgate and Dartmouth win
their first games.
8. If Amherst does not win its first game, then both Colgate and Dartmouth do
not win their first games.
9. Hither Amherst wins its first game and Colgate does not win its first game or,
if Colgate wins its first game, then Dartmouth does not win its first game.
*10. If Amherst wins its first game, then Colgate does not win its first game, but if
Colgate does not win its first game, then Dartmouth wins its first game.
11. If Amherst wins its first game, then, if Colgate does not win its first game,
then Dartmouth wins its first game.
12. Hither Amherst and Colgate win their first games or it is not the case that if
Colgate wins its first game, then Dartmouth wins it first game.
13. Amherst wins its first game only if either Colgate or Dartmouth does not win
its first game.
14. If Amherst wins its first game only if Colgate wins its first game, then
Dartmouth does not win it its first game.
15. If Amherst and Colgate both do not win their first games, then Amherst and
Colgate do, not both win their first games.

2.3 Argument Forms and Truth Tables

In this section, we develop a purely mechanical method for testing the

validity of arguments containing truth-functionally compound statements.
That method is closely related to the familiar technique of refutation by logical
analogy that was used in the first chapter to show the invalidity of the
argument

If I am President, then I am famous.

I am not President.
Therefore, I am not famous.

That argument was shown to be invalid by constructing another argument of

the same form:
20 Arguments Containing Compound Statements [Ch. 2

If Rockefeller is President, then Rockefeller is famous.

Rockefeller is not President.
Therefore, Rockefeller is not famous.

which is obviously invalid, since its premisses are true but its conclusion false.
Any argument is proved to be invalid if another argument of exactly the same
form can be constructed with true premisses and a false conclusion. This
reflects the fact that validity and invalidity are purely formal characteristics of
arguments: any two arguments having the same form are either both valid or
both invalid, regardless of any differences in their subject matter.4 The concept
of two arguments having exactly the same form is one that deserves further
examination.
It is convenient, in discussing forms of arguments, to use small letters from
the middle part of the alphabet, ‘p\ 7/’, V, V,... as statement variables, which
are defined simply to be letters for which, or in place of which, statements may
be substituted. Now, we define an argument form to be any array of symbols
that contains statement variables, such that when statements are substituted for
the statement variables—the same statement being substituted for every
occurrence of the same statement variable throughout—the result is an argu¬
ment. For definiteness, we establish the convention that in any argument form,
‘p’ shall be the first statement variable that occurs in it, 7/* shall be the second,
V the third, and so on.
Any argument which results from the substitution of statements for the
statement variables of an argument form is said to have that form, or to be a
substitution instance of that argument form. If we symbolize the simple
statement ‘The United Nations will become more responsible* as U, and the
simple statement ‘There will be a third world war’ as W, then the Disjunctive
Syllogism presented earlier can be symbolized as

UvW (1)

.*. w

It has the form -

pvq (2)
-p
..q

4 Here we assume that the simple* statements involved are neither logically true (e.g., ‘All
equilateral triangles are triangles') nor logically false (e.g., ‘Some triangles are nontriangles'). We
assume also that the only logical relations among the simple statements involved arc those
asserted or entailed by the premisses. The point of these restrictions is to limit our considerations
in Chapters 2 and 3 to truth-functional arguments alone, and to exclude other kinds of arguments
whose validity turns on more complex logical considerations to be introduced in Chapters 4 and 5.
Sec. 2.3] Argument Forms and Truth Tables 21

from which it results by substituting the statements U and W for the statement
variables p and </, respectively. But that is not the only form of which it is a
substitution instance. The same argument is obtained by substituting the
statements U vW, and W for the statement variables p, </, and r, respec¬
tively, in the argument form

V (3)
(/
r

We define the specific form of a given argument as that argument form from
which the argument results by substituting a different simple statement for
each distinct statement variable. Thus the specific form of the argument (1) is
the argument form (2). Although the argument form (3) is a form of the
argument (1), it is not the specific form of it. The technique of refutation by-
logical analogy can now be described more precisely. If the specific form of a
given argument can be shown to have any substitution instance with true
premisses and false conclusion, then the given argument is invalid.
The terms ‘valid’ and ‘invalid’ can be extended to apply to argument forms as
well as arguments. An invalid argument form is one that has at least one
substitution instance with true premisses and a false conclusion. The technique
of refutation by logical analogy presupposes that any argument of which the
specific form is an invalid argument form, is an invalid argument. Any argu¬
ment form is valid that is not invalid: So a valid argument form is one that has
no substitution instance with true premisses and false conclusion. Any given
argument can be proved valid if it can be shown that the specific form of the
given argument is a valid argument form.
To determine the validity or invalidity of an argument form, we must
examine all possible substitution instances of it to see if any of them have true
premisses and false conclusions. The arguments with which we are here
concerned contain only simple statements and truth-functional compounds of
them. We are interested only in the truth values of their premisses and
conclusions. We can obtain all possible substitution instances whose premisses
and conclusions have different truth values, by considering all possible ar¬
rangements of truth values for the statements that are substituted for the
distinct statement variables in the argument form to be tested. Possible
substitution instances can be set forth most conveniently in a truth table, with
an initial or guide column for each distinct statement variable that appears in
the argument form. Thus to prove the validity of the Disjunctive Syllogism
form

pvq
—p
:.q
22 Arguments Containing Compound Statements [Ch. 2

we construct the following truth table:

P q pyq .~p
T T T F
TFT F
F T T T
F F F T

Each row of this table represents a whole class of substitution instances. The
T’s and F’s in the two initial columns represent the truth values of statements
that can be substituted for the variables p and q in the argument form. These
determine the truth values in the other columns, the third of which is headed
by the first ‘premiss’ of the argument form, and the fourth by the second
‘premiss’. The second column’s heading is the ‘conclusion’ of the argument
form. An examination of this truth table reveals that whatever statements are
substituted for the variables p and </, the resulting argument cannot have true
premisses and a false conclusion. This is so because the third row represents the
only possible case in which both premisses are true, and there the conclusion is
true also.
Because truth tables provide a purely mechanical or effective method of
deciding the validity or invalidity of any argument of the general type here
considered, we can now justify our proposal to symbolize all conditional
statements by means of the truth-functional connective ‘ D ’. The justification
for treating all implications as though they were mere material implications is
that valid arguments containing conditional statements remain valid when
those conditionals are interpreted as expressing material implications only. The
three simplest and most intuitively valid forms of argument involving condi¬
tional statements are

Modus Ponens If p, then q

P
.\q

Modus Tollens If p, then q

.*. -p

and the

Hypothetical Syllogism If p, then q

If q, then r
. ‘. If p, then r

That they all remain valid when tliejir‘conditionals are interpreted as expressing
material implications is easily established by truth tables. The validity of
Sec. 2.3] Argument Forms and Truth Tables 23

Modus Ponens is shown by the same truth table that defines the horseshoe
symbol:

P_q P -> (I
T T T
T F F
FT T
F F T

Here the two premisses are represented by the third and first columns, and the
conclusion, by the second. Only the first row represents substitution instances
in which both premisses are true, and in that row the conclusion is true also.
The validity of Modus Pollens is shown by the truth table:

p_q_p D q _21!_ZR
T T T F F
T F F T F
FIT FT
F F T T T

Here only the fourth row represents substitution instances in which both
premisses (the third and fourth columns) are true, and there the conclusion (the
fifth column) is true also. Since the Hypothetical Syllogism form contains three
distinct statement variables, the truth table for it must have three initial
columns and will require eight rows for listing all possible substitution in¬
stances:

p_ '/ r p D q (j D r p J r
T T T T T T
T T F T F F
T F T F T T
T F F F T F
F T T T T T
F T F T F T
F F T T T T
F F F T T T

In constructing it, the three initial columns represent all possible arrangements
of truth values for the statements substituted for the statement variables p, c/,
and r. The fourth column is filled in by reference to the first and second, the
fifth by reference to the second and third, and the sixth by reference to the first
and third. The premisses are both true only in the first, fifth, seventh, and
eighth rows, and in these rows the conclusion is true also. This suffices to
demonstrate that the Hypothetical Syllogism remains valid when its con-
24 Arguments Containing Compound Statements [Ch. 2

ditionals are symbolized by means of the horseshoe symbol. Any doubts that
remain about the claim that valid arguments containing conditionals remain
valid when their conditionals are interpreted as expressing merely material
implication, can be allayed by the reader’s providing, symbolizing, and testing
his own examples by means of truth tables.
To test the validity of an argument form by a truth table requires a table
with a separate initial or guide column for each different statement variable,
and a separate row for every possible assignment of truth values to the
statement variables involved. Hence, testing an argument form containing n
distinct statement variables, requires a truth table having 2n rows. In con¬
structing truth tables, it is convenient to fix upon some uniform pattern for
inscribing the T’s and F’s in their initial or guide columns. In this book we shall
follow the practice of simply alternating T’s and F’s down the extreme
right-hand initial column and alternating pairs of T’s with pairs of F’s down the
column directly to its left. Next, we shall alternate quadruples of T’s with
quadruples of F’s,. .., and finally, we shall fill in the top half of the extreme
left-hand initial column with T’s and its bottom half with F’s.
There are two invalid argument forms that bear a superficial resemblance to
the valid argument forms Modus Ponens and Modus Pollens. These are

pDq pDq
q and —p
.'.p

and are known as the Fallacies of Affirming the Consequent and of Denying the
Antecedent, respectively. The invalidity of both can be shown by a single truth
table:

p q pD q ~~p -</
T T T F F
T F F F T
F T T T F
F F T T T

The two premisses in the Fallacy of Affirming the Consequent head the second
and third columns, and are true in both the first and third rows. But the
conclusion, which heads the first column, is false in the third row—which
shows that the argument form does have a substitution instance with true
premisses and a false conclusion, and is therefore invalid. Columns three and
four are headed by the two premisses in the Fallacy of Denying the Anteced¬
ent, which are true in both the third and fourth rows. Its conclusion heads the
fifth column and is false in the third row—which shows that the second
argument form is invalid also.
It must be emphasized that althpiigh a valid argument form has only valid
arguments as substitution instances, an invalid argument form can have both
Sec. 2.3] Argument Forms end Truth Tobies 25

valid and invalid substitution instances. So to prove that a given argument is

invalid, we must prove that the specific form of that argument is invalid.

EXERCISES

I. For each of the following arguments indicate which, if any, of the argument forms
in Exercise 11 below have the given argument as a substitution instance, and
indicate which, if any, is the specific form of the given argument:
*a. A f. M 3 (N 3 O) k. (A D B)-(C D D)
.'. A vB O 3 ~~\1 A vC
O 3 -N WBvD
b. (>D g. (P 3 QHR 3 S) I. (EDF)-(GDII)
:.c :.p 3 q ^Fv ^G
-iiv
c. E 3 (F-G) h. r 3 u m. 1 D J
-(F-G) 3 *E .'. (T 3 U) v ( V’T) D7) 3(1 37)
d. H i. W 3 X n. A' 3 (A 3 M)
1 . .X 3 (W 3 X) K 3 A
.'.H-I .’.A 3 M
*e. JD(K-L) *j. yv(z*—Y) o. /V 3 (N 3 0)
Jv(K-L) y A7 3 N
.’. K-L ~(2—n /V 3 O

II. Use truth tables to determine the validity or invalidity of each of the following
argument forms:

*1. p-q 8. p 3 q *15. (p 3 f/)*(p 3 r)

:.p P
.‘. (/vr
2. p 9. /> 3 (</**•) 16. p 3 (</ v /•)
.'.p-q ~(<Tr) 3 ~P P 3 ~<f
. pv r
3. p v <j *10. pi/ 17. (p 3 r/)-(r 3 v)
■ ■P P pvr
~t/ . qvs
4. p n. p 18. (p 3 r/)-(r 3 .v)
.'.pvq </ ~^q v ~,v
v —r

*5. p 12. p 3 q 19. (pvq) 3 (pv/)

.'.p Dq q Dp p-q
.'.pvq .'.pvq
6. p ‘ 13. p D q 20. pv(q-^p)
:.q^p PV(t V
•• C1 :. ~~(q--~p)
7. pDq 14. p D (q D r) 21. (pvr/) 3 (p-q)

• • ~P pDq Mpvq)
:.pDr ~(p*q)
26 Arguments Containing Compound Statements [Ch. 2

III. Use truth tables to determine the validity or invalidity of each of the following
arguments:

*1. If Alice is elected class president, then either Betty is elected vice-president
or Carol is elected treasurer. Betty is elected vice-president. Therefore, if
Alice is elected class-president, then Carol is not elected treasurer.
2. If Alice is elected class president, then either Betty is elected vice-president
or Carol is elected treasurer. Carol is not elected treasurer. Therefore, if
Betty is not elected vice-president, then Alice is not elected class president.
3. If Alice is elected class president, then Betty is elected vice-president and
Carol is elected treasurer. Betty is not elected vice-president. Therefore,
Alice is not elected class president.
4. If AJice is elected class president, then if Betty is elected vice-president,
then Carol is elected treasurer. Betty is not elected vice-president. There¬
fore, either Alice is elected class president or Carol is elected treasurer.
*5. If the seed catalog is correct, then if the seeds are planted in April, then the
flowers bloom in July. The flowers do not bloom in July. Therefore, if the
seeds are planted in April, then the seed catalog is not correct.
6. If the seed catalog is correct, then if the seeds are planted in April, then the
flowers bloom in July. The flowers bloom in July. Therefore, if the seed
catalog is correct, then the seeds are planted in April.
7. If the seed catalog is correct, then if the seeds are planted in April, then the
flowers bloom in July. The seeds are planted in April. Therefore, if the
flowers do not bloom in July, then the seed catalog is not correct.
8. If the seed catalog is correct, then if the seeds are planted in April, then tin*
flowers bloom in July. The flowers do not bloom in July. Therefore, if the
seeds are not planted in April, then the seed catalog is not correct.
9. If Ed wins first prize, then Fred wins second prize, and if Fred wins second
prize, then George is disappointed. Either Ed wins first prize or George is
disappointed. Therefore, Fred docs not win second prize. .
*10. If Ed wins first prize, then either Fred wins second prize or George is
disappointed. Fred does not win second prize. Therefore, if George is
disappointed, then Ed does not win first prize.
11. If Ed wins first prize, then Fred wins second prize, and if Fred wins second
prize, then George is disappointed. Either Fred does not win second prize or
George is not disappointed. Therefore, Ed does not win first prize.
12. If Ed wins first prize, then Pred wins second prize, and if Fred wins second
prize, then George is disappointed. Either Ed does not win first prize or Fred
does not win second prize. Therefore, either Fred does not win second prize
or George is not disappointed.
13. If the weather is warm and the sky is clear, then we go swimming and we go
boating. It is not the case that if the sky is clear, then we go swimming.
Therefore, the weather is not warm.
14. If the weather is warm and the sky is clear, then cither we go swimming or
we go boating. It is not the case that if the sky is clear, then we go
swimming. Therefore, if we do riot go boating, then the weather is not warm.
*15. If the weather is warm and the sky is clear, then either we go swimming or
we go boating. It is not the case’that if we do not go swimming, then the sky
is not clear. Therefore, either the weather is warm or we go boating.
Sec. 2.4] Statement Forms 27

2.4 Statement Forms

The introduction of statement variables in the preceding section enabled us

to define both argument forms in general and the specific form of a given
argument. Now, we define a statement form to be any sequence of symbols
containing statement variables, such that when statements are substituted for
the statement variables—the same statement being substituted for every
occurrence of the same statement variable throughout—the result is a state¬
ment. Again for definiteness, we establish the convention that in any statement
form, p’ shall be the first statement variable that occurs in it. The statement
variable \f shall be the second, V the third, and so on. Any statement which
results from substituting statements for the statement variables of a statement
form is said to have that form, or to be a substitution instance of it. Just as we
distinguished the specific form of a given argument, so we distinguish the
specific form of a given statement as that statement form from which the given
statement results by substituting a different simple statement for each distinct
statement variable. For example, where A, B, and C are different simple
statements, the compound statement A D {B v C) is a substitution instance of
the statement form p D </, and also of the statement form p D (q v r), but only
the latter is the specific form of the given statement.
Although the statements ‘Balboa discovered the Pacific Ocean’ (B) and
‘Balboa discovered the Pacific Ocean or else he didn’t’ (Bv —B) are both true,
we discover their truth in quite different ways. The truth of B is a matter of
history, and must be learned through empirical investigation. Moreover, events
might possibly have been such as to make B false; there is nothing necessary
about the truth of B. But the truth of the statement B v — B can be known
independently of empirical investigation, and no events could possibly have
made it false, for it is a necessary truth. The statement B v ~~B is a formal
truth, a substitution instance of a statement form all of whose substitution
instances are true. A statement form that has only true substitution instances is
said to be tautologous, or a tautology. The specific form of B v — B is p v
and is proved a tautology by the following truth table:

p_21}_E12:E
T F T
F T T

That there are only T’s in the column headed by the statement form in question
shows that all of its substitution instances are true. Any statement that is a
substitution instance of a tautologous statement form is formally true and is,
itself, said to be tautologous or a tautology.
Similarly, although the statements ‘Cortez discovered the Pacific’ (C) and
‘Cortez discovered the Pacific and Cortez did not discover the Pacific’ (C*
are both false, we discover their falsehood in quite different ways. The first
28 Arguments Containing Compound Statements [Ch. 2

simply happens to be false, and that must be learned empirically; the second is
necessarily false, and that can be known independently of empirical investiga¬
tion. The statement Cm~~C is formally false, a ‘.substitution instance of a
statement form, all of whose substitution instances are false. One statement is
said to contradict, or to be a contradiction of, another statement when it is
logically impossible for them both to be true. In this sense, contradiction is a
relation between statements. But there is another, related sense of that term.
When it is logically impossible for a particular statement to be true, that
statement itself is said to be self-contradictory, or a self-contradiction. Such
statements are also said more simply to be contradictory or contradictions. We
shall follow the latter usage here. A statement form that has only false substi¬
tution instances is said to be contradictory or a contradiction, and the same
terms are applied to its substitution instances. The statement form p • — p is
proved a contradiction by the fact that in its truth table only F's occur in the
column that it heads.
Statements and statement forms that are neither tautologous nor contradic¬
tory are said to be contingent or contingencies. For example, p, —p, p v ry, p‘(j,
and pD(/ are contingent statement forms; and B, C, —C, B*C, Bv C are
contingent statements. The term is appropriate, since their truth values are not
formally determined but are dependent or contingent upon what happens to be
the case.
It is easily proved that p D (q D p) and ~^p D (p D q) are tautologies.
When expressed in English as ‘A true statement is implied by any statement
whatever', and as ‘A false statement implies any statement whatever, they
seem rather strange. They have been called by some writers the paradoxes of
material implication. When it is kept in mind that the horseshoe symbol is a
truth-functional connective that stands for material implication rather than
either ‘implication in general’ or more usual kinds of implications such as
logical or causal, then the tautologous statement forms in question are not at all
surprising. And when the misleading English formulations are corrected by
inserting the word ‘materially’ before ‘implied’ and ‘implies’, then the air of
paradox vanishes. Material implication is a special, technical concept that the
logician introduces and uses because it simplifies his task of discriminating valid
from invalid arguments.
Two statements are said to be materially equivalent when they have the
same truth value, and we symbolize the statement that they are materially
equivalent by inserting the symbol ‘=’ between them. Being a truth-functional
connective, the three-bar symbol is defined by the following truth table:

P <1 P = <1
T T T
T . F F
F T F
F F T
Sec. 2.4] Statement Forms 29

To say that two statements are materially equivalent, is to say that they
materially imply each other, as is easily verified by a truth table. Hence the
three-bar symbol may be read ‘if and only if . A statement or statement form of
the pattern p == q is called a biconditional. Two statements are said to be
logically equivalent when the biconditional that expresses their material
equivalence is a tautology. The ‘principle of Double Negation’, expressed as
p = ~~~~p, is proved to be tautologous by a truth table.
There are two logical equivalences that express important interrelations of
conjunctions, disjunctions, and negations. Because a conjunction asserts that
both its conjuncts are true, its negation need assert only that at least one is
false. Thus negating the conjunction p*q amounts to asserting the disjunction
of the negations of p and q. This statement of equivalence is symbolized as
~~(p*<j) EH (~~p v — (f), and proved to be a tautology by the following truth
table:

P <1 P’<l ~~(p'q) ~~P ~</ ~~<l ~~(p-q) = (~~pv —q)

TTTFFF F T
T F F T F T T T
FTFTTF T T
F F F T T T T T

Similarly, because a disjunction asserts merely that at least one disjunct is true,
to negate it is to assert that both are false. Negating the disjunction pvq
amounts to asserting the conjunction of the negations of p and q. It is symbol¬
ized as ~~(p v (/) ^ p • — q) and is easily proved tautologous by a truth table.
These two equivalences are known as De Morgan’s 'Fheorems, after the English
mathematician-logician Augustus De Morgan (1806-1871) and can be stated
compendiously in English as:

The negation of the |di^unction1 } two slatei,lcnts is logically equiva¬

lent to the |disiunct*on | 0f their negations.

I conjunction)

Two statement forms are said to be logically equivalent if no matter what

statements are substituted for their statement variables—the same statement
being substituted for the same statement variable in both statement forms—the
resulting pairs of statements are materially equivalent. Thus De Morgan’s
Theorem asserts that — (p v q) and ~~pu~~q are logically equivalent statement
forms. By De Morgan’s Theorem and the principle of Double Negation
~^(p'~*~q) and —p v q are logically equivalent, hence either can be taken as
defining p D q. The second alternative is the more usual choice.
To every argument corresponds a conditional statement whose antecedent is
the conjunction of the argument’s premisses and whose consequent is the
argument’s conclusion. That corresponding conditional is a tautology if and
30 Arguments Containing Compound Statements [Ch. 2

only if the argument is valid. Thus to the valid argument form

pvq
-p
.'.q

corresponds the tautologous statement form [(pvr/)-~~p] D q; and to the

invalid argument form

pDq
<1
• *P

corresponds the nontautologous statement form [(p D q)mq) D p. An argument

form is valid if and only if its truth table has a T under its conclusion in every
row in which there are T’s under all of its premisses. Since an F can occur in
the column headed by its corresponding conditional statement only where
there are T’s under all of those premisses and an F under the conclusion, it is
clear that there can be only T’s under a conditional that corresponds to a valid
argument form. If an argument is valid, the statement that the conjunction of
its premisses implies its conclusion is a tautology.
An alternative version of the truth table test of a statement form is the
following, which corresponds to the preceding truth table.

2:_(P_ q) =_hi_E._y___22_ </)

FTTTT FTFF T
TTFFTFTTT F
TFFTTTFTF T

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Here columns (2), (4), (7), (10) are the initial or guide columns. Column (3) is
filled in by reference to columns^) and (4), and column (1) by reference to
column (3). Column (6) is filled in by reference to column (7), column (9) is
filled in by reference to column (10), and then column (8) by reference to
columns (6) and (9). Finally, column (5) is filled in by reference to columns (1)
and (8). That its main connective has only T’s under it in the truth table
establishes that the statement form being tested is a tautology.

EXERCISES

I. Use truth tables to characterize the following statement forms as tautologous,

contradictory, or contingent:

*1- PD ~~P 3. p D (p D p)
4. (p Dp) Dp
*
2. (p D —p)-(—p D p)
 Statement Forms 31
Sec. 2.4]
8. (-“P’tfHfl 3 P)
*5. p 3 (P'P)
9. [(p D q) 3 <f] 3 9
6. (p-ffP P
10. [(p 3 </) 3 Pi D P
7.

n use truth tables to decide which of the following are logical equivalences:

6. [pv(qT)]=[(pvq)T]
*1. (p 3 q) = (~P D ~(/)
7. [p v (</-r)] — [(p v9)*(pvr)]
2. (p D<j) = (~q 3 ~p)
8. (p = <?) = [(p,<?)v(~P‘~‘J)l
3. [(p-q)Dr]=[p3(q 3r)l
9. p = [p-(p 3<f)]
4. [pD(qDr)l=[(p3</P»']
‘5. [p-(qvr)] = [(p*<j)v(p*r)l 10. p = lp*(9°P)]
The Method of Deduction

3.1 Formal Proof of Validity

If arguments contain more than two or three different simple statements as

components, it is cumbersome and tedious to use truth tables to test their
validity. A more convenient method of establishing the validity of some
arguments is to deduce their conclusions from their premisses by a sequence of
shorter, more elementary arguments that are already known to be valid.
Consider, for example, the following argument, in which five different simple
statements occur:

Either the Attorney General has imposed a strict censorship or if Black

mailed the letter then Davis received a warning.
If our lines of communication have not broken down completely, then if
Davis received a warning, then Emory was informed about the matter.
If the Attorney General has imposed a strict censorship, then our lines of
communication have broken down completely.
Our lines of communication have not broken down completely.
Therefore, if Black mailed the letter, then Emory was informed about the
matter.

It may be translated into our symbolism as

Av (B D D)
D (D D E)
ADC
~~C

.\B D R

To establish the validity of this argument by means of a truth table would

require a table with thirty-two rows.* We can prove the given argument valid,
however, by deducing its conclusion from its premisses by a sequence of just
four arguments whose validity has already been established. From the third and
fourth premisses, ADC and —C, we validly infer by Modus Tottens. From

32
Sec. 3.1] Formal Proof of Validity 33

~~A and the first premiss, A v (B DD),we validly infer B D D by a Disjunctive

Syllogism. From the second and fourth premisses, D (D D E) and we
validly infer DDK by Modus Ponens. And finally, from these last two conclu¬
sions (or subconclusions), B D D and D D Ef we validly infer B D E by a
Hypothetical Syllogism. That its conclusion can be deduced from its premisses
using valid arguments exclusively, proves the original argument to be valid.
Here the elementary valid argument forms Modus Ponens (M.P.), Modus
Tollens (M.T.), Disjunctive Syllogism (D.S.), and Hypothetical Syllogism (H.S.)
are used as Rules of Inference by which conclusions are validly deduced from
premisses.
A more formal and more concise way of writing out this proof of validity is
to list the premisses and the statements deduced from them in one column,
with “justifications” for the latter written beside them. In each case the
“justification” for a statement specifies the preceding statements from which,
and the Rule of Inference by which, the statement in question was deduced. It
is convenient to put the conclusion to the right of the last premiss, separated
from it by a slanting line which automatically marks all of the statements above
it to be premisses. The formal proof of validity for the given argument can be
written as

1. Av(BDD)
2. D(DD E)
3. A D C
4. -C /.'.BDE
5. -A 3, 4, M.T.
6. BDD 1,5, D.S.
7. DDE 2, 4, M.P.
8. BDE ,
6 7, H.S.

A formal proof of validity for a given argument is defined to be a sequence of

statements, each of which is either a premiss of that argument or follows from
preceding statements by an elementary valid argument, and such that the last
statement in the sequence is the conclusion of the argument whose validity is
being proved. This definition must be completed and made definite by speci¬
fying what is to count as an ‘elementary valid argument’. We first define an
elementarxj valid argument as any argument that is a substitution instance of an
elementary valid argument form. Then, we present a list of just nine argument
forms that are sufficiently obvious to be regarded as elementary valid argument
forms and accepted as Rules of Inference.
One matter to be emphasized is that any substitution instance of an ele¬
mentary valid argument form is an elementary valid argument. Thus the
argument

—CD(DD E)

.’. D D E
34 The Method of Deduction [Ch. 3

is an elementary valid argument because it is a substitution instance of the

elementary valid argument form Modus Ponens (M.P.). It results from

pD </
p
:.q
i. . .-A.

by substituting for p and DDE for q; therefore, it is of that form even

though Modus Ponens is not the specific form of the given argument.
We begin our development of the method of deduction by presenting a list
of just nine elementary valid argument forms that can be used in constructing
formal proofs of validity:

RULES OF INFERENCE

1. Modus Ponens (M.P.) 6. Destructive Dilemma (D.D.

pDq (P q)‘(r D s)
P —q v ~s
.'.q . '. ~~p v —r
2. Modus Pollens (M.T.) 7. Simplification (Simp.)
pDq vq
~q ■'■p
~P
3. Hypothetical Syllogism (ll.S.) 8. Conjunction (Conj.) J*
pDq P
q D r </
.'.pH r :.p-q
4. Disjunctive Syllogism (D.S.) 9. Addition (Add.)< /
p\q P
~~P ..pvq
.'.q
5. Constructive Dilemma (C.D.)
(pDq)-(rDs)
p\ r
.'.qvs

These nine Rules of Inference are elementary valid argument forms, whose
validity is easily established by truth tables. They can be used to construct
formal proofs of validity for a wide range of more complicated arguments. The
names listed are standard for the most part, and the use of their abbreviations
permits formal proofs to be set down with a minimum of writing.

EXERCISES

I. For each of the following arguments, state the Rule of Inference by which its
conclusion follows from its premiss or premisses:
Sec. 3.1] Formal Proof of Validity 35

*1. (A D —B)*(~C D D) 9. (F 3 —G) 3 (-f/v -/)

.'.A D F 3 -G

2. E D ~E *10. [-(/•*) 3 -L]-(Af 3 -dV)

(E 3 «~F) v(~~G 3 //) —(/•It) v M
-Lv -iV
3. (/= —/)•(!= —7) 11. O 3 -P
-P 3 Q
.’.(O 3 -P)-(-F 3 ^))
4. Kv(LvM) 12. (-R=S)v(rv G)
. [Kv(LvM)\v[K v (Ev A/)]
Pv G
*5. /V 3 (O ~ -F) 13. [(V--W) 3 A]-l(W--V) 3 Z1
(0=-F) D() (v—vvr) v(w--y)
N 3 p . *. X v Z
6. (R == -S) 3 (T 3 V) 14. [A 3 (flvC)] 3 [(/>!') IE -F]
R = -S H(/>£)--F]
T 3 U 3 (11 v C)]
7. (V 3*W) v (X 3 V) 15. HG3(Wv/)]-[(/.K) 3L]
—(V 3 W) ■-HG3(//v/)]
.-.x d y
(A 3 ~/?)-[CV 3 (D-F)l
—Bv-(D-E)
-Av-C

11. Each of the following is a formal proof of validity for the indicated argument.
State the ‘justification’ for each line that is not a premiss:

1. "1. (ArR) 3 [A 3 (D'E 4. (—JVv -L)*(-Mv-())

2. (A-H)-C /.'.DvE /.*. (-Ov -P)'(-Mv -X)
. 3. A -R 5. (A/ 3 Ar)-(F 3 L)
4. A 3 (D-F) 6. ~~Nv~~E
5. A 7. v
6. D-E 8. —O v ~~P
7. D 9. (—Ov
y H. DvE 4. 1. 0 3(H3 S)
2. 1. Fv(GvH) 2. (R 3 S) 3 T
2. (G 3 /)•(// 3 7) 3. (S- G) 3 —V
3. (IvJ) D (FvH) 4. -V 3 (R EE ~~W)
4. -F /.’.// 5. -Tv ~~(R = -W)
5. Gv// ' —(S-C/)
6. /V/ 6. <? 3 T
7. Fv// 7. (S-C7) 3 (R= -W)
8. // 8. (p 3 7>[(S-G) 3 (R= -W)]

3. 1. K 3 L
9. ^v-iSd/)
2. MD N *5. 1. (-Xv -y) 3 [A 3 (P—^)]
3. (0 3 A1)-(F 3 L) 2. (-X--R) 3 [(F—p) 3 Z]
36 The Method of Deduction [Ch. 3

3. (-X-fl)*(-ZvA) 4. (Tv -V) 3 (W-X)

//.A 3 Z 5. Qv~~Q /.*. W-X
4. ~~X-~~R 6. -PvR
5. (P-~~<?) 3 Z 7. Sv-G
6. -X 8. Tv -V
7. -xv-y 9. W-X
8. AD(F-Q) 9. 1. [(A v v C] 3 [D 3 (E = F)]
9. ADZ 2. (A v ^B) 3 [(F = G) D II]
6. 1. A 3 B 3. A D [(E=F) D (FSC)]
2. C 3 D 4. A . D D II
3. -Bv-D 5. A v —B
4. 6. (A v — B) v C
5. (E-F) 3 C 7. D 3 (E==F)
/.'.-(E-F) 8. (E=F) 3 (F=G)
6. (A D R)-(C 3 O) 9. D 3 (F=G)
7. -Av-C 10. (FEE G) 3 II
8. -C 11. D D H
9. -(£•/♦) MO. HD (ID J)
7. 1. (G 3 //) 3 (7 = 7) 2. KD (IDJ)
2. X v —(L 3 A/) 3. (-//--X) 3 (—L v -Af)
3. (G 3 //) v ~~X 4. (—L 3 ~~N)-(~~M 3 -O)
4. JV 3 (X 3 M) 5. (P 3 N)-(p 3 O)
5. -(/ = J) 6. -(/ D/) /.*. -Pv-p
6. -(G 3 H) 7. -/7
7. -X 8. -X
8. -(/' 3 M) 9. —7/-—X
9. -N 10. —X v —Af
8. i. (O d -P)-(-y 3 R) 11. —A7v—O
2. (S 3 P)-(~G 3 -V) 12. -Pv-p
3. (~~P 3 S)-(R 3 ~~U)

111. Construct a formal proof of validity for each of the following arguments

*1. A 3 B 4. A7 3 2V
C D D iV 3 O
(-Bv-D)-(-Av^) (A7 3 O) D(N DP)
-Av-C (M DP) DQ
■■Q
v/ 2. F 3 (F- —G) *5. (fl D ~S)"(r 3 ~t/)
(FvG) 3 1/ (V D ~W)*(X D ~Y)
E (7’ D W)-(l/ D S)
Vvrt
—7'v ~(/
3. 7 3 X 6. A 3 (B-C)
7v(Xv-Z.) ~A D [(D 3 E)-(F 3 G)]
-X (B*C) v [(~A 3 D)-(~A 3 F)]
.\-/,-~~X
..EvC,
Sec. 3.1] Formal Proof of Validity 37

7. (~/f v/) 3 (/ 3 K) 9. V D W
(~~L—M) 3 (K 3 iV) xdy
(H 3 L) • (L 3 //) Z D W
—O X D /I
:.] d n W D X
[(V D Y)-(ZDA)} D (VvZ)
.'. Y v A
8. (P 3 Q)-(R 3 S) 10. (BvC) D (Dv/{)
(() D T)-(S D U) [(/)v/i)vf’J D (G v //)
(C vi/) D - D
—T K D ~~c;

IV. Construct a formal proof of validity for each of the following arguments, using the
abbreviations suggested:

*1. If cither algebra is required or geometry is required, then all students will
study mathematics. Algebra is required and trigonometry is required.
Therefore, all students will study mathematics. (A: Algebra is required.
C: Geometry is required. S: All students will study mathematics. T: Trigo¬
nometry is required.)
2. Either Smith attended the meeting or Smith was not invited to the meeting.
If the directors wanted Smith at the meeting, then Smith was invited to the
meeting. Smith did not attend the meeting. If the directors did not want
Smith at the meeting and Smith was not invited to the meeting, then Smith
is on his way out of the company. Therefore, Smith is on his way out of the
company. (A: Smith attended the meeting. /: Smith was invited to the
meeting. D< The directors wanted Smith at the meeting. W: Smith is on his
way out of the company.)
3. If a scarcity of commodities develops, then prices rise. If there is a change of
administration, then fiscal controls will not be continued. If die threat of
inflation persists, then fiscal controls will be continued. If there is overpro¬
duction, then prices do not rise. Either there is overproduction or there is a
change of administration. Therefore, either a scarcity of commodities does
not develop or the threat of inflation does not persist. fS: A scarcity of
commodities develops. P: Prices rise. C: There is a change of administration.
F: Fiscal controls will be continued. 1: The threat of inflation persists.
O: There is overproduction.)
4. If the investigation continues, then new evidence is brought to light. If new
evidence is brought to light, then several leading citizens are implicated. If
several leading citizens are implicated, then the newspapers stop publiciz¬
ing the case. If continuation of the investigation implies that the news¬
papers slop publicizing the case, then the bringing to ligfit of new evidence
implies that the investigation continues. The investigation does not con¬
tinue. Therefore, new evidence is not brought to light. (C: The investigation
continues. N: New evidence is brought to light. /: Several leading citizens
are implicated. S: The newspapers stop publicizing the case.)
 [Ch. 3

*5. If the king does not castle and the pawn advances, then either the bishop is
blocked or the rook is pinned. If the king does not castle, then if the bishop
is blocked, then the game is a draw. Either the king castles or if the rook is
pinned, then the exchange is lost. The king'does not castle and the pawn
advances. Therefore, either the game is a draw or the exchange is lost.
(K: The king castles. P: The pawn advances. B: The bishop is blocked.
R: The rook is pinned. D: The game is a draw. E: The exchange is lost.)
6. If Andrews is present, then Brown is present, and if Brown is present, then
Cohen is not present. If Cohen is present, then Davis is not present. If Brown
is present, then Emerson is present. If Davis is not present, then Farley is
present. Either Emerson is not present or Farley is not present. Therefore,
either Andrews is not present or Cohen is not present. (A: Andrews is
present. B: Brown is present. C: Cohen is present. D: Davis is present.
E: Emerson is present. F: Farley is present.)
7. If either George enrolls or Harry enrolls, then Ira does not enroll. Either Ir$
enrolls or Harry enrolls. If either Harry enrolls or George does not enroll,
then Jim enrolls. George enrolls. Therefore, either Jim enrolls or Harry does
not enroll. (GT: George enrolls. H: Harry enrolls. I: Ira enrolls. /: Jim enrolls.)
8. If Tom received the message, then Tom took the plane, but if Tom did not
take the plane, then Tom missed the meeting. If Tom missed the meeting,
then Dave was elected to the board, but if Dave was elected to the board,
then Tom received the message. If either Tom did not miss the meeting or
Tom did not receive the message, then either Tom did not take the plane or
Dave was not elected to the board. Tom did not miss the meeting. Therefore,
either Tom did not receive the message or Tom did not miss the meeting.
(R: Tom received the message. P: Tom took the plane. M: Tom missed the
meeting. D: Dave was elected to the board.)
9. If Dick was recently vaccinated, then he has a fever. Either Dick was
recently vaccinated, or if pocks begin to appear then Dick must be quaran¬
tined. Either Dick has measles or if a rash develops, then there are compli¬
cations. If Dick has measles, then he has a fever. If Dick was not recently
vaccinated and Dick does not have measles, then either a rash develops or
pocks begin to appear. Dick does not have a fever. Therefore, either there
are complications or Dick must be quarantined. (V: Dick was recently
vaccinated. F: Dick has a fever. P: Pocks begin to appear. £>: Dick must be
quarantined. M: Dick has measles. R: A rash develops. C: There are
complications.) **
*10. Either taxes are increased or if expenditures rise, then the debt ceiling is
raised. If taxes are increased, then the cost of collecting taxes increases. If a
rise in expenditures implies that the government borrows more money, then
if the debt ceiling is raised, then interest rates increase. If taxes are not
increased and the cost of collecting taxes does not increase, then if the debt
ceiling is raised, then the government borrows more money. The cost of
collecting taxes does not increase. Either interest rates do not increase or the
government does not borrow more money. Therefore, either the debt ceiling
is not raised or expenditures do not rise. (T: Taxes are increased. E: Expen¬
ditures rise. D: The debt, ceiling is raised. C: The cost of collecting taxes
increases. G: The government borrows more money. I: Interest rates
increase.)
Sec. 3.2] The Rule of Replacement 39

3.2 The Rule of Replacement

There are many valid truth-functional arguments that cannot be proved

valid using only the nine Rules of Inference that have been given thus far. For
example, a formal proof of validity for the obviously valid argument

A-R
B

requires additional Rules of Inference.

Now the only compound statements that concern us here are truth-func¬
tional compound statements. Hence if any part of a compound statement is
replaced by an expression that is logically equivalent to the part replaced, the
truth value of the resulting statement is the same as that of the original
statement. This is sometimes called the Rule of Replacement and sometimes
the Principle of Kxtcnsionality.1 We adopt the Rule of Replacement as an
additional principle of inference. It permits us to infer from any statement the
result of replacing all or part of that statement by any other statement logically
equivalent to the part replaced. Thus, using the Principle of Double Negation
(D.N.), which asserts the logical equivalence of p and--p, we can infer from
A D ——B any of the statements

ADR,-A D —R, A D-R, or-(A D-R)

by the Rule of Replacement.

To make this new rule definite, we list a number of logical equivalences with
which it can be'used. These equivalences constitute additional Rules of
Inference that can be used in proving the validity of arguments. We number
them consecutively after the first nine Rules already stated.

. Rule of Replacement: Any of the following logically equivalent expressions

ian replace each other wherever they occur:

10. De Morgan’s Theorems (De M.): ~^(P'<-j) = (~~p v

-(pvf/) = (-p—f/).
11. Commutation (Com.): (p v(/)= (q v p).
(p-q)=(q-p).
12. Association (Assoc.): [p v (q v r)] = [(p v q) v r].
[p-(q-r)] = [(p-q)-r].
13. Distribution (Dist.): [p*{q v r)] = [(p-q) v (p-r)].
[pv (9•!■)]= [(pvq)-(pvr)].

*It will be stated more formally in an appropriate context and demonstrated in Chapter 8.
40 The Method of Deduction [Ch. 3

14. Double Negation (D.N.): p = ^—p.

15. Transposition (Trans.): (p D q) = (~~q D ^p).
16. Material Implication (Impl.): (p*D q) = p v q).
17. Material Equivalence (Equiv.): (p = q) = [(p D q) • (q D p)].
(p = q) = [(p-q) v (~p—(/)]
18. Exportation (Exp.): [(p*q) 3 r] = [p D (q D r)].
19. Tautology (Taut.): p = (pvp). ,
p=(p-p).

A formal proof of validity for tlie argument given on page 39 can now Ire
written:

1. A-B /.\fl
2. B#A 1, Com.
3. B 2, Simp.

Some argument forms, although very elementary and perfectly valid, are not
included in our list of nineteen Rules of Inference. Although the argument

A-B
B

is obviously valid, its form

p-q

is not included in our list. Hence B does not follow from A-B by any single
elementary valid argument form, as defined by our list. It can be deduced,
however, by using two elementary valid arguments, as shown above. We could
add the intuitively valid argument form

- p-q

to our list, of course, but if we expanded our list in this way we might end up
with a list that was too long and, therefore, unmanageable.
The list of Rules of Inference contains several redundancies. For example.
Modus Tollens could be dropped from our list without any real weakening of
the machinery. This is so because any step deduced by its use could be deduced
by using other Rules of the list instead. Thus in our first proof, on page 33, line
5, ^A, which was deduced from lines 3 and 4, A D C and by Modus
Tollens, could have been deduced without it, since D follows from
A D Cby Transposition, and ^A from D —A and —Cby Modus Ponens.
Sec. 3.2] The Rule of Replacement 41

Modus Tollens is so common and intuitive a principle of inference, however,

that it has been included anyway. Others have also been included for conven¬
ience, despite their logical redundancy.
The test of whether or not a given sequence of statements is a formal proof,
is effective. That is, direct observation will suffice to decide, of every line
beyond the premisses, whether or not it actually does follow from preceding
lines by one of the given Rules of Inference. No ‘thinking* is required: neither
in thinking about what the statements mean, nor in using logical intuition to
check the validity of any line*s deduction. Even where the ‘justification* of a
statement is not written beside it, there is a finite, mechanical procedure for
deciding whether or not the deduction is legitimate. Each line is preceded by
only a finite number of other lines and only a finite number of Rules of
Inference have been adopted. Although time consuming, it can be verified by
inspection whether the line in question follows from any single preceding line
or from any pair of preceding lines by any one of the Rules of Inference listed.
For example, in the foregoing proof, line 2, BrA, is preceded only by line 1,
AmB. Its legitimacy can be decided by observing that although it does not
follow from A • B by Modus Ponens, nor by Modus Tollens, nor by a Hypothet¬
ical Syllogism, and so on through number 10, when we come to number 11, we
can see, simply by looking at their forms, that line 2 follows from line 1 by the
principle of Commutation. Similarly, the legitimacy of any line can be decided
by a finite number of observations, none of which involves anything more than
comparing shapes or patterns. To preserve this effectiveness, we lay down the
rule that only one Rule of Inference should be applied at a time. The explana¬
tory notation beside each statement is not, strictly speaking, part of the proof,
but it is helpful and should always be included.
Although the test of whether or not a given sequence of statements is a
formal proof is effective, constructing such a formal proof is not an effective
procedure. In this respect, the present method differs from the method of the
preceding chapter. The use of truth tables is completely mechanical: Given any
argument of the general sort with which we are now concerned, its validity can
always be tested by following the simple rules presented in Chapter 2. But in
constructing a formal proof of validity on the basis of the nineteen Rules of
Inference listed, it is necessary to think or ‘figure out* where to begin and how
to proceed. Although we have no effective or purely mechanical method of
procedure, it is usually much easier to construct a formal proof of validity than
to write out a truth table with perhaps dozens or hundreds or even thousands of
rows.
There is an important difference between the first nine and the last ten Rules
of Inference. The first nine can be applied only to whole lines of a proof. Thus
A can be inferred from A*B by Simplification only if AmB constitutes a whole
line. But neither A nor ADC follows from (A • B) D C by Simplification or by
any other Rule of Inference. A does not follow because A can be false while
(A • B) D C is true. ADC does not follow because, if A is true and B and C are
both false, (A *B) D C is true, whereas ADC is false. On the other hand, any of
42 The Method of Deduction [Ch. 3

the last ten Rules of Inference can be applied either to whole lines or to parts
of lines. Not only can the statement A D (B D C) he inferred from the whole
line(A*B) D Cby Exportation, but from the line [(4* B) D C] v D we can infer
[A D (B D C)]v D by Exportation. The Rule of Replacement authorizes
specified logically equivalent expressions to replace each other wherever they
occur, even where they do not constitute whole lines of a proof. But the first
nine Rules of Inference can be used only with whole lines of a proof serving as
premisses.
There are two very important differences between substitution and re¬
placement. First, in substituting statements for statement variables in an
argument form, any statement variable can have any statement substituted for
it. In replacement, on the other hand, a statement can be replaced only by a
statement logically equivalent to it, where the logical equivalence of the two
statements is given in Rules 10 through 19. Second, in substitution, every
occurrence of a statement variable in an argument form must have the same
statement substituted for it. In replacement, however, one (or several) occur¬
rences of a statement may be replaced by a logically equivalent statement,
without other occurrences of that statement having to be replaced also.
In the absence of mechanical rules for the construction of formal proofs of
validity, some rules of thumb or hints on procedure may be suggested. The first
is simply to begin deducing conclusions from the premisses by the given Rules
of Inference. As more and more of these subconclusions become available as
premisses for further deductions, one is more likely to see how to deduce the
conclusion of the argument that is to be proved valid.
Another hint is to try to eliminate statements that occur in the premisses, but
not in the conclusion. Such elimination can proceed only in accordance with
the Rules of Inference. The Rules contain many techniques for eliminating
statements, however. Simplification is one such rule: By its application, the
right-hand conjunct of a conjunction can be dropped—provided that the
conjunction is a whole line in the proof. By Commutation, the left-hand
conjunct can be switched over to the right-hand side, from which location it
can be dropped by Simplification. The ‘middle’ term q can be eliminated by a
Hypothetical Syllogism, given two premisses or subconclusions of the patterns
p D q and q D r. Distribution is a useful rule for transforming a disjunction of
th§ form p v (</-r) into the conjunction (p v q) «(p v r), the conjuncts of which
can be eliminated, either by Simplification alone, or by Commutation followed
by Simplification. Another rule of thumb is to introduce, by Addition, a
statement that occurs in the conclusion but not in the premisses. Another
method is to work backward from the conclusion by looking for some state¬
ment or pair of statements from which the conclusion could be deduced by one
of the Rules of Inference. Then those intermediate statements are deduced
either from the premisses or from other intermediate statements, and so on,
until some that are derivable .from the premisses are found. A judicious
combination of these methods is often the best way to proceed. Practice, of
course, is the best way to acquire facility in using the method of deduction.
Sec. 3.2] The Rule of Replecement 43
EXERCISES

I. For each of the following arguments, state the Rule of Inference by which its
conclusion follows from its premiss:
*1. (~A 3 B)-(C v ~D) *5. [P 3 (() v R)] v [P 3 (() v H)|
• • (~A 3 7i)*(~D v C) :.P D(()vR)
2. (~Fv7-’)-(Gv -H) 6. lS-(7’-in| 3 (V = -W)
(E 3 F)’(C v ~~ll) f(S•■/’)•(/] D(V=~~W)
.3. (7 3 ~/) v (~K 3 -/,) 7. [.V-(Y-Z)] 3 (A = ~«)
.'.(ID ~/) v (L 3 K) .’. X 3 [(Y-Z) 3 (A ~ -~B)]
4. M 3 ~(/Vv ~0) 8. (C—/3) 3 (/•' = ~F)
M 3 (-—AT- -O) ~~D) 3 [(/-> —-/•") v -F)]
9. (C;'v//)•(/v/)
[(Gvi/)-i]v[(Cvii)-;|
*10. (K-F) 3 {Af-I(:V-0)-/>|}
3 {m-[(cmv)-pj}
11. ~{gv ~[(H-~S)-(7’v '-(')]}
—{<;> V [—(K-~S) V ~(7’v -</)]}

12. -V 3 {VV 3 [~(X*Y) 3 ~Z]}

-V 3 {[VV-~(X-V)1 3 ~Z}
13. [A v (/I v C)| v [(DvW) v K]
.'. [A v(8vC)jv[Dv(/;vf:)l
14. (F 3 G)-{[(G 3 //)•(// 3 (.’)] 3 (7/ 3 /)}
(7*’ 3 G)*{(G = //) 3 (// 3 1)}
*15. 7^ ~{[(7t*~L)v v ~,V]>

16. O 3 [(P—(J) =? (P- -~K)1

.-.O 3 [<F—= (-P--Kj]
17. {~~T 3 [~~—Uv(~TS)])
~S~ {—-Tv [~- O v (—• 7’*S)|}
18. V 3 {(~VF 3 ~~X)v[(~Y 3 Z)v(~Z 3 ~Y)|}
V3 {(~X 3 W) v [(— Y 3 Z) v (~~Z 3 —■ Y)|}
19. (A—B) 3 l(C-C) 3 (C 3 D)]
.'. (A ■ ~fl) 3 [C 3 (C 3 0)1
20. (73 ~F) 3 [C 3 (G 3 77)]
A. (/<;—F) 3 [(G-G) 3 7/)]

II. Each of the following is a formal proof of validity for the indicated argument.
State the ‘justification’ for each line that is not a premiss:

*1. 1. (A vB) 3 (C‘D) 7. —fl* ~A

2. ~C /.’. -71 8.
3. ~Cv ~D 2. 1. (E-F)-G
4. ~{C-D) 2. (F~ G) 3 (77v 7) /. ./v//
5. ~(A v 77) 3. F-(F-G) •
6. ~A—B 4, (F-G)-F
44 The Method of Deduction (Ch. 3

5. F-G 3. C 3 (-C 3 ~D)

6. (F-G) v (—F-~G) 4. C 3 (C 3 ~D)
7. F=G • 5. (C-C) 3 ~D
8. Hvl 6. C 3 ~D
9. IwH 7. ~C v ~D
3. 1. (7-K) 3 L 8. ~(C‘F>)
2. (7 3 L) 3 M 9. (C-D)v(~C-~D)
3. ~K v N K 3 (M’N) 10. ~C-~D
4. (K-J) 3 L 8. 1. -E.-(FvG)
5. KD(JDL) 2. (F-G) 3 ~(//v/)
6. K DM 3. (~//v~/) D —(E-F)
7. —X v M /■'■ H= I
8. (~XvM)-(~XvlV) 4. (E-G) 3 (~//-~/)
9. —X v(M-lV) 5. ~(W*/) 3 ME-F)
10. X 3 (M-1V) 6. (E-F) 3 (//•/)
/4. 1.(03 ~P)-(P 3 (1) 7. [(E-F) D (//•/)]• [(E-C
2. g 3 O 3 (-//•-/)]
3. ~R 3 /.\R 8. (E-F) v (E-G)
4. ~QvO 9. (il-I) v (~H--~7)
5. O v 10. //=/
6. (O 3 ~P)-(~Q 3 ~P)
9. 1. 7v(~Xv7)
7. —Fv ~P 2. Xv(~7vX)
8. ~F /.-.(/•K)v (-/•-/
9. -R 3. (~Kv/)v/
10. R 4. —X v (/v/)
*5. 1. S 3 (T 3 U) 5. ~Xv7
2. (/ 3 —17 6. KDJ
3. (V 3 S)-(W 3 T) /.\VD ~W 7. (~/v X) v X
4. (S-T) 3 1/ 8. v (X v X)
5. ~~Uv~~U 9. ~7vX
6. ~l/ 10. 7 3 X
7. ~~(S'T) 11. (7 3 X)-(X 3 7)
8. ~S v —T 12. 7=X
9. ~V v ~W 13. (7-K)v(~7-~K)
10. V 3 ~W 10. 1. (LvM)v(N-O)
6. 1. X 3 (Y 3 Z) 2. (~L-0)—(~L-Af)
2. X 3 (A 3 B) /.'. ~L-AT
3. X-(YvA) 3. ~L-[0—(~L-M)]
4. ~Z /.\B 4. ~L
5. (X-Y) 3 Z 5. Lv [Mv(Af-O)]
6. (X-A) 3 B 6. Mv (A7-0'
7. (X-Y)v(X-A) .7. (MvN)-(MvO)
8. [(X-Y) 3 Z]-[(X-A) 3 B] 8. Mv N
9. ZvB 9. ~L*(Mv IV)
10. B 10. (~L-M) v (~L-W)
7. 1. C 3.(A 3 ~C). 11. —(—L-Af)-(~L-0)
2. C = D /.-.—C'-f'-D .12. ~(~L-M)
13.
Sec. 3.2] The Rule of Replacement 49
III. Construct a formal proof of validity for each of the following arguments:
*1. -A 6. N D O 11. E 3 F
A 3 B .'.(N-p) d a E 3 G
E 3 (F-G)
2. C 7. ((Jvfl) D s 12. // 3 (/ v 7)
.D 3 C .-.pDS
. .HD]
3. E 3 (F 3 C) 8. T 3 -((./ D V) 13. (KvL) D ^(M-N)
E D (E 3 G) .*. T 3 V (-A/v-,V) 3 (0= /»
(OeeP) 3 (Q-K)
/:(LvK) D (K-(>)
4. H D(l-J) 9. YV D (X--V) 14. S D T
:.n 3 / W 3 (Y D X) Sv T
.;. r
*5. K 3 L *10. A D D C) — C7v V)-((^v W)
K 3 (E v Af) 3 C -X 3 -w
D . ‘. V v X C

16. AD(BD C) 24. (F D C)-(H D /)

C D (D-E) Ev //
. .A D (B D D) (E .3 —/)•(// 3 —G)
17. E D F .‘.G== -/ t
G D F ‘25. j-(K-L)
.(Ev G) D F (K-J) D \Mv(NvO)]
18. [(//-/) 3/]-[-K 3(Z—7)] —O)
H D K .\N
19. [L-(Af v 2V)] 3 (AhN) 26. <;)v(K*S)
:.L D (M D N) (() 3 E)*(r D S)
20. O D (P D Q)
s
P D({) D R) 27. (1/ 3 V)-(W 3 X)
O 3 (P 3 R) (U v W) D(VvX)
21. S3(T-r/) 28. (Y 3 Z)*(A 3 fl)
(Tv (/) 3 V .*.(Y-A) 3(Z-B)
.\S 3 V 29. (C 3 D)*(E 3 E)
22. —Wv [(X D Y)-(Z 3 X)] C 3 (CvE) ^
.'.G 3 (DvE)
W-(XvZ)
/. y 30. (// 3 /)•(/ 3 X)
//vy
23. (AvB)D (C-D)
—A 3 (li 3 ~~E) (// 3 -K)-(7 3 -W
-C (I—X) 3 L
.'. -E X 3 (IvAf)
E v A#

IV. Construct a formal proof of validity for each of the following arguments, using the
suggested notation in each case.
46 The Method of Deduction [Ch. 3

*1. If I study, I make good grades. If I do not study, I enjoy myself. Therefore,
either I make good grades or I enjoy myself. (S, G, E)
2. If the supply of silver remains constant and the use of silver increases, then
the price of silver rises. If an increase in the use of silver implies that the
price of silver rises, then there will be a windfall for speculators. The supply
of silver remains constant. Therefore, there will be a windfall for specula¬
tors. (S, (7, P, W)
3. Either Adams is elected chairman or both Brown and Clark are elected to
the board. If either Adams is elected chairman or Brown is elected to the
board, then Davis will lodge a protest. Therefore, either Clark is elected to
the board or Davis will lodge a protest. (A, 8, C, D)
4. If he uses good bait, then if the fish are biting, then he catches the legal
limit. He uses good bait, but he does not catch the legal limit. Therefore, the
fish are not biting. (G, ft, C)
*5. Either the governor and the lieutenant-governor will botkrun for reelection
or the primary race will be wide open and the part^,will be torn by
dissension. The governor will not run for reelection. Therefore, the party
will be torn by dissension. (G, L, W, T)
6. If the Dodgers win the pennant, then they will win the Series. Therefore, if
the Dodgers win the pennant, then if they continue to hit, then they will win
the Series. (P, S, H) „
7. If he attracts the farm vote, then he will carry the rural areas, and if he
attracts the labor vote, then he will carry the urban centers. If he carries
both the urban centers and the rural areas, then he is certain to be elected.
He is not certain to be elected. Therefore, either he does not attract the farmf
vote or he does not attract the labor vote. (F, ft, L, (7, C)
8. Either Argentina does not join the alliance or Brazil boycotts it, but if
Argentina joins the alliance, then Chile boycotts it. If Brazil boycotts the
alliance, then if Chile boycotts it, then Ecuador will boycott it. Therefore, if
Argentina joins the alliance then Ecuador will boycott it. (A, ft, C, E)
9. If Argentina joins the alliance, then both Brazil and Chile will boycott it. If
either Brazil or Chile boycotts the alliance, then the alliance will be
ineffective..Therefore, if Argentina joins the alliance, then the alliance will
be ineffective. (A, ft, C, 1)
*10. Steve took either the bus or the train. If he took the bus or drove his own car,
then he arrived late and missed the first session. He did not arrive late.
Therefore, he took the ttain. (ft, T, C, L,M)
11. If you enroll in the course and study hard, then you will pass, but if you
enroll in the course and do not study hard, then you will not pass. Therefore,
if you enroll in the course, then either you study hard and pass or you do not
study hard and do not pass. (E, S, P)
12. If Argentina joins the alliance, then either Brazil or Chile will boycott it. If
Brazil boycotts the alliance, then Chile will boycott it, also. Therefore, if
Argentina joins the alliance, then Chile will boycott it. (A, ft, C)
13. If either Argentina or Brazil joins the alliance, then both Chile and Ecuador
will boycott it. Therefore, if Argentina joins the alliance, then Chile will
boycott it. (A, ft, C, E\
14. If prices fall or wages rise, Ithen both retail sales and advertising activities
 increase. If retail sales increase, then jobbers make more money, but jobbers
do not make more money. Therefore, prices do not fall. (P, W, ft, A, ])
*15. If I work, then I earn money, but if I am idle, then I enjoy myself. Either I
work or I am idle. However, if I work, then I do not enjoy myself, while if
I am idle, then I do nbt earn money. Therefore, 1 enjoy myself if and only if
I do not earn money. (W, M, /, E)
16. If you enter the primary, then if you campaign vigorously, then you win the
nomination. If you win the nomination and receive the support of the party
regulars, then you will be elected. If you take the party platform seriously,
then you will receive the support of the party regulars but will not be
elected. Therefore, if you enter the primary, then if you campaign vigor¬
ously, then you do not take the party platform serioilsly. (P, C, JV, ft, E„T)
17. Either the tariff is lowered or imports continue to decrease and our own
industries prosper. If the tariff is lowered, then our own industries prosper.
Therefore, our own industries prosper. (Ty /, O)
18. Either he has his old car repaired or he buys a new car. If he has his ojd car
repaired, then he will owe a lot of money to the garage. If he owes a lot of
money to the garage, then he will not soon be out of debt. If he buys a new
car, then he must borrow money from the bank, and if he must borrow
money from the bank, then he will not soon be out of debt. Either he will
soon be out of debt or his creditors will force him into bankruptcy. There¬
fore, his creditors will force him into bankruptcy, (ft, zV, G, S, ft, C)
19. If she goes on a picnic, then she wears sport clothes. If she wears sport
clothes, then she does not attend both the banquet and the dance. If she does
not attend the banquet, then she still has her ticket, but she does not still
have her ticket. She does attend the dance. Therefore, she does not go on a
picnic. (P, S, ft, D, T)
20. If she studies the sciences, then she prepares to earn a good living, and if she
studies the humanities, then she prepares to live a good life. If she prepares
to earn a good living or she prepares to live a good life, then her college
years are well spent. But her college years are not well spent. Therefore, she
does not study either the sciences or the humanities. (S, E, //, L, C)
*21. If you plant tulips, then your garden will bloom early, and if you plant
asters, then your garden will bloom late. So if you plant either tulips or
asters, then your garden will bloom either early or late. (!\ E, A, L)
22. If you plant tulips, then ymir garden will bloom early, and if you plant
asters, then your garden will bloom late. So if you plant both tulips and
asters, then your garden will bloom both early and late. (P, E, A, L)
23. If we go to Europe, then we tour Scandinavia. If we go to Europe, then if
we tour Scandinavia, then we visit Norway. If we tour Scandinavia, then if
we visit Norway, then we shall take a trip on a fiord. Therefore, if we go to
Europe, then we sh^ll take a trip on a fiord. (E, S, N, F)
24. If Argentina joins the alliance, then either Brazil or Chile boycotts it. If
Ecuador joins the alliance, then either Chile or Peru boycotts it. Chile does
not boycott it. Therefore, if neither Brazil nor Peru boycotts it, then neither
Argentina nor Ecuador joins the alliance. (A, ft, C, E, P)
25. If either Argentina or Brazil joins the alliance, then if either Chile or
Ecuador boycotts it, then although Peru does not boycott it, Venezuela
 [Ch. 3
boycotts it. If either Peru or Nicaragua does not boycott it, the Uruguay
will join the alliance, 'flierefore, if Argentina joins the alliance, then if Chile
boycotts it, then Uruguay will join the alliance. {A, B, C, E, P, V, N, U)
26. If either Saudi Arabia raises the price of oil or Saudi Arabia does not raise
the price of oil, then America will be in grave difficulties. Therefore,
America will be in grave difficulties. (S, A)

3.3 Proving Invalidity

We can establish the invalidity of an argument by using a truth table to show

that the specific form of that argument is invalid. The truth table proves
invalidity if it contains at least one row in which truth values are assigned to
the statement variables in such a way that the premisses are made true and the
conclusion false. If we can devise such a truth value assignment without
constructing the entire truth table, we shall have a shorter method of proving
invalidity.
Consider the invalid argument

If the Senator votes against this bill, then he is opposed to more severe
penalties against tax evaders.
If the Senator is a tax evader himself, then he is opposed to more severe
penalties against tax evaders.
Therefore, if the Senator votes against this bill, he is a tax evader himself,

which may be symbolized as

V D O
H D O
.'. V D II

Instead of constructing a truth table for the specific form of this argument, we
can prove its invalidity by making an assignment of truth values to the
component simple statements V, O, and II, which will make the premisses true
and the conclusion false. The conclusion is made false by assigning T to V and
F to H; and both premisses are made true by assigning T to O. This method of
proving invalidity is closely related to the truth table method. In effect, making
the indicated truth value assignment amounts to describing one row of the
relevant truth table—a row that suffices to establish the invalidity of the
argument being tested. The relationship appears more clearly, perhaps, when
the truth value assignments are written out horizontally:

V O H VO O HDO VDH
. — . ... . ... ... .i . . . —. —
T T F T T F
Sec. 3.4] The Rule of Conditions Proof 49
This new method of proving invalidity is shorter than writing out a complete
truth table. The amount of time and work saved is proportionally greater for
more complicated arguments. In proving the invalidity of more extended
arguments, a certain amount of trial and error may be needed to discover a
truth value assignment that works, i But even so, this method is quicker and
easier than writing out the entire truth table. It is obvious that the present
method will suffice to prove the invalidity of any argument that can be shown
to be invalid by a truth table.

EXERCISES

Prove the invalidity of each of the following arguments by the method of assigning
truth values: ..
► *1. ADD / 3 (K 3 L) *5. u
C3D K ~J (-/. 3 M) U== (V-W)
BvC (L v M) 3 N Vf (7’ v A‘)
/.AvD 3 N Tv X
T-X
2. E 3 (Fv G) 4. (O v P) 3 {) 6. X ^ (V 3 Z)
G 3 (Ihl) Q 3 (Tv R) y ■=(-*—/)
O 3 (~~S 3 P) Ze(Xv-V)
E 3 / (S 3 O) 3 -Zt y
..F=Q . ’. X v Z

7. (ADH)-(C 3 D) 9. P=K)^-/0
A vC 3 (-/iv -S)
(«vD) 3 (E-F) [« D(()v-r.)]-(lJD9)
E 3 (F 3 G) [G 3 (S-T)l-(i 3 -V)
G 3 (A 3 //) [«>•«) -f/|-[i/D(yvii)]
// (()vVJ—V

8. / v (V'K)
(ivy) 3 (/.= -*#) X = (Z 3 Y)
(L 3 —Af) 3 y = (z=-A)
(/V 3 0)-(0 3 -V/) Z nE (A 3 fl)
(/ 3 K) 3 O A ~ (Z* = Z)
.'.O Z* v - W
U ~ B

3.4 The Rule of Conditional Proof

Next, we introduce a new rule to use in the method of deduction: the rule of
Conditional Proof. In this section, the new rule will be applied only to
arguments whose conclusions are conditional statements. The new rule can best
be explained and justified by reference to the principle of Exportation and the
correspondence, noted in Chapter 2, between valid argument forms and
tautologies.
50 The Method of Deduction [Ch. 3
To every argument there corresponds a conditional statement whose ante¬
cedent is the conjunction of the argument’s premisses and whose consequent is
the argument’s conclusion. As has been remarked, ,an argument is valid if and
only if its corresponding conditional is a tautology. If an argument has a
conditional statement for its conclusion, which we may symbolize as A D C,
then if we symbolize the conjunction of its premisses as P2 the argument is
valid if and only if the conditional

p => (A D C) (1)

is a tautology. If we can deduce the conclusion A D C by a sequence of

elementary valid arguments, from the premisses conjoined in P\ we thereby
prove the argument to be valid and the associated conditional (1) to be a
tautology. By the principle of Exportation, (1) is logically equivalent to

(PmA) D C (2)

But (2) is the conditional associated with a somewhat different argument. This
.second argument has, as its premisses, all of the premisses of the first argument
plus an additional premiss that is the antecedent of the conclusion of the first
argument. The conclusion of the second argument is the consequent of the
conclusion of the first argument. Now, if we deduce the conclusion of the
second argument, C, from the premisses conjoined in P*A by a sequence of
elementary valid arguments, we thereby prove that its associated conditional
statement (2) is a tautology. But since (1) and (2) are logically equivalent, this
fact proves that (1) is a tautology also, from which it follows that the original
argument—with one less premiss and the conditional conclusion A D C—is
valid also. Now, the rule of Conditional Proof permits us to infer the validity of
any argument

P
.'.A D C

from a formal proof of validity fbr the argument

P
A
.*.C

Given any argument whose conclusion is a conditional statement, a proof of

its validity, using the rule of Conditional Proof—that is, a conditional proof of
its validity—is constructed by assuming the antecedent of its conclusion as an

2For definiteness, we specify that the conjunction has its conjuncts associated to the left, that
is, if the premisses are Plt F2, and P3, the conjunction is (fVP2)*P3. If there is a fourth premiss, P4,
the conjunction is [(Pj'PJ’Isl’P* an^ so on-
Sac. 3.4] The Rule of CondNtonal Proof Sf
additional premiss, and then deducing the consequent of its conclusion by a
sequence of elementary valid arguments. Thus a conditional proof of validity
for the argument

(AvB)D (C-D)
(DvE) D F
.’.A D F

may be written as

1. (AvB)D (C-D)
2. (DvE) 3 F /.’.ADF
3. A /.' .F (C.P.)
4. AvB 3, Add.
5. C-D 1, 4, M.P.
6. D-C 5, Com.
7. D 6, Simp.
8. DvE 7, Add.
9. F 2, 8, M.P.

Here, the second slant line and three-dot ‘therefore’ symbol, as well as the
parenthesized ‘C.P.’, indicate that the rule of Conditional Proof is being used.
Since the rule of Conditional Proof can be used in dealing with any valid
argument having a conditional statement as conclusion, it can be applied more
than once in the course of the same deduction. Thus a conditional proof of
validity for

AD(BD C)
B D (C 3 D)
.’.A D(BDD)

will be a proof of validity for

A D (B D C)
B D (C D D)
A
B D D

and since the latter, itself, has a conditional conclusion, it can be given a
conditional proof by proving the validity of

AD(BDC)
BD (CD D)
A
B
:.d
52 The Method of Deduction [Ch. 3
Each use of the conditional method should be signaled by an additional slant
line and ‘therefore’ sign, in addition to the notation ‘(C.P.)’. The suggested
proof would be written:

1. AD (BDC)
2. B D (C D D) //.A D(BDD)
3. A /:.BDD (C.P.)
4. B /.-.D (C.P.)
5. BDC 1, 3, M.P.
6. C 5, 4, M.P.
7. CD D 2, 4, M.P.
8. D 7, 6, M.P.

The rule of Conditional Proof is a genuine addition to the proof apparatus of

Sections 3.1 and 3.2. Not only does it permit the construction of shorter proofs
of validity for arguments that could be proved valid by appealing to the
original list of nineteen Rules of Inference alone, but it permits us to establish
the validity of valid arguments whose validity could not be proved by refer¬
ence to the original list alone.
It is proved in Appendix A3 that the obviously valid argument

AD B
.’.AD( A-B)

cannot be proved valid using only the original list of nineteen Rules of
Inference. Rut it is easily proved valid using the new rule of Conditional Proof.
Its conditional proof of validity is

1. ADB /.'.A D (A-B)

2. A /..A-B (C-P.
3. B 1, 2, M.P.
4. A-B 2, 3, Conj.

EXERCISES

Give conditional proofs of validity for Exercises *21, 22, 23, 24, and 25 on pages
47-48.

3.5 The Rule of Indirect Proof

The method of indirect proof, often called the method of proof by reductio
ad ahsurdutn, is familiar to all w'ho have studied elementary geometry. In
deriving his theorems, Euclid.often begins by assuming the opposite of what

3 See pages 333-336.

Sec. 3.5] The Rule of Indirect Proof 53
he wants to prove. If that assumption leads to a contradiction, or ‘reduces to an
absurdity, then that assumption must be false, and so its negation—the
theorem to be proved—must be true.
An indirect proof of validity for a given argument is constructed by assum¬
ing, as an additional premiss, the negation of its conclusion, and then deriving
an explicit contradiction from the augmented set of premisses. Thus an indirect
proof of validity for the argument

A D (B-C)
(B v D) D E
D\A
.*. e

may be set down as follows:

1. AD (B-C)
2. (B v D) D E
3. DvA /.'.£
4. ~~E I.P. (Indirect Proof)
5. ~~(B v D) 2, 4, M.T.
6. 5, De M.
7. 6, Com.
8. 7, Simp.
9. A 3, 8, D.S.
10. B-C 1, 9, M.P.
11. B 10, Simp.
12. —B 6, Simp.
13. B-~~B ’ll, 12, Conj.

Here line 13 is an explicit contradiction, so the demonstration is complete. The

validity of the original argument follows by the rule of Indirect Proof.
Adding the rule of Indirect Proof serves to strengthen our proof apparatus
still further. Any argument whose conclusion is a tautology can be shown to be
valid, by the method of truth tables, regardless of what its premisses may be.
But if the tautologous conclusion of an argument is not a conditional statement,
and the premisses are consistent with each other and quite irrelevant to that
conclusion,4 then the argument cannot be proved valid using only the methods
of Sections 3.1, 3.2, and 3.4. The rule of Indirect Proof must be used. Although
the argument

A
.*. B v (B D C)

4 In this context, one statement is irrelevant to another if there is no simple statement that is
a component of both.
cannot be proved valid by the methods set forth in the preceding sections, its
validity is easily established by using the rule of Indirect Proof. One proof of
its validity is this:

1. A A’ . B v (B D
2. ~[B v (BD C)] I.P.
3. ~[Bv (~B'/C)] 2, Impl.
4. ' ~B) vC] 3, Assoc.
5. ~(B v -*)• ~c 4, De M.
6. ~(B v 5, Simp.
7. ~B-~B 6, De M.

Our nineteen Rules of Inference, plus the rules of Conditional and Indirect
Proof, provide us with a method of deduction that is complete. Any argument
whose validity can be established by the use of truth tables can be proved valid
by the method of deduction as set forth in Sections 3.1, 3.2, 3.4 and 3.5. This
will not be proved, however, until the end of Chapter 8.

EXERCISES

For each of the following arguments, construct both a formal proof of validity and
an indirect proof and compare their lengths:

1. A v (B'C) 4. (MvN)D (O-P)

ADC (Ovp) D (-R-S)
.\C (R v T) D (M-U)
~~R
2. (Dv E) D(FD G) *5. (V D — W)-(X D Y)
(~~GvH) D (D-F) (~~W D Z)-(Y D -A)
G (Z D -8)-(-A D C)
v-x

*3. (H D !)•(/ D K)
(I v K) DL

.*.-(//v ])

3.6 Proofs of Tautologies

The conditional and indp^ct methods of proof can be used not only^ to
establish the validity of^rfguments, but also to prove that certain statements
and statement forarrfare tautologies’. Any conditional statement corresponds, in
a sense, to an argument whose single premiss is the antecedent of the condi¬
tional, and whose conclusion is the conditional’s consequent. The conditional is
Sec. 3.6] Proofs of Tautologies 55
a tautology if and only if that argument is valid. Hence a conditional is proved
tautologous by deducing its consequent from its antecedent by a sequence of
elementary valid arguments. Thus the statement (A*B) DA is proved tau¬
tologous by the same sequence of lines that proves the validity of the argument

A-B
.'. A

It has already been noted that the conditional method can be used repeatedly
in a single proof. Thus the conditional statement

(V [(P D Q) D(PD R)]

is proved tautologous by the following:

1. QDR /.*. (P D Q) D {r D R) (C.P.)

2. PDQ /.. P D R (C.P.)
3. P D R 2, 1, H.S.

For some complicated conditional statements, this method of proving them

tautologous is shorter and easier than constructing truth tables.
There are many tautologies that are not conditional in form. To these, the
preceding method cannot be applied. But any tautology can be established as
tautologous by the indirect method. As applied to an argument, the indirect
method of proving validity proceeds by adding the negation of its conclusion to
the argument’s premisses and then deducing a contradiction by a sequence of
elementary valid arguments. As applied to a statement, the indirect method of
proving it tautologous proceeds by taking its negation as premiss and then
deducing an explicit contradiction by a sequence of elementary valid argu¬
ments. Thus the statement B v ~B is proved to be a tautology by the follow¬
ing:

1. -(Bv-B) /.\Bv-B (I.P.)

2. —B 1, De M.

To say that a statement is a tautology is to assert that its truth is uncondi¬

tional, so it can be established without appealing to any other statements as
premisses. Another, perhaps not too misleading, way of saying the same thing
is to assert the validity of the argument’ that has the statement in question as
‘conclusion’, but has no premisses at all. If the ‘conclusion’ is a tautology, then
the method of deduction permits us to prove that the ‘argument’ is valid, even
though it has no premisses—using either the rule of Conditional Proof or the
rule of Indirect Proof. Any tautology can be established by the method of
deduction, although this will not be proved until the end of Chapter 8.
 56 The Method of Deduction [Ch. 3
EXERCISES
'/ . .
I. Use the method of conditional proof to verify that the following are tautologies:
*1. PD(QD P)
•2. [PD(QDR)]D[(PDQ)D(PDR)]
•3. [P D (Q D R)] D [£ D (P D R)]
4. (PDQ)D (-Q D ~P)
*5. —~b d p
6.PD-B
7. (A D B) D [(B D C) D (A D C)]
8. [(A D B)-(A D C)] D [A D (B v C)]
9. [(A D B)-(A D Q] D [A D (B-C)]
.*10. (A D B) D [A D (A-B)]
11. (AD B) D [(-A D 0) D BJ
12. (A D B) D [(A-C) D (B-C)|
13. [(A D il) D »1 D (A v B)
14. (B D C) D [(A vB) D (CvA)]
;i5. [A D (B-C)] D {[B D (D*£)j D (A D D)}
16. [(A vB) D C] D {[(CvD) D £] D (A D £))
17. [(A D B) D A] D A
18. P D (B-B)
19. (P-Q) D P
20. (B D <)) D [-(<?•«) =5 -(B-B)]

“T&t. Use the method of indirect proof to verify that the following are tautologies:
*1. (A D B) v (A D —B) 6. A v (A D B)
2. (A D B) v (—A D B) 7. P=-B
3. (A D B)v(B D A) * 8. A = [A*(A v £)]
4. (ADB)v(BDC) j* 9. A = [A v (A-B)]
*5. (A D B) v (~A D C) 10. -[(A D -A)-(-A D A)]

3.7 The Strengthened Rule of Conditional Proof

In the preceding sections, the method of Conditional Proof was applied only
to arguments whose conclusions were conditional in form. But in the next
chapter, it will be convenient to use something like the method of Conditional
Proof for arguments whose conclusions are not explicit conditional statements.
To accomplish this purpose, we strengthen our rule of Conditional Proof and
thereby give it wider applicability.
To formulate our Strengthened. Rule of Conditional Proof, it is useful to
adopt a new method of writing-out proofs that make use of the Conditional
Method. As explained in Section 3.5, we used the method of Conditional Proof
Sec. 3.7] The Strengthened Rule of Conditional Proof 57
to establish the validity of an argument that has a conditional as conclusion by
adding the antecedent of that conditional to the argument s premisses as an
assumption, and then deducing the conditional’s consequent. The notation in
Section 3.5 involved the use of an additional slant line and an extra therefore
sign, as in proving the validity of the argument

AD B
..A D A-B)

by the four-line proof on page 52.

A Conditional Proof of validity for that same argument is set down in our
new notation as the following sequence of five lines:

1. ADB /.’.AD (A-B)

2. A assumption
3. B 1, 2, M.P.
4. A-B 2, 3, Conj.
5. AD (A-B) 2-4, C.P.

Here, the fifth line is inferred not from any one or two of the preceding lines,
but from the sequence of lines 2, 3, 4, which constitutes a valid deduction of
line 4 from lines 1 and 2. In lino 5, we infer the validity of the argument

/\ D B
/.A D (A-B)

from the demonstrated validity of the argument

A D B
A
.’.A-B

That inference is ‘justified’ by noting the sequence of lines to which appeal is

made, using the letters ‘C.P.’ to show that the principle of Conditional Proof is
being used.
In the second of the preceding proofs, line 2, the assumption, has lines 3 and
4 dependent upon it. Line 5, however, is not dependent upon line 2, but only
upon line 1. Line 5 is, therefore, outside or heyoncl the scope of the assumption
made as line 2. When an assumption is made in a Conditional Proof of validity,
its ‘scope’ is always limited' never extending all the way to the last line of the
demonstration.
A notation is now introduced that is very helpful in keeping track of
assumptions and their scopes. A bent arrow is used for this purpose, with its
head pointing at the assumption from the left, its shaft bent down to run along
all lines within the scope of the assumption, and then bent inward to mark the
58 The Method of Deduction [Ch. 3
end of the scope of that assumption. The scope of the assumption in the
preceding proof is indicated thus: - •

1. ADB A D (A*B)
P> 2. A assumption
3. B 1, 2, M.P.
4. A*B 2, 3, Conj.
5. A D (A'B) 2-4, C.P.

It should be observed that only a line inferred by the principle of Conditional

Proof ends the scope of an assumption, and that every use of the rule of
Conditional Proof serves to end the scope of an assumption. When the scope of
an assumption has been ended, the assumption is said to have been discharged.
No subsequent line can be justified by reference to it or to any line lying
between it and the line inferred by the rule of Conditional Proof that dis¬
charges it. Only lines lying between an assumption of limited scope and the
line that discharges it, can be justified by reference to that assumption. After
one assumption of limited scope has been discharged, another such assumption
may be made and then discharged. Or a second assumption of limited scope
may be written within the scope of a first. Scopes of different assumptions may
follow each other, or one scope may be contained entirely within another.
If the scope of an assumption does not extend all the way to the end of a
proof, then the final line of the proof does not depend on that assumption, but
has been proved to follow from the original premisses alone. Hence we need
not restrict ourselves to using as assumptions only the antecedents of condi¬
tional conclusions. Any proposition can be taken as an assumption of limited
scope, for the final line that is the conclusion will always be beyond its scope
and independent of it.
A more complex demonstration that involves making two assumptions is the
following (incidentally, when our bent arrow notation is used, the word
'assumption' need not be written, since each assumption is sufficiently identi¬
fied by the arrowhead on its left):

1. (AvB)D[(CvD)DE] /. A D [(C-D) D E]
i->2. A
3. AvB 2, Add.
4. (CvD)DE 1,3, M.P.
r-5. C-D
6. Q 5, Simp.
7. CvD Q, Add.
8. E_ 4, 7, M.P.
9. (C-D) D E 5-8, C.P.
10. A D [(C-D) D E] 2-9, C.P.
Sec. 3.7] The Strengthened Rule of Conditional Proof 59
In this proof, lines 2 through 9 lie within the scope of the first assumption,
while lines 5, 6, 7, and 8 lie within the scope of the second assumption. From
these examples, it is clear that the scope of an assumption a in a proof consists
of all lines a through <p, where the line following <p is of the form a D <p and is
inferred by C.P. from that sequence of lines. In the preceding proof, the second
assumption lies within the scope of the first because it lies between the first
assumption and line 10, which is inferred by C.P. from the sequence of lines 2
through 9.
When we use this new method of writing out a Conditional Proof of validity,
the scope of every original premiss extends all the way to the end of the proof.
The original premisses may be supplemented by additional assumptions,
provided that the latter’s scopes are limited and do not extend to the end of the
proof. Each line of a formal proof of validity must be either a premiss, or an
assumption of limited scope, or must follow validly from one or two preceding
lines by a Rule of Inference, or must follow from a sequence of preceding lines
by the principle of Conditional Proof.
Some arguments can be proved valid by the Strengthened Rule of Condi¬
tional Proof, although the original Rule of Conditional Proof, as stated in
Section 3.4, cannot be applied to them. Consider the following argument and
its proof:

L AD B
2. [AD(A-B)]DC /.*.C

1, 3, M.P.
3, 4, Conj.
6. AD(A-B) 3-5, C.P.
7. C ' 2, 6, M.P.

Here the conclusion, C, is not a conditional statement, so the original Rule of

Conditional Proof as stated in Section 3.4 is simply not applicable. But the
Strengthened Rule of Conditional Proof, which permits any statement to be
taken as an assumption of limited scope, here permits a quick and easy proof
of validity.
Of course, the Rule of Indirect Proof could be used here, but in the present
case, it would result in a longer proof. Because the Strengthened Rule of
Conditional Proof permits any assumption of limited scope to be made, we
can, if we wish, assume the negation of the argument's conclusion. Thus the
Strengthened Rule of Conditional Proof subsumes the Rule of Indirect Proof,
although when C.P. rather than I.P. is used and a contradiction is deduced, we
are required to go on through or past the contradiction, instead of stopping
with it. We must do that in order to discharge the assumption and end its scope.
We go on through the contradiction by deducing the conclusion of the argu¬
ment from that contradiction.
60 The Method of Deduction [Ch. 3

It is easy to show that, from a contradiction, any conclusion can validly be

deduced. In other words, any argument of the form

.\q

is valid, no matter what statements are substituted for the variables p and q.
Thus from lines 11 and 12 in the proof on page 53, the conclusion E can be
derived in just two more lines. Such a continuation would proceed:

14. BwE 11, Add.

15. E 14, 12, D.S.

Hence it is possible to regard an indirect proof of the validity of a given

argument, not as the deduction of its validity from the fact that a contradiction
was obtained, but rather as the deduction of the conclusion of that argument
from the contradiction itself. To transform the proof of validity on page 53 into
a proof using the Strengthened Rule of Conditional Proof, simply make the
following changes. Insert an arrow head pointing to line 4 to label it an
assumption. Then delete line 13 altogether, because the contradiction is
already explicit in lines 11 and 12. Then deduce B v E (as line 13) from B (line
11) by Addition. Then deduce E (as line 14) from line 13 and line 12 by a
Disjunctive Syllogism. We are still within the scope of our assumption. To
discharge that assumption, we bend the arrow’s shaft in underneath line 12 and
write beneath it:

15. —E D E 4-12, C.P.

We do not yet have the desired conclusion, but we can get it from line 15 in
just three easy steps.

16. -Kvf; 15, Impl.

17. E v E 16, O.M.
18. E 17, Taut.

From now on, the Strengthened Rule of Conditional Proof will be referred to
simply as the Rule of Conditional Proof.5 * * * * 10

5 For further discussion of the relations between the Rule of Indirect Proof and the (Strength¬
ened) Rule of Conditional Proof, see Donald Scherer, ‘The Form of Reductio ad Absurdurn, Mind,
vol. 80 (1971), pp. 247-252; Paiil Foulkes, ‘The Form of Reductio ad Absurdurn, Mind, vol. 82
(1973), pp. 579-580; Charles II. Lambros; ‘Scherer on Reductio ad Absurdurn, Mind, vol. 82
(1973), pp. 581-585; J. M. Lee, ‘The Form of Reductio ad Absurdurn, Notre Dame Journal of
Formal Logic, vol. 14 (1973), pp. 381-386; and Lyman C. D. K. Kulathungam, ‘Reductio-ad-
Absurdurn: A Family Feud between Copi and Scherer’, Notre Dame Journal of Formal Logic, vol.
10 (1975), pp. 245-254.
Sec. 3.8] Shorter Truth Table Technfque-ft<toc//o Ad Absurdum Method 61
EXERCISES

^ Use the strengthened method of conditional proof to prove the validity of the
following arguments:

*1. A D B . 4. Q\ (R 3 S)
B D f(C 3 -C) 3 D] [K D(/US)1 Df/’vU)
/.ADD (r D 3 V)
<;>v v
2. (E v F) 3 G 5. [W D (-V*-y)]-[Z D -(XvV)l
II 3 (/•/) (-A D \V)-{^B D /)
.-.(E 3 (;)•(// D /) (A D X)m(B D V)
..A — y
3. (X D L)m(M 3 N) 6. (C v D) D (X D F)
(LvN) 3 {[O D (Ov I*)] D (K-M)} [E D (/<>F)| D C;
c; D [(*-//V --//) D (OH)]

3.8 Shorter Truth Table Technique—Reductio Ad Absurdum Method

There is still another method of testing the validity of arguments and of

classifying statements as tautologous, contradictory, or contingent. In the
preceding section, it was pointed out that an argument is invalid if and only if
it is possible to assign truth values to its component simple statements in such a
way as to make all its premisses true and its conclusion false. It is impossible to
make such truth value assignments in the case of a valid argument. Hence to
prove the validity of an argument, it suffices to prove that no such truth values
can be assigned. We do so by showing that its premisses can be made true and
its conclusion false only by assigning truth values inconsistently, so that some
component statement is assigned both a T and an F. If the truth value T is
assigned to each premiss of a valid argument and the truth value F is assigned
to its conclusion, this will necessitate assigning both T and F to some compo¬
nent statement, which is, of course, a contradiction.
Thus to prove the validity of the argument

(AvB)D(C-D)
(D v E) D F
:.A 3 F

we assign T to each premiss and F to the conclusion. Assigning F to the

conclusion requires that T be assigned to A and F be assigned to F. Since T is
assigned to A, the antecedent of the first premiss is true, and since the premiss
has been assigned T, its consequent must be true also, so T must be assigned to
both C and D. Since T is assigned to D, the antecedent of the second premiss is
true, and since the premiss has been assigned T, its consequent must also be
true, so T must be assigned to F. We have already been forced to assign F to
62 The Method of Deduction [Ch. 3

F, however, to make the conclusion false. Hence the argument is invalid only if
the statement F is both true and false, which is impossible. This method of
proving the validity of an argument is a version of the reductio ad absurdum
technique that uses truth value assignments rather than Rules of Inference.
It is easy to extend the use of this method to the classification of statements
(and statement forms). Thus to certify that Peirce’s Law [(p D q) D p] D p is
a tautology, we assign it the truth value F, which requires us to assign T to its
antecedent [(p D q) D p] and F to its consequent p. For the conditional
[(p D q) D p] to be true while its consequent p is false, its antecedent (p D q)
must be assigned the truth value F also. But for vhe conditional p D q to be
false, its antecedent p must be assigned T and its consequent q assigned F.
However, we were previously forced to assign F to p, so assuming Peirce’s
Law false leads to a contradiction, which proves it a tautology.
If it is possible to assign truth values consistently to its components on the
assumption that it is false, then the expression in question is not a tautology, but
must be either contradictory or contingent. In such a case, we attempt to assign
truth values to make it true. If this attempt leads to a contradiction, the
expression cannot possibly be true and must be a contradiction. But if truth
values can be assigned to make it true and other truth values can be assigned to
make it false, then it is neither a tautology nor a contradiction, but is contin¬
gent.
The reductio ad absurdum m#iipd of assigning truth values is by far the
quickest and easiest method of testing arguments and classifying statements. It
is, however, more readily applied in some cases than in others. If F is assigned
to a disjunction, F must be assigned to both disjuncts, and where T is assigned
to a conjunction, T must be assigned to both conjuncts. Here the sequence of
assignments is forced. But where T is assigned to a disjunction or F to a
conjunction, that assignment, by itself, does not determine which disjunct is
true or which conjunct is false. At this point we shall have to experiment and
make various ‘trial assignments’, which will tend to diminish the advantage of
the method for such cases. Despite these complications, however, in the vast
majority of cases, the reductio ad absurdum method is superior to any other
method known.

EXERCISES

1. Use the reductio ad absurdum method of assigning truth values to decide the
validity or invalidity of the arguments and argument forms in the exercises on
pages 25-26.
2. Use the reductio ad absurdum method of assigning truth values to establish that the
statements in Exercises I and II on page 56 are tautologies.
3. Use the reductio ad absurdum method of assigning truth values to classify the
statement forms in Exercise I on pages 30-31 as tautologous, contradictory, or con¬
tingent.
Quantification Theory

4.1 Singular Propositions and General Propositions

1 he logical techniques developed thus far apply only to arguments whose

validity depends upon the way in which simple statements are truth-function-
ally combined into compound statements. Those techniques cannot be applied
to such arguments as the following:

All humans are mortal.

Socrates is human.
Therefore, Socrates is mortal.

The validity of such an argument depends upon the inner logical structure of
the noncompound statements it contains. To appraise such arguments, we must
develop methods for analyzing noncompound statements and symbolizing their
inner structures.
The second premiss of the preceding argument is a singular proposition; it
says that the individual Socrates has the attribute of being human. We call
‘Socrates’ the subject term and ‘human’ the predicate term. Any (affirmative)
singular proposition states that the individual referred to by its subject term has
the attribute designated by its predicate term. We regard as individuals not
only persons, but any things, such as animals, cities, nations, planets, or stars, of
which attributes can significantly be predicated. Attributes can be designated
not only by adjectives, but by nouns or even verbs. Thus ‘Leslie is a gossip’ and
‘Leslie gossips’ have the same meaning, which can also be expressed as ‘Leslie
is gossipy’.
In symbolizing singular propositions, we use the small letters V through
to denote individuals, ordinarily using the first letter of an individual’s name to
denote that individual. Because these symbols denote individuals, we call them
individual constants’. To designate attributes, we use capital letters, being
guided by the same principle in their selection. Thus in the context of the
preceding argument, we denote Socrates by the small letter V and symbolize
the attributes human and mortal by the capital letters 7T and ‘M’. To express
a singular proposition in our symbolism, we write the symbol for its predicate

63
64 Quantification Theory [Ch. 4

term to the left of the symbol for its subject term. Thus we symbolize ‘Socrates
is human' as 7/s’ and ‘Socrates is mortal' as ‘Ms’.1
Examining the symbolic formulations of singular propositions having the
same predicate term, we observe them to have a common pattern. The
symbolic formulations of the singular propositions ‘Aristotle is human', ‘Boston
is human', ‘California is human’, ‘Descartes is human’,..., which are 7/a’,
7/b’, 7fc\ 7/d', ..., each consists of the attribute symbol 7/' followed by an
individual constant. We use the expression 7/x\ to symbolize the pattern
common to all such singular propositions. The small letter Y—called an
‘individual variable'—is a mere place marker that serves to indicate where an
individual constant can be written to produce a singular proposition. The
singular propositions *Ha\ ‘Hh‘Hc\ 7/d’,... are either true or false; but ‘Hx
is neither true nor false, since it is not a proposition. Such expressions as 7/x’ are
called ‘propositional functions'. These are defined to be expressions that
contain individual variables and become propositions when their individual
variables are replaced by individual constants.2 Any singular proposition can
be regarded as a substitution instance of the propositional function from which
it results by the substitution of an individual constant for the individual
variable in the propositional function. The process of obtaining a proposition
from a propositional function by substituting a constant for a variable is called
‘instantiation’. The negative singular propositions ‘Aristotle is not human’ and
‘Boston is not human', symbolized as ‘—//a’ and Hb\ result by instantiation
from the propositional function of which they are substitution in¬
stances. Thus we see that symbols other than attribute symbols and individual
variables can occur in propositional functions.
. General propositions such as ‘Everything is mortal’ and ‘Something is mortal'
differ from singular propositions in not containing the names of any individuals.
However, they also can be regarded as resulting from propositional functions,
not by instantiation, but by the process called ‘generalization’ or ‘quantifica¬
tion’. The first example, ‘Everything is mortal', can alternatively be ex¬
pressed as

Given any individual thing whatever, it is mortal.

Here the relative pronoun ‘it’ refers back to the word ‘thing’, which precedes it
in the statement. We can use the individual variable Y in place of the pronoun
‘it’ and its antecedent to paraphrase the first general proposition as

Given any x, x is mortal.

Then we can use the notation already introduced to rewrite it as

Given any x, Mx.

1 Some logicians enclose the individual constant in parentheses, symbolizing 'Socrates is human’
as *H{s)\ but we shall not follow that practice here.
2Some writers have defined 'propositiohal functions’ to be the meanings of such expressions; but
here we define them as the expressions themselves.
Scc- 4,11 Singular Propositions and General Propositions 65

The phrase Given any x is called a ‘universal quantifier’, and is symbolized by

(*) • Using this new symbol, we can completely symbolize our first general
proposition as

(x)Mx

We can, similarly, paraphrase the second general proposition, ‘Something is

mortal’, successively as

There is at least one thing that is mortal.

There is at least one thing such that it is mortal.
There is at least one x such that x is mortal.

and as

There is at least one x such that Mx.

The phrase There is at least one x such that’ is called an ‘existential quantifier'
and is symbolized as ‘(3:*)’. Using the new symbol, we can completely symbol¬
ize our second general proposition as

(3x)Mx

A general proposition is formed from a propositional function by placing either

a universal or an existential quantifier before it. It is obvious that the universal
quantification of a propositional function is true if and only if all of its
substitution instances are true. Also, the existential quantification of a propo¬
sitional function is true if and only if it has at least one true substitution
instance. If we grant that there is at least one individual, then every proposi¬
tional function has at least one substitution instance (true or false). Under this
assumption, if the universal quantification of a propositional function is true,
then its existential quantification must be true as well.
A further relationship between universal and existential quantification can
be shown by considering two additional general propositions, ‘Something is not
mortal’ and ‘Nothing is mortal’, which are the respective negations of the first
two general propositions just considered. ‘Something is not mortal’ is symbol¬
ized as ‘(3x)~~Mx’ and ‘Nothing is mortal’ is symbolized as \x)~~Mx\ These
show that the negation of the universal (existential) quantification of a propo¬
sitional function is logically equivalent to the existential (universal) quantifi¬
cation of the new propositional function that results from placing a negation
symbol in front of the first propositional function. Where we use the Greek
letter phi to represent any attribute symbol whatever, the general connections
between universal and existential quantification can be described in terms of
the following square array:
66 Quantification Theory [Oh. 4

Assuming the existence of at least one individual, we can say that the two top
propositions are contraries, that is, they might both be false but cannot both be
true; the two bottom propositions are subcontraries, that is, they can both be
true but cannot both be false. Propositions that are at opposite ends of the
diagonals are contradictories, of which one must be true and the other false.
Finally, on each side, the truth of the lower proposition is implied by the truth
of the proposition directly above it.
Traditional logic emphasized four types of subject-predicate propositions
illustrated by the following:

All humans are mortal.

No humans are mortal.
Some humans are mortal.
Some humans are not mortal.

These were classified as universal affirmative’, ‘universal negative’, ‘particular

affirmative’, and ‘particular negative’, respectively, and their types abbreviated
as ‘A’, *E\ 7’, ‘O’, again respectively. (The letter names have been presumed
to come from the Latin ‘Afflrmo’ and ‘n£gO\ meaning I affirm and l deny.)
These four special forms of subject-predicate propositions are easily symbol¬
ized by means of propositional functions and quantifiers.3 The first of them, the
A proposition, can successively be paraphrased as

Given any individual thing whatever, if it is human, then it is mortal.

Given any jc, if x is human, then x is mortal.
Given any xf x is human D x is mortal.

and finally symbolized as

(x)(Hx D Mx)

3 An alternative method of symbolizing them is presented in Section 7.1.

Sbc- 4-1] Singular Propositions and General Propositions 67

Our symbolic formulation of the A proposition is the universal quantification of

the complex propositional function 7/x D Mx\ which has as its substitution
instances not singular propositions but conditionals whose antecedents and
consequents are singular propositions that have the same subject terms. Among
the substitution instances of the propositional function ‘Hx D Mx' are the
conditionals gHa D Ma\ ‘Hb D Mb\ 7/c D Mc\ and so on. In symbolizing an
A proposition the parentheses (or square brackets or braces) serve as punctu¬
ation marks to indicate that the universal quantifier ‘(a)* applies to or has within
its scope the whole of the complex propositional function 7/x D Mx*. The
notion of the scope of a quantifier is very important, for differences in scope
correspond to differences in meaning. The expression \x)(Hx D Mx)' is a
proposition which asserts that all substitution instances of the propositional
function 4Hx D Mx are true. On the other hand, the expression \x)Hx D Mx'
is a propositional function whose substitution instances are \x)Hx D Ma\
g(x)Hx D Mb', \x)Hx D Mc\ and so forth.4
The E proposition ‘No humans are mortal* may similarly be paraphrased
successively as

Given any individual thing whatever, if it is human, then it is not mortal.

Given any x, if x is human, then x is not mortal.
Given any x, x is human D x is not mortal.

and then symbolized as

(x)(llx ID ~~Mx)

Similarly, the 1 proposition ‘Some humans are mortal*, may be para¬

phrased as

There is at least one thing that is human and mortal.

There is at least one thing such that it is human and it is mortal.
There is at least one x such that x is human and x is mortal.
There is at least one x such that x is human -x is mortal.

and completely symbolized as

(3 .v)(//x*Afx)

Finally, the O proposition ‘Some humans are not mortal, becomes

There is at least one thing that is human but not mortal.

There is at least one thing such that it is human and it is not mortal.
There is at least one x such that x is human and x is not mortal.

4 We have the same symbolic convention for quantifiers (both universal and existential) that we
established for negation on page 12: a quantifier applies to or has for its scope the smallest
component that the punctuation permits.
68 Quantification Theory [Ch. 4

and then symbolized as the existential quantification of a complex function

(3x)(ffx-^Afx) '

Where the Greek letters phi and psi are used to represent any attribute
symbols whatever, the four general subject-predicate propositions of tradi¬
tional logic may be represented in a square array as

(.v) (4»x D \l/x) (x) (<t»x D ~\J/x)

E

()
(3.0 (4h- • 0.v) (30 (4>v • *4,x)

Of these, the A and the O are contradictories, and the E and the I are
contradictories also. But none of the other relationships discussed in connection
with the square array on page 66 hold for the traditional A, E, Z, and O
propositions, even if we assume that there is at least one individual in the
universe. Where 4 <l>x* is a propositional function that has no true substitution
instances, then regardless of what attribute is symbolized by the proposi¬
tional functions *<bx D tyx’ and ‘4>x D —'I'x’ have only true substitution
instances. This is so because all their substitution instances are conditional
statements with false antecedents. In such cases both the A and E propositions
that are the universal quantifications of these complex propositional functions
are true, so A and E propositions are not contraries. Again, where ‘<I>x is a
propositional function that has no true substitution instances, then, regardless
of what €yVx might be, the propositional functions ‘<J>x • ♦x’ and ‘<t>x • ~~'Vx have
only false substitution instances,*Ior their substitution instances are conjunc¬
tions whose first conjuncts are false. In such cases the 1 and O propositions that
are the existential quantifications of these complex propositional functions are
false. Thus I and O propositions are not subcontraries. In all such cases, then,
since the A and E propositions are true and the / and O propositions are false,
the truth of a universal does not imply the truth of the corresponding particu¬
lar; no implication relation holds between them.
If we make the assumption that there is at least one individual, then
‘(x)(<t>x D 'I'*)’ does imply \3x)(<bx O ^x)\ The latter, however, is not an f
proposition. An / proposition qf the form 'Some <J>’s are 'k’s is symbolized as
‘(3x)(<I>x-'I,x)\ which asserts that .there is at least one thing having both the
attribute $ and the attribute 'V. But the proposition D asserts
Sec' Singular Propositions and General Propositions 69

only that there is at least one object that either has the attribute or does not
have the attribute <1>, which is a very different and much weaker assertion.
The four traditional subject-predicate forms A, E, 7, and O are not the only
forms of general propositions. There are others that involve the quantification
of more complicated propositional functions. Thus the general proposition ‘All
members are either parents or teachers’, which does not mean the same as ‘All
members are parents or all members are teachers’, is symbolized as
(x)[Mx D (Pxv lx)]. And the general proposition ‘Some Senators are either
disloyal or misguided’, is symbolized as '(3x)[Sx-(Dxv Mx)]\ It should be
observed that such a proposition as ‘Apples and bananas are nourish¬
ing’ can be symbolized either as the conjunction of two A propositions,
*[(*)(-A* Nx)] • [(x)(Bx D Nx)]’, or as a single noncompound general proposi¬
tion, ‘(x)[(Ax v Bx) D Nx\\ It should not, however, be symbolized as
<(x)[(Ax*Bx) D Nx]\ since to say that apples and bananas are nourishing is to
say that anything is nourishing which is cither an apple or a banana, not that
anything is nourishing which is both an apple and a banana (whatever that
might be). It must be emphasized that there are no mechanical rules for
translating statements from English into our logical notation. In every case one
must understand the meaning of the English sentence and then re-express that
meaning in terms of propositional functions and quantifiers.

EXERCISES

I. Translate each of the following into the logical notation of propositional functions
and quantifiers. In each case, use the abbreviations suggested and begin each
formula with a quantifier, not with a negation symbol:

*1. Snakes are reptiles. (Sx: x is a snake. Bx: x is a reptile.)

2. SnalJes are not all poisonous. (Sx: x is a snake. Bx: x is poisonous.)
3. Children are present. (Cx: x is a child. Px: x is present.)
4. Executives all have secretaries. (Ex: x is an executive. Sx: x has a secretary.)
*5. Only executives have secretaries. (Ex: x is an executive. Sx: x has a secretary.)
6. Only property owners may vote in special municipal elections. ' Px: x is a
property owner. Vx: x may vote in special municipal elections.)
7. Employees may use only the service elevator (Ux: x is an elevator that
employees may use. Sx: x is a service elevator.)
8. Only employees may use the service elevator. (Ex: x is an employee. Vx: x
may use the service elevator.)
9. All that glitters is not gold. (Cx: x glitters. Ax: x is gold.)
*10. None but the brave deserve the fair. (Bx: x is brave. Dx: x deserves the fair.)
11. Not every visitor stayed for dinner. ( Vx: x is a visitor. Sx: x stayed for dinner.)
12. Not any visitor stayed for dinner. (Vx: x is a visitor. Sx: x stayed for dinnett)
13. Nothing in the house escaped destruction, (//x: x was in the house. Ex: x
escaped destruction.)
14. Some students are both intelligent and hard workers. (Sx: x is a student, lx: x
is intelligent. Hx: x is a hard worker.)
*15. No coat is waterproof, unless it has been specially treated. (Cx: x is a coat.
Wx: x is waterproof. Sx: x has been specially treated.)
70 Quantification Theory [Ch. 4

16. Some medicines are dangerous only if taken in excessive amounts. (Mx: x is
a medicine. Dx: x is dangerous. Ex: x is taken in excessive amounts.)
17. All fruits and vegetables are wholesome and delicious. (Fx: x is a fruit. Vx: x
is a vegetable. Wx: x is wholesome. Dx: x is delicious.)
18. Everything enjoyable is either immoral, illegal, or fattening. (Ex: x is enjoya¬
ble. Mx; x is moral. Lx: x is legal. Ex: x is fattening.)
19. A professor is a good lecturer if and only if he is both well informed and
entertaining. (Px: x is a professor. Gx: x is a good lecturer. Wx: x is well
informed. Ex: x is entertaining.)
*20. Only policemen and firemen are both indispensable and underpaid. (Px: x is
a policeman. Fx: x is a fireman. Ix: x is indispensable. Ux: x is underpaid.)
21. Not every actor is talented who is famous. (Ax: x is an actor. Tx: x is talented.
Fx: x is famous.)
22. Any girl is healthy if1 she is well nourished and exercises regularly. (Gx: x is
a girl. IIx: x is healthy. Wx: x is well nourished. Ex: x exercises regularly.)
23. It is not true that every watch will keep good time if and only if it is wound
regularly and not abused. (Wx: x is a watch. Kx: x keeps good time. Rx: x is
wound regularly. Ax: x is abused.)
24. Not every person who talks a great deal has a great deal to say. (Px: x is a
person. Tx: x talks a great deal. Hx: x has a great deal to say.)
*25. No automobile that is over ten years old will be repaired if it is severely
damaged. (Ax: x is an automobile. Ox: x is over ten years old. Rx: x will be
repaired. Dx: x is severely damaged.)

In symbolizing the following, use the abbreviations: Hx: x is a horse. Gx: x is

gentle. Tx: x has been well trained.

26. Some horses are gentle and have been well trained.
27. Some horses are gentle only if they have been well trained.
28. Some horses are gentle if they have been well trained.
29. Any horse is gentle that has been well trained.
*30. Any horse that is gentle has been well trained.
31. No horse is gentle unless it has been well trained.
32. Any horse is gentle if it has been well trained.
33. Any horse has been well trained if it is gentle.
34. Any horse is gentle if ancfonly if it has been well trained.
*35. Gentle horses have all been well trained.
36. Only well-trained horses are gentle.
37. Only gentle horses have been well trained.
38. Only horses are gentle if they have been well trained.
39. Some horses are gentle even though they have not been well trained.
*40. If something is a well-trained horse, then it must be gentle.
41. Some horses that are well trained are not gentle.
42. Some horses are neither gentle nor well trained.
43. No horse that is well trained fails to be gentle.
44. A horse is gentle only if it has been well trained.
45. If anything is a gentle horse, then it has been well trained.
46. If any horse is well trained, then it is gentle.
Sec. 4.2] Proving Validity: Preliminary Quantification Rules 71

II. Symbolize the following, using propositional functions and quantifiers:

*1. Blessed is he that considered the poor. (Psalm 41:1)

2. He that hath knowledge spareth his words. (Proverbs 17:27)
3. Whoso findeth a wife findeth a good thing. (Proverbs 18:22)
4. He that maketh haste to be rich shall not be innocent. (Proverbs 28:20)
*5. They shall sit every man under his vine and under his fig-tree. (Micah 4:4)
6. He that increased knowledge increased sorrow. (Ecclesiastes 1:18)
7. Nothing is secret which shall not be made manifest. (Luke 8:17)
8. Whom The Lord loveth He chastened. (Hebrews 12:6)
9. If a man desire the office of a bishop, he desircth a good work. (1 Timothy 3:1)
10. He that hateth dissembled with his lips, and layeth rtp deceit within him.
(Proverbs 26:24)

4.2 Proving Validity: Preliminary Quantification Rules

To construct formal proofs of validity for arguments symbolized by means of

quantifiers and propositional functions, we must augment our list of Rules of
Inference. We shall add four rules governing quantification, offering an over¬
simplified preliminary statement of them in this section and giving a more
adequate formulation in Section 4.5.

1. Universal Instantiation (Preliminary Version). Because the universal

quantification of a propositional function is true if and only if all substitution
instances of that propositional function are true, we can add to our list of Rules
of Inference the principle that any substitution instance of a propositional
function can validly be inferred from its universal quantification. We can
express this rule symbolically as

(a*)(<p.v)
(where v is any individual symbol)

Since this rule permits substitution instances to be inferred from universal

quantifications, we refer to it as the ‘principle of Universal Instantiation’, and
abbreviate it as ‘UI.’5 The addition of UI permits us to give a formal proof of
validity for the argument: ‘AH humans are mortal; Socrates is human; there¬
fore, Socrates is mortal’.

1. (x)(Hx D Mx)
2. Hs //.Ms
3. Hs D Ms 1, UI
4. Ms 3, 2, M.P.

5This rule and the three that follow are variants of rules for natural deduction’, which were
devised independently by Gerhard Gentzen and Stanislaw Jankowski in 1934.
72 Quantification Theory [Ch. 4

2. Universal Generalization (Preliminary Version). We can explain our

next rule by analogy with fairly standard mathematical practice! A geometer
may begin a proof by saying, ‘Let ABC be any arbitrarily selected triangle’.
Then he may go on to prove that the triangle ABC has some specified attribute,
and concludes that all triangles have that attribute. Now, what justifies his final
conclusion? Why does it follow from triangle ABC*s having a specified attri¬
bute that all triangles do? The answer is that if no assumption other than its
triangularity is made about ABC, then the expression 'ABC* can denote any
triangle whatsoever. And if the argument has established that any triangle
whatsoever must have the attribute in question, then it follows that all triangles
do. We now introduce a notation analogous to that of the geometer in his
reference to ‘any arbitrarily selected triangle’. The hitherto unused small letter
‘y will be used to denote any arbitrarily selected individual In this usage, the
expression is a substitution instance of the propositional function ‘4>x’, and
it asserts that any arbitrarily selected individual has the property <I>. Clearly,
‘4>i/’ follows validly from ‘(x)(Ox)’ by UI, since what is true of all individuals is
true of any arbitrarily selected individual. The inference is equally valid in the
other direction, since what is true of any arbitrarily selected individual must be
true of all individuals. We augment our list of Rules of Inference further by
adding the principle that the universal quantification of a propositional
function can validly be inferred from its substitution instance with respect to
the symbol ‘y’. Since this rule permits the inference of general propositions that
are. universal quantifications, we refer to it as the ‘principle of Universal
Generalization’, and abbreviate it as ‘UG\ Our symbolic expression for this
second quantification rule is

<I>iy
(where ‘xj denotes any arbitrarily selected individual,
.■.(*)(<!>*) and Or/ is not within the scope of any assumption
containing the special symbol ‘r/’)

We can use the new notation and the additional rule UG to construct a
formal proof of validity for the argument: ‘No mortals are perfect; all humans
are mortal; therefore, no human? are perfect’.

1. (x)(Mx D ~~Px)
2. (x)(Hx DMx) /. *. (x)(Hx D -Px)
3. lly D My 2, UI
4. My D ~~Py 1,UI
5. Ily D ~~Py 3, 4, H.S.
6. (x)(IIx D ~Px) 5, UG

We can explain the need for the two restrictions in our statement of UG by
considering two obviously invalid arguments. First, the silly argument ‘Chicago
is large, therefore, everything is large’. If we ignored the restriction on UG that
Sec. 4.2] Proving Validity: Preliminary Quantification Rules 73

the premiss must contain the special symbol y\ we might be led to construct
the following ‘proof’:

1. Ijc /.'. (x)Lx

2. (x)Lx 1, UG (wrong)

The mistake occurs at line 2. The individual symbol in line 1 is <\ but UG can
be used only on a line containing
The second invalid argument to consider is the obviously invalid: ‘Not
everything is edible; therefore, nothing is edible’. If we ignored the restriction
on UG that the Of/ must not be within the scope of an assumption containing
the special symbol ‘i/\ we might be led to construct the following ‘proof’:

1. ~~(x)Ex (x)*Ex
2. Etj
3. (x)Ex 2, UG (wrong)
47 Ey D (x)Ex 2-3, C.P.
5. ^Eij 4, 1, M.T.
6. (x)^Ex 5, UG

The mistake here occurs at line 3. The line 2 to which UG is (mis)applied, is

within the scope of an assumption containing the special symbol Uj. This
shows, incidentally, that we regard every assumption as being (or lying) within
its own scope. To avoid errors like these, we must obey the indicated restric¬
tions in using UG.

3. Existential Generalization (Preliminary Version). Because the existential

quantification of a propositional function is true if and only if that proposi¬
tional function has at least one true substitution instance, we can add to our list
of Rules of Inference the principle that the existential quantification of a
propositional function can validly be inferred from any substitution instance of
that propositional function. This rule permits the inference of general propo¬
sitions that are existentially quantified, so we call it the ‘principle of Existential
Generalization’ and abbreviate it as ‘EG’. Its symbolic formulation is

(where v is anv individual symbol)

4. Existential Instantiation (Preliminary Version). One further quantifica¬

tion rule is required. The existential quantification of a propositional function
asserts that there exists at least one individual the substitution of whose name
for the variable *x in that propositional function will yield a true substitution
instance of it. Of course, we may not know anything else about that individual.
74 Quantification Theory [Ch. 4

But we can take any individual constant other than ‘y\ say, *w\ which has had
no prior occurrence in that context and use it to denote the individual or one of
the individuals whose existence has been asserted by the existential quantifica¬
tion. Knowing that there is such an individual, and having agreed to denote it
by *w\ we can infer from the existential quantification of a propositional
function the substitution instance of that propositional function with respect to
the individual symbol We add, as our final quantification rule, the principle
that from the existential quantification of a propositional function we may
validly infer the truth of its substitution instance with respect to an individual
constant that has no prior occurrence in that context. The new rule may be
written as

(3x)(<fr3c) (where v is an individual constant, other than 1 2 3 4y\

.'.$v that has no prior occurrence in the context)

It is referred to as the ‘principle of Existential Instantiation' and abbreviated as

‘EI\
We make use of the last two quantification rules in constructing a formal
proof of validity for the argument: ‘All dogs are carnivorous; some animals are
dogs; therefore, some animals are carnivorous'.

1. (x)(Dx D Cx)
2. (3 x)(Ax*Dx) /.’ . (3x)(Ax'Cx)
3. Aw-Dw 2, El
4. Dw D Cw 1,UI
5. DwAw 3, Com.
6. Dw 5, Simp.
7. Cw 4, 6, M.P.
8. Aw 3, Simp.
9. AwCw 8, 7, Conj.
10. (3 x)(Ax-Cx) 9, EG

We can show the need for tffe indicated restriction on the use of El by
considering the obviously invalid argument: ‘Some cats are animals; some dogs
are animals; therefore, some cats are dogs'. If we ignored the restriction on El
that the substitution instance inferred by it can contain only an individual
constant that had no prior occurrence in the context, we might be led to
construct the following ‘proof’:

1. (3 x)(Cx-Ax)
2. (3 x)(Dx*Ax) /.‘.(3 x)(Cx9Dx)
3. CwAw 1, El.
4. Dw •Aw 2, El (wrong)
Sec. 4*2] Proving Validity: Preliminary Quantification Rules 75

5. Cw 3, Simp.
6. Dw 4, Simp.
7. CwDw 5, 6, Conj.
8. (3x)(Cx-Dx) 7, EG

The mistake here occurs at line 4. The second premiss assures us that there is
at least one thing that is both a dog and an animal. We are not free to use the
symbol w to denote that thing, however, because *w' has already been used to
denote one of the things that was asserted by the first premiss to be both a cat
and an animal. To avoid errors of this sort, we must ojbey the indicated
restriction in using El. It should be clear that whenever we use both El and UI
in a proof to instantiate with respect to the same individual constant, we must
use El first. (It has been suggested that we could avoid the necessity of using El
before UI by changing the restriction on El to read ‘where v is an individual
constant, other than ‘t/’, that was not introduced into the context by any
previous use of El/ Even apart from its apparent circularity, that formulation
would not prevent the construction of an erroneous ‘formal proof of validity'
for such an invalid argument as ‘Some men are handsome. Socrates is a man.
Therefore, Socrates is handsome.')
Like the first nine Rules of Inference presented in Section 3.1, the four
quantification rules UI, UG, EG, and El can be applied only to whole lines in
a proof.
Any assumption of limited scope may be made in a Conditional Proof of
validity and, in particular, we are free to make an assumption of the form ‘4>t/\
Thus the argument ‘All freshmen and sophomores are invited and will be
welcome; therefore, all freshmen are invited’ is proved valid by the following
Conditional Proof:.

1. (x)f(FxvSx) 3 (Ix'Wx)] /:. (x)(Fx 3 lx)

2. *'y
3. (FyvSy) 3 (ly-Wy) 1,UI
4. FyvSy 2, Add.
5. IyWy 3,4, M.P.
6. iy 5, Simp.

7. FyDly 2-6, C.P.

(x)(Fx 3 lx) 7, UG

More than one assumption of limited scope can be made in proving the validity
of arguments involving quantifiers, as in the following Conditional Proof:

1. (x)[(AxvBx) D (Cx*Dx)]
2. (x){(CxvEx) D [(FxvGx) D IIx) //.(x)[Ax D (Fx D Hx)]
3. (Ay v By) D (Cy *Dy) 1, UI
4. (Cy v Ey) D [(Fy v Gy) D Hy] 2, UI
 [Ch. 4

—> 5. Ay
6. Ay v By 5, Add
7. CyDy 3, 6, M.P.
8. Cy 7, Simp.
9. Cy v Ey 8, Add.
10. (Fy v Gy) 3 Hy 4,9, M.P.
pll. Fy
12. FyvGy 11, Add.
13. Hy 10,12, M.P.
14. Fy 5 Ily 11-13, C.P.
15. Ay 3 (Fy 3 i/y) 5-14, C.P.
16. (x)[Ax 3 (Fx 3 J/x)] 15, UG

XERCISES

I. Construct formal proofs of validity for the following arguments, using the rule of
Conditional Proof wherever you wish:

*1. (x)(Ax 3 Bx) *5. (xj(Kx D Lx)

—Bt (x)l(Kx-Lx) D Mx]
—At . (x)(Kx D Mx)
2. (x)(Cx 3 Dx) 6. (x)(Nx D Ox)
(x)(Ex 3 ~Dx) (x)(l'x 3 Ox)
(x)(Ex 3 -~Cx) .-.(x)[(NxvPx) 3 Ox]
3. (x)(Fx D ~~Cx) 7. (x)(\)x 3 Hx)
(3 x)(Hx‘Gx) (3x){QxvRx)
:.(3x)(Hx-~-Fx) (3x)Rx

J4. (x)(Ix D Jx) 8. (x)[Sx 3 (Tx 3 Ux)]

(3x)(Ix—Jx) (x)[Ux 3 (Vx*Wx)]

.\(x)(Jx D lx) (x)[Sx 3 (Tx 3 Vx)]

9. (*)[(Xxv Yx) 3 (Zx-Ax)]

(x)[(ZxvAx) 3 (Xx- Yx)]
,\(x)(Xx=Zx)
10. (x)[(Bx 3 Cx)-(Dx 3 Ex)f
(x)[(Cxv Ex) 3 {[Fx 3 (Gx 3 Ft)] 3 (Bx-Dx)}]
:.(x)(Bx=Dx)

I. Construct formal proofs of validity for the following arguments, using the rule of
Conditional Proof wherever you wish:

*1. All athletes are brawny. Charles is not brawny. Therefore, Charles is not an
athlete. (Ax, Bx, c)
2. No contractors are dependable. Some contractors are engineers. Therefore,
some engineers are not dependable. (Cx, Dx, Ex)
3. All fiddlers are gay. Soni^ hunters are not gay. Therefore, some hunters are
not fiddlers. (Fx, Gx, I lx)
Sec. 4.2] Proving Validity: Preliminary Quantification Rules 77
4. No judges are idiots. Kanter is an idiot. Therefore, Ranter is not a judge.
(/*, /*, k)
5. All liars are mendacious. Some liars are newspapermen. Therefore, some
newspapermen are mendacious. (Lx, M.v, \Jx)
6. No osteopaths are pediatric ians. Some quacks are pediatricians. Therefore,
some quacks are not osteopaths. (Ox, Px, (Jx)
7. Only salesmen are retailers. Not all retailers are travelers. Therefore, some
salesmen are not travelers. (S.v, Rx, Tx)
8. There are no uniforms that are not washable. There are no washable velvets.
Therefore, there are no velvet uniforms. ((■ v, W'\\ V.v)
9. Only authoritarians are bureaucrats. Authoritarians are all churlish. There¬
fore, any bureaucrat is churlish. (Ax, Rx, Cx)
*10. Dates are edible. Only items of food are edible. All items of food are good.
Therefore, all dates arc good. (Dx, Ex, Fx, Cx)
11* All dancers are graceful. Mary is a student. Marv is a dancer. Therefore, some
students are graceful. (Dx, Cx, S.v, in)
12. Tigers are fierce and dangerous. Some tigers are beautiful. Therefore, some
dangerous things are beautiful. (Tx, Fx, Dx, Rx)
St 13. Bananas and grapes are fruits. Fruits and vegetables are nourishing. There¬
fore, bananas are nourishing. (Rx, Cx, Fx, Vx, \Tx)
14. A communist is either a fool or a knave. Fools are naive. Not all communists
are naive. Therefore, some communists are knaves. (Cx, Fx, Kx, Nx)
*15. All butlers and valets are both obsequious and dignified. Therefore, all butlers
are dignified. (Rx, Vx, Ox, Dx)
16. All houses built of brick are warm and cozv. All houses in Engle¬
wood are built of brick. Therefore, all houses in Englewood are warm.
(lix, Rx, Wx, Cx, Fx)
17. All professors are learned. All learned professors are savants. Therefore, all
professors are learned savants, i Tx, 1.x, Sx)
18. All diplomats are public servants. Some diplomats are eloquent. All
eloquent public servants are orators. Therefore, some diplomats are orators.
(Dx, Fx, Ex, Ox)
Y 19. Doctors and lawyers are college graduates. Any altruist is an idealist. Some
lawyers are not idealists. Some doctors are altruists. Therefore, some college
graduates are idealists. (Dx, Lx, Cx, Av, lx)
*20. Bees and wasps sting if they are either angry or frightened. Therefore, any
bee stings if it is angry. (Rx, W'r, Sv, Ax, Fx)
21. Any authors are successful if and only if they are well read. All authors are
intellectuals. Some authors are successful but not well read. Therefore, all
intellectuals are authors. [Ax, Sx, Wx, lx)
22. Every passenger is either in first class or in tourist class. Each passenger is in
tourist Mass if and only if he is not wealthy. Some passengers are wealthy. Not
all passengers are wealthy. Therefore, some passengers are in tourist class.
(Px, Fx, Tx, Wx)
23. All members are both officers and gentlemen. All officers are fighters. Only a
pacifist is either a gentleman or not a fighter. No pacifists are gentlemen if
they are fighters. Some members are fighters if and only if they are officers.
Therefore, not all members are fighters. (Mx, Ox, Cx, lx, Px)
78 Quantification Theory [Ch. 4

24. Wolfhounds and terriers are hunting dogs. Hunting dogs and lap dogs are
domesticated animals. Domesticated animals, are gentle and useful. Some
wolfhounds are neither gentle nor small. Therefore, some terriers are small
but not gentle. ( Wx, Tx, Hx, Lx, Dx, Gx, Ux, Sx)
25. No man who is a candidate will be defeated if he is a good campaigner. Any
man who runs for office is a candidate. Any candidate who is not defeated
will be elected. Every man who is elected is a good campaigner. Therefore,
any man who runs for office will be elected if and only if he is a good
campaigner. (Mx, Cx, Dx, Gx, Rx, Ex)

4.3 Proving Invalidity

In the preceding chapter we proved the invalidity of invalid arguments

containing truth-functional compound statements by assigning truth values to
their component simple statements in such a way as to make their premisses
true and their conclusions false. We use a closely related method to prove the
invalidity of invalid arguments involving quantifiers. The method of proving
invalidity about to be described is connected with our basic assumption that
there is at least one individual.
The assumption that there is at least one individual could be satisfied in
infinitely many different ways: if there is exactly one individual, or if there are
exactly two individuals, or if there are exactly three individuals, and sp forth.
For any such case, there is a strict logical equivalence between npncompound
general propositions and truth-functional compounds of singular propositions.
If there is exactly one individual, say, a, then

(x)(<t>x) = Ofl and (3x)(<t>x) = <I>rt

If there are exactly two individuals, say a and b, then

(x)(<I>x) = (<I>a •<!>&) and (3x)(<I>x) = (<l>a v 4>b)

And for any number k, if there are exactly k individuals, say a, b, c,..., k, then

(x)(<l>x) =
and
(3jc)(4>x) = (<t>a v <I>h v v ... v 4>fc)

The truth of these biconditionals is an immediate consequence of our defini¬

tions of the universal and existential quantifiers. No use is made here of the four
quantification rules presented in the preceding section. So for any possible
nonempty universe or model containing any finite number of individuals, every
general proposition is logically equivalent to some truth-functional compound
of singular propositions. Hence, for any such model, every argument whose
Sec. 4.3] Proving Invalidity 79
propositions contain quantifiers is logically equivalent to some argument
containing only singular propositions and truth-functional compounds of them.
An argument whose propositions contain quantifiers is valid if and only if it
is valid no matter how many individuals there are, so long as there is at least
one individual. So an argument involving quantifiers is valid if and only if for
every possible nonempty universe or model it is logically equivalent to a
truth-functional argument that is valid. Hence we can prove the invalidity of
a given argument by displaying or describing a model for which the given
argument is logically equivalent to an invalid truth-functional argument. We
can accomplish this purpose by translating the given argument, whose propo¬
sitions contain quantifiers, into a logically equivalent argument containing only
singular propositions and truth-functional compounds of them, and then using
the method of assigning truth values to prove the truth-functional argument
invalid. For example, given the argument

All whales are heavy.

All elephants are heavy.
Therefore, all whales are elephants.

we first symbolize it as

(x)(Wx D Hx)
(;x)(Ex D Hx)
(x)(Wx D Ex)

In the case of a model containing exactly one individual, say, a, the given
argument is logically equivalent to

V Va D 11a
Ei D 11a
.’. Wa D Ea

which is proved invalid by assigning the truth value T to *Wa and 7/r/’ and F
to ‘Ea’. (Tliis assignment of truth values is a shorthand way of describing the
model in question as one that contains only the one individual a, which is \V (a
whale), and H (heavy), but not E (an elephant).)6 Hence the original argument
is not valid for a model containing exactly one individual and is, therefore,
invalid.

6Here wc assume that the simple propositional functions Ax', 'Bx\ 'Cx\ ... arc neither
necessary, that is, logically true of all individuals (for example, x is identical with itself), nor
impossible, that is, logically false of all individual* (for example, x is different from itself). We also
assume that the only logical relations among the simple propositional functions are those asserted
or logically implied by the premisses of the argument being proved invalid. The point of these
restrictions is to permit the arbitrary assignment of truth values to substitution instances of these
simple propositional functions without inconsistency—for our model-descriptions must, of course,
he consistent.
80 Quantification Theory [Ch. 4

It must be emphasized that, in proving the invalidity of arguments involving

quantifiers, no use is made of our quantification rides. For a model containing
just the one individual a, we do not infer the statement *Wa D Ha9 from the
statement \x)(Wx D Hx)' by UI. Those two statements are logically equivalent
for that model, because in it ‘ Wa D Ha is the only substitution instance of the
propositional function ‘Wx D Hx'.
It can happen that an invalid argument involving quantifiers is logically
equivalent, for any model containing exactly one individual, to a valid truth-
functional argument, although it will be logically equivalent, for any model
containing more than one individual, to an invalid truth-functional argument.
For example, consider the argument

All whales are heavy.

Some elephants arc heavy.
Therefore, all whales are elephants.

which is symbolized as

(x)(Wx D Hx)
(3.t)( Ex • Hx)
/. (x)(Wx D Ex)

For a model containing just the one individual a, this argument is logically
equivalent to

Wa D Ha
Eam Ha
. ’. Wa D Ea

which is a valid argument. But for a model consisting of the two individuals a
and fo, the given argument is logically equivalent to

(Wa DjIa)-(Wb D Hh)

(Ea*Ha) v (Eb*Hb)
.*. (Wa D Ea)-(Wb => Eb)

which is proved invalid by assigning the truth value T to €Wa\ ‘Wb\ 1Ha', 7/b’,
‘Eb’, and the truth value F to ‘Ea. Hence the original argument is invalid,
because there is a model for which it is logically equivalent to an invalid
truth-functional argument.
Another illustration is

Some dogs are pointers.

Some dogs are spaniels. •
Therefore, some pointers are spaniels.
Sec. 4.3] Proving Invalidity 81
which we symbolize as

(3 x)(Dx-Px)
(3x)(Dx'Sx)
(3x)(PxmSx)

For a model containing just the one individual a, it is logically equivalent to

Da-Pa
Da • Sa
. ’. Pa • Sa

which is valid. But tor a model consisting of the two individuals a and b, it is
equivalent to

(Da-Pa)v(Db-Ph)
(Da • Sa) v (Db • Sb)
:.(Pa-Sa)v(Pb-Sb)

which is proved invalid by assigning the truth value T to 'Da\ 1Db\ ‘Pa\ 'Sh\
and the truth value F to T// and 'Sa. Here, too, the original argument is
invalid, because there is a model for which it is logically equivalent to an
invalid truth-functional argument.
An invalid argument involving quantifiers may be valid for any model
containing fewer than k individuals, even though it must be invalid for every
model containing k or more individuals. Hence in using this method to prove
the invalidity of an argument involving quantifiers, it may be necessary to
consider larger and larger models. The question naturally arises: How large a
model must we consider in trying to prove the invalidity of an argument of this
type? A theoretically satisfactory answer to this question has been found. If an
argument contains n different predicate symbols, then if it is valid for a model
containing 2n individuals, then it is valid for every model or universally valid.7
This result holds only for propositional functions of one variable and is not true
of the relational predicates discussed in Chapter 5. Although theoretically
satisfactory, this solution is not of much practical help. If we were to go
straight to the theoretically crucial case for deciding the validity or invalidity
of any of the arguments already considered in this section, we should have to
consider models containing eight individuals. And for some of the following
exercises, the theoretically crucial case would be a model containing 28 = 256
individuals. In fact, however, none of the following exercises requires consid¬
eration of models containing more than three individuals to prove their
invalidity.

7 See Paul Bernays and Moses Schonfinkel. ‘Zuin Entscheidungsproblcm der inathematischen
Logik’. Mathematische Annaltm.. vol. 99 (1928), and Wilhelm Ackerinann. Solvable Cases of the
Decision Problem, North-Holland Publishing Company, Amsterdam, 1954, chap. 4.
J/Z Quantification Theory [Ch. 4
EXERCISES

I. Prove that each of the following arguments is invalid:

*1. (3x)(Ax*Bx) 6. (x)[Mx D (Nx-Ox)]

Ac (3x)(Px-Nx)
..Be (3 x)(Px-~Ox)
(x)(Mx D ~~Px)
2. (x)(Cx D ~Dx) 7. (x)[@x D (Rx-Sx)]
~~Cj (3*)(7\-Rx)
:.Dj (3x)(Tx‘ ~-Sx)
.-. (x)(Qx D Tx)
3. (x)( Ex D Fx) 8. (*)[ Ux D (VxD Wx)]
(x)(Gx D Fx) (x)[ Vx D (Ux D ~Wx)]
(x)(Ex D Gx) (3x)( Ux-Wx)
.'. (3x)(l/x-Vx)
4. (x)(Hx D ~~Ix) 9. (lx)(Xx-Yx)
(3x)(Zx*~/x) (x)(Xx D Zx)
.\(x)(Hx DJx) (3x)(Zx-~Xx)
(3x)(Zx,~Yx)
*5. (3x)(Kx-Lx) 10. (x)(Ax D Bx)
(3 x)(~Kx-~Lx) (3 x)(Cx-Bx)
.'. (lx)(Lx-~~Kx) (3 x)(Cx-~Bx)
(x)(Ax D Cx)

II. Prove that each of the following arguments is invalid:

*1. All astronauts are brave. Jim is brave. Therefore, Jim is an astronaut.
2. No cowboys are dudes. Bill is not a dude. Therefore, Bill is a cowboy.
3. All evergreens are fragrant. Some gum trees are not fragrant. Therefore, some
evergreens are not gum trees.
4. All heathens are idolaters. No heathen is joyful. Therefore, no idolater is
joyful.
*5. No kittens are large. Some mammals are large. Therefore, no kittens are
mammals.
6. All novelists are observant. Some poets are not observant. Therefore, no
novelists are poets.
7. All statesmen are intelligent. Some politicians are intelligent. Not all politi¬
cians are intelligent. Therefore, no statesmen are politicians.
8. All statesmen are intelligent. Some politicians are intelligent. Not all politi¬
cians are intelligent. Therefore, all stktesmen are politicians.
9. All statesmen are politicians. Some statesmen are intelligent. Some politi¬
cians are not statesmen. Therefore, some politicians are not intelligent.
10. Horses and cows are mammals. Some animals are mammals. Some animals are
not mammals. Therefore, all horses are animals.

III. Prove the validity or prove the 'invalidity of each of the following arguments:

*1. All aviators are brave: Jones is brave. Therefore, Jones is an aviator.
2. All collegians are debonair. Smith is a collegian. Therefore, Smith is debo-
Sec. 4.4] Multlply-General Propositions 83
3. No educators are fools. All gamblers are fools. Therefore, no educators are
gamblers.
4. No historians are illiterates. All illiterates are underprivileged. Therefore, no
historians are underprivileged.
*5. Only citizens are voters. Not all residents are citizens. Therefore, some
residents are not voters.
6. Only citizens are voters. Not all citizens are residents. Therefore, some
voters are not residents.
7. Automobiles and wagons are vehicles. Some automobiles are Fords. Some
automobiles arc trucks. All trucks are vehicles. Therefore, some Fords are
trucks.
8. Automobiles and wagons are vehicles. Some automobiles are Fords. Some
automobiles are trucks. All vehicles are trucks. Therefore, some Fords are
trucks.
9. Automobiles and wagons are vehicles. Some automobiles are Fords. Some
automobiles are trucks. Some wagons arc not vehicles. Therefore, some
Fords are trucks.
*10. All tenors arc either overweight or effeminate. No overweight tenor is
effeminate. Some tenors are effeminate. Therefore, some tenors are over¬
weight.
11. All tenors are either overweight or effeminate. No overweight tenor is
effeminate. Some tenors are effeminate. Therefore;, some tenors are not
overweight.
12. No applicant will be either hired or considered who is either untrained or
inexperienced. Some applicants are inexperienced beginners. All applicants
who are women will be disappointed if they are not hired. Every applicant
is a woman. Some women will be hired. Therefore, some applicants will be
disappointed.
13. No candidate is cither elected or appointed who is cither a liberal or a
radical. Some candidates arc wealthy liberals. All candidates who are
politicians arc disappointed if they are not elected. Every candidate is a
politician. Some politicians arc elected. Therefore, some candidates are not
disappointed.
14. Abbots and bishops are churchmen. No churchmen are either dowdy or
elegant. Some bishops are elegant and fastidious. Some abbots arc not
fastidious. Therefore, some abbots are dowdy.
15. Abbots and bishops are churchmen. No churchmen are both dowdy and
elegant. Some bishops arc elegant and fastidious. Some abbots are not
fastidious. Therefore, some abbots are not dowdy.

4.4 Multiply-General Propositions

Thus far, we have limited our attention to general propositions containing

only a single quantifier. A general proposition that contains exactly one
quantifier is said to be singly general. We turn next to multiply-general
propositions, which contain two or more quantifiers. In our use of the term, any
compound statement whose components are general propositions is to be
84 Quantification Theory [Ch. 4

counted as a multiply-general proposition. For example, the conditional ‘If all

dogs are carnivorous then some animals are. carnivorous’, symbolized as
\x)(Dx D Cx) D (3x)(Ax*Cx)’, is a multiply-general proposition. Other multi¬
ply-general propositions are more complex and require a more complicated
notation. To develop the new notation, we must turn again to the notion of a
propositional function.
All propositional functions considered up to now have had as substitution
instances either singular propositions or truth-functional compounds of singular
propositions having the same subject terms. If we consider a compound
statement whose components are singular propositions having different subject
terms, such as *Fa *Gb\ we can regard it as a substitution instance either of the
propositional function ‘Fx • Gby or of the propositional function ‘Fa-Gx’. Some
propositional functions, we see, may contain singular propositions as parts. And
if we consider a compound statement of which one component is a general
proposition and the other component is a singular proposition, such as
“If all dogs are carnivorous, then Rover is carnivorous’, symbolized as
\x)(Dx D Cx) D GV, we can regard it as a substitution instance of the propo¬
sitional function ‘(x)(Dx D Cx) D Cx. Thus we see that some propositional
functions may contain general propositions as parts.
At this point, two new technical terms may properly be introduced. An
occurrence of the variable ‘Y, which does not occur within or lie within the
scope of a universal or existential quantifier8 ‘(x)’ or ‘(3x)\ will be called a free
occurrence of that variable. On the other hand, an occurrence of the variable
V, which is either part of a quantifier or lies within the scope of a quantifier
‘(x)’ or ‘(3x)’, will be called a bound occurrence of that variable.9 Thus in the
expression ‘(x)(Dx D Cx) D GY, the first occurrence of the variable ‘x’ is part of
a quantifier and is, therefore, considered to be bound. The second and third
occurrences are bound occurrences also. But the fourth occurrence is a free
occurrence. Thus we see that propositional functions may contain both free
and bound occurrences of variables. On the other hand, all occurrences of
variables in propositions must be bound, since every proposition must be either
true or false. A propositional function must contain at least one free occurrence
of a variable, but no proposition can contain any free occurrences of any
variable.
The proposition ‘Fa-Gb’ can also be regarded as a substitution instance of
‘Fx*Gi/’, where the latter is a propositional function containing two different
variables. Up to now, we have explicitly admitted only one individual variable,
the letter ‘x’. However, in our previous use of the letter *y to denote any
arbitrarily selected individual, we were, in effect, using it as a variable without
acknowledging this fact. In introducing a letter by El to denote some particu¬
lar individual having a specified attribute without knowing which individual
was denoted by it, we were, in effect, using that letter as a variable, also. We

8 As explained on page 67.

9 An alternative, less common nomenclature refers to free variables as ‘real’ variables, and to
bound variables as ‘apparent’ variables.
Sec. 4.4] Multiply-General Propositions 85
now proceed to acknowledge explicitly what was implicit in our former usage.
Some propositional functions may contain two or more different individual
variables. It will be convenient to have a larger supply of individual variables
available, so we readjust our notational conventions to include the letters V,
V, ‘u>\ Y, 'y\ and 'z as individual variables. Propositional functions now
include such expressions as TV, 'Fu v Ctv, '(Fx-Gy) D Hz, ‘Fxv (Gyllx)',
and the like.
In substituting constants for variables to obtain a proposition from a propo¬
sitional function, the same constant must be substituted for every free occur¬
rence of the same variable. Thus among the substitution instances of the
propositional function ‘Fx v (Gyllx)' are

Fa v (Cb • Ha), Fa v (Cc • Ha), Fa v (Cd • I la ),...

Fh v (Ca • Hb), Fb v (Cc • Hb), Fb v (Cd-Hb),...
Fc v (Ca • lie), Fc v (Cb • lie), Fc v (Cd • lie),.. .

but not such propositions as KFa v (Cb •!!()'. On the other hand, the same
constant can be substituted for free occurrences of different variables, pro¬
vided, of course, that if it is substituted for any free occurrence of a variable, it
must be substituted for all free occurrences of that variable. Thus additional
substitution instances of the propositional function 'Fxv (Cyllx)' are
'Fa v (Ca-Ha)’, 'Fb v (Cb-Hb)\ 'Fc v (Ccllcf,...
Having admitted the letters *u\ *v\ *w , Uj , and ‘z as individual variables in
addition to Y, we now adjust our notation for universal and existential
quantification to conform to our expanded stock of variables. The proposition
"All F’s are GY may be alternatively symbolized as \u)(Fu D Cu)\
\v)(Fv 7D Cv)\ \w)(Fw D Gw)\ \x)(Fx D Cx)\ \i/)(Fy D Gy)', or
'(z)(Fz D Cz)\ Similarly the proposition, ‘There are some //V, may be alterna¬
tively symbolized as ‘(3u)Hu\ '(3v)hv\ '(3w)Ilw, (3x)Hx, (3y)Hy^ or
‘(3z)llz. The difference between \x)Fx and \\f)Fij [as between \3x)Cx and
‘(3f/)Gi/ ] is purely notational and either may be written in place of the other.
Of course, where a propositional function contains free occurrences of two or
more different variables, such as Fx-Cy\ the two propositional functions that
result from quantifying it differently as

(x)( Fx • Gy) and (y)( Fx • Gy)

are very different indeed, and their difference is moie than merely notational.
The substitution instances of the first are

(x)(Fx-Ga), (x)(Fx• Cb), (x)(Fx-Cc),...

whereas the substitution instances of the second aie

(tj)(Fa • Gy), (y)(Fb • Gy). (y)( Fc • Gy),...

If every individual has the attribute F, and some, but not all, individuals have
the attribute G, then some substitution instances of the first will be true
propositions, whereas all substitution instance's- of the second will be false, a
considerable difference. This example should serve to indicate the need for
speaking, not of “the universal (or existential) quantification of a propositional
function*, but rather of “the universal (or existential) quantification of a
propositional function with respect to the variable Y ** or “the universal (or
existential) quantification of a propositional function with respect to the
variable ‘y ”, and so on.
It should be clear that since \x)(Fx D Gx)’ and '(y)(Fy D Gy)9 are alternative
translations of the proposition ‘Everything that is an F is also a G\ the
universal quantification of ‘Fx D Gx with respect to Y has the same meaning
and is logically equivalent to the universal quantification with respect to ‘i/* of
the propositional function that results from substituting *y for every free
occurrence of Y in *Fx D Gx—for the result of that substitution is ‘Fy D Gy\
In the early stages of our work, it will be desirable to have at most one
quantification with respect to a given variable in a single proposition. This is
not strictly necessary, but it is helpful in preventing confusion. Thus the first
multiply-general proposition considered, ‘If all dogs are carnivorous, then some
animals are carnivorous’, is more conveniently symbolized as \x)(Dx D Cx) D
(3y)(AyCy)’ than as \x)(Dx D Cx) D (3x)(Ax#Cx)\ although neither is in¬
correct.
It has been remarked that no proposition can contain a free occurrence of
any variable. Hence in symbolizing any proposition, we must take care that
every occurrence of every variable used lies within the scope of a quantifier
with respect to that variable. Some examples will help to make the matter
clear. The proposition

If something is wrong with the house, then everyone in the house

complains.

is properly symbolized as a conditional whose antecedent and consequent

contain different quantifiers;.

(3x)(.r is wrong with the house) D (y) [(y is a person in the house)
D (y complains)]

Here the scope of the initial quantifier does not extend past the main implica¬
tion sign. But if we turn now to another proposition which bears a superficial
resemblance to the first:

If something is wrong, then it should be rectified,

it would be incorrect to symbolize it as

(3x)(% is wrong) D (x should be rectified)

Sec. 4.4] Multlply-General Propositions 87
For since the scope of the initial quantifier ends at the implication sign, the
occurrence of x in the consequent cannot refer back to the initial quantifier
because it does not lie within the scope of the quantifier. We have here a free
occurrence of a variable, which means that the proposed symbolization is not a
proposition and, therefore, not an adequate translation of the given statement.
The error is not to be corrected by simply extending the scope of the initial
quantifier through rcbracketing, for the symbolic expression,

(3a)|(.y is wrong) D (x should be rectified)]

although a proposition, does not have the same meaning as the original
proposition in English. Instead, it says merely that there is at least one thing
that should be rectified if it is wrong. The sense of the English sentence,
however, is clearly that anything which is wrong should be rectified. Hence a
correct symbolization is neither of the preceding, but rather

(.v)[(.v is wrong) D (.v should be rectified)]

The situation is more complicated, but no different in principle, when one

quantifier occurs within the scope of another quantifier. I lore the same warning
against dangling or unquantified variables must be sounded. The proposition

If something is missing, then if nobody calls the police, someone will be

unhappy.

is properly symbolized as

(3.y;(.y is missing) D {(y)[(y is a person) ~) —(1/ calls the police)]

D (3c)](^ is a person)m(z will be unhappy)]}

But the following proposition, which is superficially analogous to the preced-

ing:

If something is missing, then if nobody calls the police, it will not be

recovered.

is not to be symbolized as

(3.x)(.v is missing) D {(</)[([/ is a person) D ~~(y calls the police)] D

—(.v will be recovered)}

for the last occurrence of the variable V is outside the scope of the initial
quantifier and is left dangling. It cannot be corrected simply by rebracketing,
as

(3x){(x is missing) D {([/)[(// *s a person) J ~~(y calls the police)] D

'—(x will be recovered)}}
88 Quantification Theory [Ch. 4

for this expression fails equally to preserve the sense of the English sentence, in
the same way as in the previous example. -That sense is expressed by the
formula

(*){(* is missing) D {(y)[(y ifs a person) D —(y calls the police)] D

~~(x will be recovered)}}

which is, therefore, a correct symbolization of the given proposition.

EXERCISES

Symbolize each of the following propositions. In each case, use the suggested
notation and make the symbolic formula parallel the English as closely as possible:
*1. If anything is damaged, someone will be blamed. (Dx: x is damaged. Px: x is^
person. Bx: x will be blamed.)
2. If anything is damaged, the tenant will be charged for it. (Dx: x is damaged. Cx:
x will be charged to the tenant.)
3. If nothing is damaged, nobody will be blamed. (Dx: x is damaged. Px: x is a
person. Bx: x will be blamed.)
4. If something is damaged, but nobody is blamed, the tenant will not be charged
for it. (Dx: x is damaged. Px: x is a person. Bx: x is blamed. Cx: x will be charged
to the tenant.)
*5. If any bananas are yellow, they are ripe. (Bx: x is a banana. Yx: x is yellow. Rx: x
is ripe.)
6. If any bananas are yellow, then some bananas are ripe. (Bx: x is a banana. Yx: x
is yellow. Rx: x is ripe.)
7. If any bananas are yellow, then if all yellow bananas are ripe, they are ripe. (Bx:
x is a banana. Yx: x is yellow. Rx: x is ripe.)
8. If all ripe bananas are yellow, some yellow things are ripe. (Rx: x is ripe. Bx: x
is a banana. Yx: x is yellow.)
9. If all officers present are either captains or majors, then either some captains are
present or some majors are present. (Ox: x is an officer. Px: x is present. Cx: x is
a captain. Mx: x is a major.)
*10. If any officer is present, then either no majors are present or he is a major. (Ox: x
is an officer. Px: x is present. Mx: x is a major.)
11. If some officers are present, then if all officers present are captains, then some
captains are present. (Ox: x is an officer. Px: x is present. Cx: x is a captain.)
12. If some officers are present, then if all officers present are captains, then they are
captains. (Ox: x is an officer. Px: x is present. Cx: x is a captain.)
13. If all survivors are fortunate and only women were survivors, then if there are any
survivors, then some women are fortunate. (Sx: x is a survivor. Px: x is fortunate.
Wx: x is a woman.)
14. If any survivors are women, then if all women are fortunate, they are fortunate.
(Sx: x is a survivor. Wx: x is a woman. Fx: x is fortunate.)
*15. If there are any survivors and only women are survivors, then they are women.
(Sx: x is a survivor. Wx: xis a woman.)
16. If every position has a future and no employees are lazy, then some employees
will be successful. (Px: x is a position. Fx: x has a future. Ex: x is an employee. Lx:
x is lazy. Sx: x will be successful.)
Sec. 4.5] Quantification Rules 89

17. If any employees are lazy, then if some positions have no future, then they will
not be successful. (Ex: x is an employee. Lx: x is lazy. Px: x is a position. Fx: x
has a future. Sx: x will be successful.)
18. If any employees are lazy and some positions have no future, then some em¬
ployees will not be successful. (Ex: x is an employee. Lx: x is lazy. Px: x is a
position. Fx: x has a future. Sx: x will be successful.)
19. If any husband is unsuccessful, then if all wives are ambitious, then some wives
will be disappointed. (Ilx: x is a husband. Sx: x is successful. Wx: x is a wife. Ax:
x is ambitious. Dx: x will be disappointed.)
20. If any husband is unsuccessful, then if some wives are ambitious he will be
unhappy. (Hx: x is a husband. Sx: x is successful. Wx: x is a wife. Ax: x is
ambitious. Ux: x will be unhappy.)

4.5 Quantification Rules

1. Inferences Involving Propositional Functions. In constructing a formal

proof of validity for a given argument, the premisses with which we begin and
the conclusion with which we end are all propositions. But wherever the rules
of Existential Instantiation or Universal Generalization are used, at least some
of the intermediate lines must contain free variables and will, therefore, be
propositional functions rather than propositions. Each line of a formal proof of
validity must be either a premiss, or an assumption of limited scope, or follow
validly from preceding lines by an elementary valid argument form accepted as
a Rule of Inference, or follow from an immediately preceding sequence of lines
by the principle of Conditional Proof. Three questions naturally arise at this
point: In what sense can a propositional function be said to follow validly from
other propositional functions? In what sense can a propositional function be
said to follow valuily from propositions? And in what sense can a proposition
be said to follow validly from propositional functions?
To answer these questions, we shall introduce a revised, more general sense
of the word Valid’. Propositional functions contain free variables and are,
therefore, neither true nor false. But a propositional function is made into a
proposition by substituting constants for all of its free variables, and the
resulting substitution instance is either true or false. One propositional function
can be said to follow validly, as conclusion, from one or more other proposi¬
tional functions, as premisses, when every substitution of constants tor free
occurrences of variables in it (the same constants being substituted for the
same variables in both premisses and conclusion, of course) results in a valid
argument. For example, the propositional function 'Ox follows validly from
the propositional functions 'Fx D GY and 'Fx\ because every substitution of a
constant for Y results in an argument of the form Modus Ponens. We may say
of such an inference that it is valid by the principle of Modus Ponens despite
the fact that propositional functioas, rather than propositions, are involved. It
should be clear that any inference is valid that proceeds by way of any of the
nineteen Rules of Inference on our original list, regardless of whether the
premisses and conclusion are propositions or propositional functions. Iij pass-
90 Quantification Theory [Ch. 4

ing, we may note that this is so, even when the conclusion contains more free
variables than are present in the premisses. For example, by the principle of
Addition, we validly infer the propositional furtction of two variables ‘Fx v Gy'
from the propositional function of one variable ‘Fx'.
The original list of nineteen Rules of Inference also permits the inference of
propositional functions from propositions. By the principle of Addition, for
example, we can infer the propositional function ‘Fa v Gx' from the proposi¬
tion ‘Fa'. That such inferences as these are valid, in the new sense explained, is
obvious. Moreover, propositions can validly be inferred from propositional
functions by our Rules of Inference. By the principle of Simplification, for
example, we can infer the proposition ‘Fa' from the propositional function
‘Fa*Gx'. Hence the letters ‘p‘if, V, V in our nineteen Rules of Inference
now range over, or represent, either propositions or propositional functions.
We can now adopt a more general definition of formal proof of validity,
which parallels our earlier definition exactly except that lines of a proof can be
either propositions or propositional functions. If each line after the initial
premisses follows validly from preceding lines, in the generalized sense of
‘valid’ already explained, then the last line validly follows from the initial
premisses. If our initial premisses and conclusion are propositions rather than
propositional functions, then the conclusion validly follows from the initial
premisses, in the original sense of ‘valid’ which applies to arguments whose
premisses and conclusions are all statements or propositions. The following
considerations demonstrate this conclusion. As we go from our original prem¬
isses to propositional functions—if we go validly—then, if the premisses are
true, all substitution instances of the inferred propositional functions must be
true also. And as we proceed from previously inferred propositional functions
to other propositional functions, if we proceed validly, then all substitution
instances of these other propositional functions must be true also. Finally, when
we go from validly inferred propositional functions to the final conclusion,
which is a proposition, then if we proceed validly, since all substitution
instances of the former are true, the final conclusion must be true also.
The preceding remarks require some modification to take account of as¬
sumptions of limited scope thjt contain free variables, but the modifications are
best introduced as alterations in the quantification rules themselves. The
preliminary versions of our quantification rules must be replaced in any case,
for as stated, they apply only to propositions and not to propositional functions.
The two rules for generalizing, UG and EG, must now permit the quantifica¬
tion (or binding) of free variables. The instantiation rules UI and El, on the
other hand, must now permit the freeing of bound variables to permit the
introduction of propositional functions themselves, rather than (pretended)
substitution instances of them.
In our earlier discussion of propositional functions of the single variable V,
we introduced the Greek letters phi and psi, and let ‘<t>x' and N'x’ denote any
propositional function of V, such as ‘Fx', ‘Gx', ‘Fx*llx', ‘(Fx v Gx) D Hx',...
no matter how complicated such functions might be. It will be useful to
Sec. 4.5] Quantification Rules 91

continue to use these Greek letters: typically will denote any proposi¬
tional function that contains at least one free occurrence of the variable \v\
even including those propositional functions that contain free occurrences of
other variables. Thus 4>x may denote any of the following:

Fx, Fx v Gx, Cm D //a, FwFx, {3z)(Gz = //v), ...

Similarly, ^1/ may denote any of the propositional functions

py> hJ v c;y> 3 //;/, FwFtj, (3z)(Gz ~ Ihj),...

To be able to refer to any propositional function in either of the preceding

groups, we shall introduce the Cheek letters mu and nu (>' and V) to denote
individual symbols. Thus ‘<!>j\i may denote any of the preceding propositional
functions of either V or ‘if, according as V1 is taken to denote ‘a or 'ij\
Similarly, according as ‘/a’ is taken to denote ‘a’ or 'if, 4(/.i)(<l>/i)’ will denote the
universal quantification with respect to V or ‘//’ of any of the preceding
propositional functions of V or *i/\ The existential cjuantification will be
denoted by ‘(3/a.)(<1>/a)\
For convenience, the term 4fI>/i’ will also denote either a proposition or a
propositional function that contains no free occurrence of the variable repre¬
sented by ‘/a’. In such a case, (/a)(<I>/a) and (3/a)(<1>u) will be called ‘vacuous*
quantifications, and will be equivalent to each other and to <t>/i itself. This
somewhat unnatural notation is included only for the sake ol the complete¬
ness described and proved for the axiomatic development of Quantification
Theory that will be presented in Chapter 10. We shall make no use of it until
that time.
The Greek letters phi and psi may also be used with an individual constant to
denote either propositions or propositional functions containing that constant.
Thus may symbolize any of the expressions

Fa, Fa v Ga, Gc D 7/a/, Fur Fa, (3z)(Gz ~ Ha),...

and *<!>// may denote any of the expressions

Fh, Fh v Gh, Gc D // />, Fw • Fb, (3z)(Gz = lib),. ..

By the convention to be introduced immediately below, ‘<l>p’ may denote any

expression of the two preceding groups according as V is taken to denote 4a
or This notation will be useful in reformulating our quantification rules.
It will be convenient to establish two definite conventions governing the
expressions ‘<J>/a’ and so that each expression may be used in the same sense
in the statements of all four quantification rules. The first convention is that mu
(V) denotes individual variables exclusively, whereas nu (V) can denote either
an individual variable or an individual constant. The second convention is this:
92 Quantification Theory [Ch. 4

The expression Oj\i denotes any proposition or propositional function.

The expression denotes the result of replacing every free occurrence
of n in Op by p, provided that if p is a variable, it must occur free in Op
at all places that p occurs free in Op. (If Op contains no free occurrence
of p, then Op and Op are identical. The variables p and p may, of course,
be the same. If they are, in this case too, Op and Op are identical.)

This general convention helps to prevent unwanted (that is, invalid) infer¬
ences from being allowed by our four quantification rules. How it contributes
to this end will be explained after formulating each of the four rules.

EXERCISES

I. For each formula Op (where /u is V) in the left-hand column, indicate which

formulas in the right-hand column are legitimate instances of a <1>p for the
given Op:

"1. Fx v Gx a. Fa v Gy
2. Fx v Gy b. (y)(Fy Gy)
3. Fx D (3y)Gy c. Fa • Gx
4. (y)(Fx D Gy) d. Fb v Gh
*5. Fx=(3y)(GxvHy) e. Fx D (3y)Gy
6. (x)Fx D (3y)(Cx-Hy) f. Fy v Gy
g- (y)(Fa D Gy)
h. Fy => (3y)c*!/
i. (y)(Fz D Gy)
j- Fz=(3y)(Gx-IIy)
k. (x)Fx D (ly)(Gz-IIy)
LU

*—
ill

1.
a

a
<

m. (x)(Fx D Gy)

II. For each formula (where v is ‘i/’) in the left-hand column, indicate which
formulas in the right-hand column are legitimate instances of a Op for the
given <t>v:

1. Fy a. Fa
2. Fx-Gy b. Fx
3. FyGtj c. Fy
4. Fy v (3x)(Gx D Hy) d. Fx<mX
*5. Fy v (3y)(Gc v Hy) e. Fx • Gz
6. Fy D (3y)(GyHy) f. Fx v (3x)(Gx D Hx)
g. Fxv(3y)(GcvIIy)
h. Fx D (3y)(GyIix)
i. Fz v (3x)(Gx D Hz)
j. Fx D (3y)(GyHy)
k. Fzv(3x)(Gx D Hy)
l. Fx D (3y)(GyIlz)
m. Fy v (3x)(Gx D Hy)
Sec. 4.5] Quantification Rules 93

2. Universal Instantiation. The presentation of our quantification rules

will be accompanied by examples of valid arguments which they must permit,
and also by examples of invalid arguments to be avoided by restrictions placed
on those rules. The following inferences are clearly valid:

(*)** (yXjjlv Gb) (z)(Fz D Gb) (x)(Fx~Gy) (x)[Fx-(3x)(Gx-Hy)]

/.Fa' Fa v Gb /. Fb D Gb ’ . *. Fc = Gy 9 /.Fh• (3 x)(Cx • lly)

They may generally be described as possessing the form

.'. <bi>

where /a is an individual variable, v is an individual constant, and <1 >r results

from by replacing all free occurrences of /i in <1>/j by r. Of course, there can
be no free occurrence of /a in (|u.)(<]>/a), but there may be any number of free
occurrences of /la in <1>/a. On the other hand, not every occurrence of /i in
4>/a need be free: for example, where V denotes V and ‘<1>/a’ denotes
\Fjc D (3x)(C*xv Hty)\ only the first occurrence of /a in is free. This is the
case because the second occurrence is part of the existential quantifier (3/a)

and the third is within the scope of that quantifier.

Also valid are such inferences as

(x)Fx (x)Fx (y)(Fy v Cb) (z)(Fz D Gx) ix)[Fx-(3x)(Cx-Hy))

/. Fy ’ Fx ’ Fxv Gb ’ Fx D Gx ’ . . Fz*(3x)(Gx•Hy)

which are also of the form

except that in these instances both //. and v are individual variables. Here, the
premiss (fA)(Ou) may be a proposition, but the conclusion <l>r must be a
propositional function.
Just as we count valid the inference*

(z)(Fz v Gb)
/. Fhv Gb

where the instantiating constant ‘fo’ occurs in the premiss as well as in the
conclusion, so we wish to count as valid such inferences as
94 Quantification Theory [Ch. 4

in which the instantiating variable occurs free in the premiss as well as in the
conclusion. In general, when Or is inferred from (/x)(Ojx) validly, r must occur
free in Or at every place where /x occurs free in O/x. There may be more free
occurrences of r in Or than there are free occurrences of fi in O/x, however.
There will be more occurrences whenever r occurs free in O/x. All the preced¬
ing inferences must be legitimized by our principle of Universal Instantiation.
Our first Quantification Rule, Universal Instantiation, is stated as

ui MM
The general convention governing O/x and <1 >r serves to prevent such an
erroneous inference as

(x)[(3y)(Fx= —Fy)]
.*.(3 y)(Fy=-Fy)

from being allowed by UI, because r (lif) does not occur free in Or
(\3y)(Fy ~ —Fy)’) at all places that /lx (V) occurs free in <1>/x
(‘(3y)(Fx= —Fy)'). Hence ‘(3y)(Fy = —Fy)' is not a legitimate Or for use in
applying UI where (/x)(0/x) is \x)\(3y)(Fx = ~~Fy)]\ It should be obvious that
the inference is invalid because it fails for a model or possible universe
containing some things that are F and some things that are not l \ which would
make the premiss true, whereas the conclusion, being self-contradictory, is false
for any model or possible universe.

3. Existential Generalization. Turning now to Existential Generalization,

we observe that all the following are valid inferences:

Fa Fa Fa v Gb Fa D Gb
.’.(3x)Fx9 .‘.(3y)Fy' .\(3x)(Fx v Gb) ’ .*. (3y)(Fa D Gy) ’

They may generally be described as possessing the form

Or
V. (3m)(<M
Here both premisses and conclusions are propositions. Also valid are such
inferences involving propositional functions as

Fx Fay Gy Fx D Gy Fx-Gx
(3y)Fy ’ (3.r)(Fa v Gx) ’ (3y)(Fx D Gy) ’ (3y)(Fy-Gx)

which are of the same pattern as the preceding, except that here r is a variable
instead of a constant.
Sec. 4.5] Quantification Rules 95

Our second Quantification Rule, Existential Generalization, is stated as

The general convention governing <l>/i and <l>r serves to prevent such an
erroneous inference as

Fx= —Ftj
.‘.(3 x)(Fx= -Fx)

from being allowed by EG, because v (‘tj) does not occur free in (I)/'
(‘Fx == ~~Fif) at all places that fi (V) occurs free in <l>/i (*Fx EE —Fx). Hence
‘Fx EE -Ft/ is not a legitimate <l>r from which ‘(3r)( Fx EE —Fx)' can be
obtained as (3/x)(<I>/li) by EG. It should be obvious that the inference is invalid.
Even though the premiss has substitution instances that are true, the conclusion
is self-contradictory, hence false.
With both UI and EG available, we may illustrate their use in proving the
validity of the argument

All humans are mortal.

Therefore, if Socrates is human, then some humans are mortal,

by the following conditional proof:

1. (x)(HxDMx) /.’.//.* E) (d.v)i

2. Hs
3. Us D Ms 1, Ut
4. Ms 3, 2, M.P.
5. IIs9Ms 2, 4, Conj.
6. (3x)(lIx-Mx) 5, EG
T.Usl^3x)(Ilx-Mxj 2-6, C.P.

A simple pattern of deduction sudices to establish the validity of any argument

of the form

VFX*/1)

Here the pattern of proof is

1. (M)(‘V)
2. 1, UI
3. (3/i2. EG
96 Quantification Theory [Ch. 4

4. Existential Instantiation. Before we discuss our new formulation of

Existential Instantiation, we shall establish the logical truth of equivalences of
the form

(E) Dp) = D p]

where v occurs free in <!>*> at all and only those places that jut occurs.free in <l>/i,
and where p contains no free occurrence of the variable v. In case p is a true
proposition, or for any true substitution instance of p, if p is a propositional
function, both sides of the equivalence must be true. This is so because if p is
true, then D p is true for every value of v, whence (r)(4>^ D p) is true. The
truth of (3/x)(4>/i) D p also follows immediately from the truth of p. In case p
is a false proposition, or for any false substitution instance of p if p is a
propositional function, and (p)(<\>v D p) is true, every substitution instance of
$iO p is true and so every substitution instance of <1>p must be false, whence
(3/i)(0/i) must be false and so (3/x)(4>/i) D p is true. In case p is a false proposi¬
tion, or for any false substitution instance of p if p is a propositional function,
and (3/i)(<I>/i) D p is true, (3/i)(4>/i) must be false. Thus every substitution
instance of Or must be false, whence every substitution instance of Or D p
must be true. Therefore, (r)(Or D p) is true also. This argument establishes the
truth of every equivalence of the form (E) because if p is true, both sides are
true. If p is false, each side implies the other.
Turning now to Existential Instantiation, we want to permit going from
(3 x)Fx to Fx or Fy only under very stringent restrictions. Not only must the
instantiating variable not have any prior free occurrence (as discussed on pages
74-75), but we must not permit (x)Fx to be inferred from (3x)Fx by way of
Existential Instantiation and Universal Generalization. There are many ways of
imposing such restrictions. One method is to formulate the rule of Existential
Instantiation in such a way that the formula or line finally inferred by its means
contains no free variables that are introduced by it. The feasibility of this
procedure can be seen by the following considerations.
Here, as in earlier sections, we are concerned with constructing proofs of
validity only for arguments whose premisses and conclusions are propositions.
We are not concerned with propositional functions that contain free variables.
Hence we never end a proof with a propositional function that contains a free
variable. Therefore, in any proof of validity in which the rule of Existential
Instantiation involves going from (3/i)(0/i) to <J>y, the propositional function Or
serves only to permit the subsequent inference of a formula that does not
contain any free occurrence of the variable v.
Suppose we already have (3jti)(0/z) as a line in a proof. Suppose further that
in the presence of other lines already obtained, if we also had Or, we would be
able to derive a desired formula p that contains no free occurrence of the
variable r. We can proceed by writing down Or as an assumption of limited
scope. Then, after p has been derived, we can close the scope of the assumption
and infer the formula Or Dp by the strengthened Rule of Conditional Proof.
Sec. 4.5] Quantification Rules 97

From this line (subject to reasonable restrictions), the formula (r)(4>i' D p) can
be inferred by Universal Generalization. And from the latter formula, by
equivalence (E), we can infer (3p)(<f>p) D p. Now, from this formula and the
earlier line (3/x)(4>p), we can obtain p by Modus Ponens. This whole process can
be represented schematically as

i. (3/i)(<t>m)
—> j- $*1’ (where v is a variable and has no free
occurrence in any preceding line)

k. p (where p contains no free occurrence of v

k + 1. 4j—k, C.P.
k+2. W(#Op) k + 1, UG
k+3. (3/i)(4>jll) Dp k+2, Equivalence (E)
k+4. p k+3, i, M.P.

The preceding discussion can be regarded as providing an informal justifi¬

cation for the rule of Existential Instantiation, which we now state as

El: (3]*)(<V)
i—* <!>/'

provided that v is a variable and does not occur free either in p or in any line
preceding

Before discussing the restrictions in the statement of El, it may be helpful to

present a proof of validity that makes use of the new rule:

1. (x)(Fx D Gx)
2. (3y)Fy A‘.(3 z)Gz
3. Fu
4. Fu D Gu 1, UI
5. Gu 4, 3, M.P.
6. (3z)Gz 5, EG
7. (3z)Gz 2, 3-6, El

There can be, though there need not be, other lines intervening between the
premiss (3ja)(4>p) (line 2) and the propositional function <\>v (line 3) which is
98 Quantification Theory [Ch. 4

marked as an assumption of limited scope. When the desired formula p (line 6)

is reached, the next line is simply p again. The justification written next to this
line is the number of the line consisting of (3ja)(0ja),ithe hyphenated numbers of
the lines consisting of Or and the first line consisting of p, and, finally, the
notation El.
The general convention that governs Oja and Or serves to prevent such an
erroneous “proof” as

1. (3x)(Fx*Gx) /.*. (x)Fx

— 2. Fx-Gy (wrong as part of El)
3. Fx 2, Simp.
4. Fx 1, 2-3, El (wrong)
5. (x)Fx 4, UG

from being allowed by El. This is so because r ('y') does not occur free in Or
(‘Fx'Gy ) at all places that ja (V) occurs free in Oju, (‘Fx*Gx'). Hence *FxmGy' is
not a legitimate Or for use in applying El, where (3ja)(Oja) is *(3x)(Fx mGx)\ It
should be obvious that the argument is invalid. That something is both an F and
a G clearly does not entail that everything is an F
The restriction that r does not occur free in any line preceding Or, ensures
that r does not occur free in (3ja)(Oja). Therefore, if r is different from ja, v does
not occur free in Oja either. The general convention already assures us that
there are no free occurrences of ja in Or. Thus the present restriction on El,
together with the general convention, entails the restriction associated with the
logical equivalence (E), namely, that r occurs free in Or at all and only those
places that p occurs free in Oja.
The restriction that r does not occur free in p serves to prevent any
subsequent and erroneous use of Universal Generalization to derive (ja)(Oja) as
conclusion, from (3ja)(Oja), as premiss. Without the restriction, such an error
would be possible because if p were allowed to contain a free occurrence of
r, it could be Or itself, from which UG could be used to derive the conclusion

The restriction that r does notaccur free in any line preceding Or serves to
prevent such an erroneous “proof” as

1- (x)(3y)(Fx = ~Fi/) /.*. (3x)(Fx= ~~Fx)

2. (3y)(Fx = —Fy) 1, UI
—»3. Fx= ~~Fx (wrong as part of El)
4. (3x)(Fx= —Fx) 3, EG
5. (lx)(Fx= ~~Fx) 2, 3-4, El (wrong)

from being allowed by El. This js so because v (V) occurs free in line 2, which
precedes Or (*Fx = ~~Fx’) in line 3. Hence ‘Fx = ~~Fx* is not a legitimate Or
for the given Oja for use in applying El, where (3ja)(Oja) is ‘(3y)(Fx = ~*~Fy)\ It
Sec. 4.5] Quantification Rules 99

should be obvious that the argument is invalid, since it fails for a model
containing some things that are F and some things that are not F, which would
make the premiss true, whereas the conclusion, being self-contradictory, is false
for every model.

5. Universal Generalization. The complicated and highly restricted for¬

mulation of El just given permits a somewhat less restricted formulation of the
rule of Universal Generalization, which we now state as

<l>r
UG:

provided that v is a variable and does not occur free cither in (ja)(0/i) or in any
assumption within whose scope lies.

The general convention governing <l>/t and <!>/> serves to prevent such an
erroneous “proof” as

1. (3x)(y)(Fx D —Fy) /.'. (a)(Fa D —Fa)

r~> 2. (y)(Fx D ~~Fy)
3. Fx D —F?/ 2, UI
4. (x)(Fx D — Fx) 3, UG (wrong)
5 (a)(Fa D —Fa) 1, 2-4, El

from being allowed by UG, because i» (‘i/’) does not occur free in <1>y
(‘Fa D —Fy') at all places that g (‘a') occurs free in <I>/i (‘Fa D —Fa’). Hence
line 3, ‘Fa D — Fy\ is not a legitimate <l>r, for the given *!>/*, from which line 4,
‘(a)(Fa D —Fa)\ can be obtained as by UG. It should be obvious that
the argument is invalid, for its premiss is true if there is at least one :hing that
is not an F, whereas its conclusion states that there are no F’s at all.
The restriction that v does not occur free in (g)(<I>/i) serves to prevent such an
erroneous “proof” as

1. (x)(Fx Fx) /.'. (x)(y)(Fx Fy)

2. Fx = Fx 1, UI
3. (y)(Fx = Fy) 2, UG (wrong)
4. (x)(y)(Fx = Fy) 3, UG

from being allowed by UG, because v (‘a’) does occur free in (/a)(<1>/x) at line 3,
(\y)(Fx= Fy)'). Ilence line 2, *Fx=Fx\ is not a legitimate for the given
(hja, from which ‘(y)(Fx EE Fy) can be derived as ([i)(<P/i) by UG. It should be
obvious that the argument is invalid, for its premiss states only that any thing is
an F if and only if it is an F, whereas its conclusion slates that of any things x
and «/, a is an F if and only if y is an F also.
100 Quantification Theory [Ch. 4

The restriction that v does not occur free in any assumption within whose
scope Or lies, serves to prevent such an erroneous “proof” as

(3x)Fx /.\(x)Fx
Fy
(x )Fx 2, UG (wrong)
(x)Fx 1, 2-3, El

from being allowed by UG, because r (*y') occurs free in the assumption lF\j
within whose scope the premiss Or (*Fy) lies. Hence, in this case, ‘Fi/’ is not a
legitimate Or for the given O/x, from which \x)Fx* can be derived as (ju,)(0]u,) by
UG. The argument is, of course, obviously invalid.
Like their preliminary versions, these final versions of our four quantification
rules are similar to the first nine Rules of Inference in that they can be applied
only to whole lines in a proof.

EXERCISES

Identify and explain all of the mistakes in the following erroneous “proofs”:
1. r—* 1. Fx
2- (tj)Fij 1, UG
3. Fx D (ij)Fij 1-2, C.P.
4. (x)[Fx D (y)Fy] 3, UG
2. 1. (3x)(Fx-Gx) /.\(3x)Fx
|—* 2. Fx • Cx
| 3. Fx 2, Simp.
4. Fx 1, 2-3, El
5. (3x)Fx 4, EG
*3. 1. (x)(3y)(Fx=Gy) /.*. (3y)(x)(Fx= Gy)
2. (3y)(Fx = Gy) 1, UI
3. Fx ~ Gy
IN -
III

3, UG
X

5. (3y)(x)(Fx=Gy) 4, EG
jar*
'x

2? 3-5, El
vD

III

1. (x)(3y)(Fx D Gy) /: . (3y)(x)(Fx D Gy)

2. (3y)(Fx D Gy) 1, UI
—>3. Fx D Cx
4. (x)(Fx D Gx) 3, UG
5. (3;i)(x)(Fx D Gy) 4, EG
6. (3y)(x)(/<x 3 Gy) 2, 3-5, El
1. (y)(3x)(Fxv Gy) (3 x)(y)(FxvGy)
2. (3x)(Fxv Gy) 1, UI
i—>3. Fx v Gx
4. (y)(Fx v Gy) 3..UG
5. (3x)(t/)(Fx v Gy) 4, EG
6. (3x)(y)(FxvGy) 2, 3-5, El
Sec. 4.5] Quantification Rules 101

*6. 1. (lx)(y)[(Fx-Gx) D Hy] /;. (3x)[(Fx-Gx) D Ilx\

p* 2. (y)[(Fz-Gz) D Hy]
3. (Fz-Gz) D Hy 2, UI
4. (3x)[(Fx-Gz) D Hy] 3, EG
5. (y)(3x)|(Fx-Gy) 3 ljy] 4, UG

6- (y)(3x)[(Fjr-Gi/) 73 lly] 1. 2-5, El

7. (3x)[(Fx-Gx) 3 ;/.t] 6, UI
7. 1. (3x)Fx
2. (3x)Cx /.'.(3x)(Fx*Cx)
-* 3.
r-» 4. Cl/
5. FyGy 3, 4. Conj.
6. (3x)(Fx*Cx) 5, EG
7. (3x)(Fx*Gx) 2, 4-6, El
8. (3x)(Fx*Gx) 1, 3-7, El
8. 1. (3x)(3y)|(Fx v Gy)*Ily] /■’• (x)(u)(Ftj v Gx)
I-* 2. (3i/)[(Fxv Gy)’Hy]
3. (FxvGx)-Hx
4. Fx v Cx 3, Simp.
5. Fx v Gx 2, 3- 4, El

a Fx v Cx 1, 2-5, El
7. ('/)(/•>/ V Cx) 6, UG
8. (*)(?/)(% v Gx) 7, UG
*9. 1. (3 x)(Fx*Gx)
2. (3 x)(—Fx*Cx) /.\(3.t )(Fx-~ Fx)
Fx'Gy
r 4- Fx
5. Fx
3, Simp.
1, 3-4, El
f * —Fx*Gx
Ll -Fx
8. -Fx
6, Simp.
2, 6-7, El
9. Fx * - Fx 5, 8, Conj.
10. (3x)(Fx*—Fx) 9, EG
10. 1. (x)[(Fx D Cx)* ~~Ga]
r» 2. (x)[(Fx D Cx)* ~~Cy]
3. (Fx D Gz)—Gy 2, UI
4- (y)l(Fy 3 Gy) -.~G«/] 3, UG
5. (Fu D Cm)*—Cm 4, UI
6. Fu D Gu 5, Simp.
7. —Gm*(Fm D Cm) 5, Com.
8. -Cm 7, Simp.
9. -Fm 6, 8, M.T.
10. (x)—Fx 9, UG

11. (x)[(Fx D Gx)—GyJ D (x)~Fx 2 10, C.P.

12. (ti)){(x)[(Fx 3 Gx)—Gu>] D (x)~Fxj 11, UG
13. (x)[(Fx D Gx)—Ga] D (x)-Fx 12, UI
14. (x)~Fx 13, 1, M.P.
102 Quantification Theory [Ch. 4

6. Shorter Proofs of Validity. At this stage of our work, it is desirable to cut

down the length of our formal proofs of validity by. permitting shortcuts in the
application of the original list of Rules of Inference. We can combine any use
of the principle of Double Negation with any other step, w'hich will permit us
to go directly from ‘—A D B’ to ‘A v B’ or vice versa, without having to write
down the intermediate step ‘——A v B\ We can short-cut tedious uses of the
principle of Commutation by permitting not only "A . *. A v B’ by the principle
of Addition, but also such inferences as ‘A . Bv A\ Also, we permit such
inferences as ‘A v #, ~~B .'. A’ by the principle of the Disjunctive Syllogism as
well as inferences such as ‘A v B, —A B\ Since the definition of Material
Implication and the principle of Distribution can always be used to obtain
‘A D (B-C)9 from ‘(A D B)*(A D C)\ and vice versa, our demonstrations can
be shortened by permitting Distribution to be applied directly to conditionals
that have conjunctions as their consequents. This is tantamount to adding the
form *\p D (</*r)] = [(p D q)'(p D r)]’ to our list as an alternative version of
the principle of Distribution. By repeated application of the principles of
Association and Commutation, we can rearrange the terms of any conjunction
or disjunction in any way we please. Hence we can shorten our demonstrations
by omitting parentheses, brackets, and so forth, from conjunctions of any
three or more propositions. Thus such propositions as ‘A*{Bm[C*(D*E)]}\
\A-B)-[C-(D-E)]\ \A-B)-[(C-D)-E}\ '\(A-B)-C]-(D-E)\ ‘[A-(B-C)]-
(DmE)\ ‘{[(A-B)*C]*D}*£\ 4A*[(B*C)b(D*E)]\ . .. will all be written indif¬
ferently as ‘A9B*C*DmE\ and any permutation will be justified simply by the
principle of Commutation. Moreover, if we wish to infer the conjunction of
some of the indicated conjuncts, in any order, we can do so in one step and
justify it by the principle of Simplification. Thus from tA*B*C*D*E\ we can
infer *E*B'D’ in a single step. Also, the principle of Conjunction can be
applied to any number of lines to produce a new line that is the conjunction of
all of them. Finally, we shall permit the telescoping of the rules of Material
Implication, De Morgan, and Double Negation to obtain ‘—(A D B)’ from
*Am~~B\ and vice versa, which amounts to adding the form
*~~(p D q) = (p'~~q)' to our list as an alternative version of Material Impli¬
cation.
In proving the validity of some arguments, all four of our quantification rules
must be used. Consider the following moderately complex argument:

If all drugs are contaminated, then all negligent technicians are scoun¬
drels. If there are any drugs that are contaminated, then all drugs are
contaminated and unsafe. All germicides are drugs. Only the negligent
are absent-minded. Therefore, if any technician is absent-minded, then
if some germicides are contaminated, then he is a scoundrel.

By the use of fairly obvious abbreviations, it may be symbolized and proved

valid as follows:
Sec. 4.5] Quantification Rules 103
1. (x)(Dx 3 Cx) 3 (y)[(NyTy) 3 Sy]
2. (3x)(Dx-Cx) 3 (y)[Dy 3 (CyUy)]
3. (x)(C.x 3 Dx)
4.
• 5.
(x)(A* 3 Nx)
TwAu
/:. (x){(Tx-Ax) 3 U3y)(GyCu) 3 : s*n
6. (3y)(GyCy)
7. Gw • Cw
8. Gw D Dw 3, UI
9. Gw 7, Simp.
10. Dw 8, 9. M.P.
11. Cw 7, Simp.
12. Dw • Cw 10, 11, Conj.
13. (3 x)(DxmCx) 12, EG
14. (y)[Dy 3 (CyUy)] 2, 13, M.P.
15. Dz 3 (Cz-Uz) 14, IJI
16. (1Dz 3 Cz)-(Dz 3 Uz) 15, Dist.
17. Dz 3 Cz 16, Simp.
18. (x)(Dx 3 Cx) 17, UG
19. («/)[(Nyty => Sy] 1, 18, M.P.
20. (.Nu’Tu) 3 S« 19, UI
21. Au 5, Simp.
22. Aw 3 iVw 4, UI
23. Nu 22, 21, M.P.
24. Tu 5, Simp.
25. Nu • Tu 23, 24, Conj.
26. Sw 20, 25, M.P.
27. Sw 6, 7-26, El
28. (I'M'.yCy) 3 Su 6-27, C.P.
Il9. (Tu‘Au) 3 \(3y)((',yCy) 3 Sw] 5-28, C.P.
30. (x){(Tx-Ax) 3 [(3y)(C-y-Q/) 3 Sx]} 29, UG

EXERCISES

I. Construct a formal proof of validity for each of the following arguments:

*1. (x)(Ax 3 Bx) *5. (3x)Lx 3 (y)Mtj
(x)(Bx 3 Cx) 3 (Ak 3 Ck) (x)[Lx 3 (y)My]
2. (x)(Dx 3 Ex) 6. (x)(/Vx 3 Ox)
.'.Da D [(y)(Ey 3 /'?/) D M • ■ (x)fTx 3 [(y)(Py 3 Afy) 3 Ox]}
3. (x)[Gx 3 (y)(Hy 3 /?/)] 7. (x)(0x 3 ilx)
(x)Gx 3 (y)(Hy 3 ft/) (x)(Sx 3 7x)
(x)(7lx 3 Sx) 3 (y)(Qy D Ty)

4. (3x)/x 3 (3y)Ky 8. (3x)Ux 3 (i/)[((/[/ v Vij) D Wy]

• • (3x)[/x 3 (ly)Ky] (3x)t/x*(3x)Wx
(3x)(Lrx* Wx)
104 Quantification Theory [Ch. 4

9. (3x)X* D (y)(Y,j D 7aj) 13. (3xV* v (3y)K^

■ .\(3x)(Xx-Yx) D(3y)(XyZy) (x)(Jx.D Kx)
•'•(3 y)Ky
*10. (3x)Ax D (y)(By D Cy) 14. (x)(Lx D Mx)
(3.t)Dx D (3y)By (x)(Mx D Nx)
/. (3x)(Ax-Dx) D (3y)Cy (3x)Lx D (ly)Ny
11. (x)(3y)(Ex\Fy) #15. (*){Ox D [(y)(Py D 3 Rx]}
[x)Ex v (3y)Fy (x){Rx D [(i/)(Pi, D St/) D Tx]}
.■.(t/)[Pt/D((?t/-St/)]D(x)(OxDTx)
12. (3x)Cxv(y)(Cy D Hy) 16. (3x)[[/x-(t/)(Vt/ D Wt/)]
(x)(lx 3 ~~Cx) (x){t/x D'lOj/XXyV^) D Yxl)
• (x)(Gx D lx) D (y)(Gy D Hy) .'.(3t/)(XirVt/) D(3x)Yx

17. (x){Ax D l(3y)By D Cxi}

(x){Cx D \(3y)Dy D £x|}
(3x)(Bx-l<x) D [(y)(Fy D Dy) (s)(As 3 Es)]
18. (x)(3y)(Cx-lly)
. (x)Gx'(3y)IIy
19. (3x)(y)(Ix = Jy)
(»/)(3ar )(/x = /t/)
20. (x)(3y)(Kx-Ly)
(3 y)(x)(Kx-Ly)

II. (Construct a formal proof of validity for each of the following arguments. In each
case use the suggested notation and make the symbolic formulas parallel the
English as closely as possible:

1. No acrobats are clumsy. Therefore, if A1 is a waiter, then if all waiters are

clumsy, Al is not an acrobat. (Ax, Cx, Wx, a.)
2. All pets are gentle. Therefore, if any dogs are excitable and no excitable dogs
are gentle, they are not pets. (Px, Cx, Dx, Ex.)
3. All the accused are guilty. All who are convicted will hang. Therefore, if all
who are guilty are convicted, then all the accused will hang. (Ax, Cx, Cx, //x.)
*4. If there are any geniuses, then all great composers are geniuses. If anyone is
temperamental, all geniuses are temperamental. Therefore, if anyone is a
temperamental genius, then all great composers are temperamental. (Cx: x
is a genius. Cx: x is a great composer. Px: x is a person. Tx: x is temperamen¬
tal.)
5. Any car with good brakes is safe to drive and safe to ride in. So, if a car is
new, then if all new cars have good brakes, it is safe to drive. (Cx: x is a car.
Bx: x has good brakes. Dx: x is safe to drive. Rx: x is safe to ride in. Nx: x is
new.)
Either all the guests enjoyed themselves or some of the guests concealed their
real feelings. No honest person would conceal his real feelings. Therefore, if
the guests were all honest persons, then all the guests enjoyed themselves.
(Cx: x is a guest. Ex: x enjoyed himself. Cx: x conceals his real feelings. Hx: x
is honest. Px: x is a person.)
7. Any businessman who is a poet must be a wealthy man. Wealthy men are all
Sec. 4.q

conservatives. If some conservative does not like poetry, then no poets are
conservatives. Therefore, if there is a wealthy man who does not like poetry,
then no businessmen are poets. (Rx: x is a businessman. Px: x is a poet. Wx: x
is a wealthy man. Cx: x is a conservative. Lx: x likes poetry.)
*8. All radioactive substances either have a very short life or have medical value.
No uranium isotope that is radioactive has a very short life. Therefore, if all
uranium isotopes are radioactive, then all uranium isotopes have medical
value. (Rx: xis radioactive. Sx: xhas very short life. Mx: x has medical value.
\ Ux: x is a uranium isotope.)
sane witness would lie if his lying would implicate him in a crime.
Therefore, if any witness implicated himself in a crime, thpn if all witnesses
were sane, that witness did not lie. (Sx: x is sane. Wx: x is a witness. Lx: x
lies, lx: x implicates himself in a crime.)
.
10 If any jewelry is missing, then if all the servants are honest, it will be
returned. If any servant is honest, they all are. So if any jewelry is missing,
then if at least one servant is honest, it will be returned. (Jx: x is jewelry. Mx:
x is missing. Sx: x is a servant. Hx: x is honest. Rx: x will be returned.)
.
11 If there are any liberals, then all philosophers are liberals. If there are any
humanitarians, then all liberals are humanitarians. So if there are any
humanitarians who are liberals, then all philosophers are humanitarians. (Lx:
x is a liberal. Px: x is a philosopher. Hx: x is a humanitarian.)
*12. If something is lost, then if everyone values his possessions, it will be missed.
If anyone values his possessions, so does everyone. Therefore, if something is
lost, then if someone values his possessions, then something will be missed.
(Lx: x is lost. Px: x is a person. Vx: x values his possessions. Mx: x is missed.)

4.6 More on Proving Invalidity

In Section 4.3 a method was presented for proving the invalidity of invalid
arguments that involve quantifiers. There, we were dealing only with singly
general propositions. That method, however, is easily adapted to deal with
arguments that contain multiply general propositions.
Here, as before, we assume that there is at least one individual. That
assumption is satisfied by models containing one, two, or any larger number of
objects. For each such model, any given proposition containing quantifiers is
logically equivalent to a truth-functional compound of singular propositions.
Hence for each such model, any given argument whose propositions contain
quantifiers is logically equivalent to some argument that contains only singular
propositions and truth-functional compounds of singular propositions. An
argument involving quantifiers is valid if and only if, for every model, it is
logically equivalent to a valid truth-functional argument. Hence a given
argument that involves quantifiers is proved invalid by describing a model for
which the given argument is logically equivalent to an invalid truth-functional
argument. We do this by translating the given argument involving quantifiers
into a logically equivalent argument that contaias only singular propositions
and their truth-functional compounds. Then, we use the method of assigning
106 Quantification Theory [Ch. 4

truth values to prove that the truth-functional argument is invalid. The

assignment of truth values to the singular propositions that are involved
adequately describes the model that is being use<$.
We already know how to translate singly-general propositions into logically
equivalent truth-functional propositions for any given model. As explained on
page 78 for a model containing exactly k objects, a, b, c, ..., k, we use the
logical equivalences

(x)($x) = (0a'0f7*0c-...*0fc)
and
(3x)(0x) = (Oa v 0fo v 0c v ... v 0k)

There is nothing fundamentally new in translating multiply-general proposi¬

tions into logically equivalent truth-functional propositions. The translation
process is only more complicated. For example, the multiply-general proposi¬
tion

(x)Fx D (3jf)Gy

is logically equivalent for a one-element model that contains just a to

Fa D Ga

For a two-element model containing just a and b, it is logically equivalent to

(Fa ■ Fb) D (Ga v Gb)

And for a three-element model that contains a, b, and c, it is logically equiva¬

lent to

(Fa • Fb • Fc) D (GavGb v Gc)

For a multiply-general proposition with nested quantifiers, that is, some of

whose quantifiers lie within the^cope of other quantifiers, a stepwise process
will be found convenient, beginning with the outer or leftmost quantifier. For
example, although the multiply-general proposition

(x)(3y)(Fx D Gy)

is logically equivalent for the one-element model that contains just a to

Fa D Ga,

for the two-element model containing just a and b, it is logically equivalent

first to

[(ly)(Fa D GyM(3y)(Fb D Gy)]

Sec. 4.6] More on Proving Invalidity 107
and then to

[(Fa D Go) v (Fa D Gb)]-[(Fb D Ga) v (Fb D Gb)]

Some invalid arguments that contain multiply-general propositions can be

proved invalid using a one-element model. For example, the argument

(x) (ly)(Px D Qy)

(y) (3z)(Ry D Qz)
.-.(3x)(z)(FxDRz)

is logically equivalent in the one-element model that contains just a to the

truth-functional argument

Pa 3 Qa
Ra 3 Qa
Pa 3 Ra

which is proved invalid by assigning the truth value T to 7V and \)a’ and the
truth value F to 'Ra9.
As in the c:ase of arguments containing singly-general propositions, a given
argument may be valid for a one-element model but invalid for a two-element
model. For example,

(3x)(y'(Fx 3 Gy)
(ij){3z)(Gy D Hz)
W(3z)(FxDllz)

is logically ecjuivalent for the one-element model that contains just a to the
valid truth-functional argument

Fa D Ga
Ga D Ha
Fa Z) Ha

but, for the two-element model that contains just a and b, it is logically
equivalent to the truth-functional argument

[(Fa D Ga)-(Fa D Gb)] v [(Fb D Ga)'(Fb D Clb)]

[(Ga D Ha) v (Ga D Hb)]-[(Gb D Ila) v (Gb 3 Hb)]
/. [(Fa D Ha) v (Fa D Hb)]-[(Fb D Ha) v (Fb 3 Hb)J

which is proved invalid by assigning the truth value T to 'Fa9 and the truth
value F to all the others, that is, to 'Fb\ 'Ga\ 'Gb\ ‘Ha\ and lib9.
108 Quantification Theory [Ch. 4

EXERCISES

Prove that each of the following arguments is invalicj:

*1. (x)(y)[Ax 3 (By v Cy)} 6. (x)Nx 3 (3y)Oy

(z){[(y)By v (y)Cy] 3 Dz} (y)Oy 3 (3z)Pz
(3x)(3z)(Ax 3 Dz) (3x)Nx 3 (z)Pz
2. (x)(3y)(Kx 3 Fy) 7. (x)^x 3 ((3y)Ry(3y)Sy}
(3;f)(z)(Fy 3 ~~Gz) (3y)(RySy) 3 (z)7z
(x)(z)(~Ex 3 Gz) .\(x)Qx 3 (3z)Tz
3. (x)(3y)(Hx 3 ly) 8. (3x)(3y){(/x 3 [Vy(3y)Wy]}
(3y)(z)(ly 3 /z) (y)(z)((VyWy) 3 Xz]
.'. (x)ffx 3 (z)/z (x)(3z)(f/x 3 Xz)
4. D Lx) 9. (3x)(3y)(Yx 3 Zy)
(3x)(ly)(Lx-My) (3y)(z)(Zy 3 As)
.'.(y)(Kx 3 My) (3x)Yx 3 (z)As
*5. (x)(3y)(Fx=Cy) 10. (x)(y)(Bx 3 Cy)
•‘•(3!/)(x)(Fx=Gy) (x)Cx 3 [(3y)(DyEy)-(3z)(Dz--~F,z)]
(x)(Bx 3 Dx)

4.7 Logical Truths Involving Quantifiers

In Chapter 2 truth tables were used, not only to establish the validity of
arguments, but also to certify the logical truth of propositions (tautologies such
as ‘A v ~A’). The notion of a logically true proposition is, therefore, familiar.
As we have seen, not every valid argument can be established by the method of
truth tables. Some of them are proved valid by using quantification rules.
Similarly, not every logically true proposition can be certified by the method of
truth tables. Some of them are demonstrated by using quantification rules.
The method used in demonstrating the logical truth of tautologies was set
forth in Chapter 3. A demonstration of the logical truth of the tautology
‘A D (A v B)’ can be set down as

r-> 1. A
1 2. AvB 1, AddT
3. A D (A v B) 1-2, C.P.

In demonstrating the logical truth of propositions involving quantifiers, we

appeal not only to the original list of elementary valid argument forms and the
strengthened principle of Conditional Proof, but to our quantification rules
as well. Thus a demonstration of the logical truth of the proposition
<(x)Fx 3 (3x)Fx’ can be set down as

-*1. (x)Fx
2. Fy 1..UI
3. (3x)Fx 2, EG'
4. (x)Fx D (3x)Fx 1-3, C.P.
Sec. 4.7] Logical Truths Involving Quantifiers 109

( Just as in discussing the validity of arguments, so in discussing the logical truth

of propositions, we explicitly limit our consideration to possible nonempty
universes or models.)
Other logically true propositions that involve quantifiers require more
complicated demonstrations. For example, the logically true proposition
\x)Fx D —(3a:)—Fa:’ has the following demonstration:

1. (3 a:)— Fx
— 2. —Ft/
3. (x)Fx
I_4. Fy_ 3, III
5. (x)Fx D Fy 3 -4, C.P.
6. —(a*)Fv 5, 2, M.T.
7. — (x)Fx 1, 2- 6, El
8. (3a)-Fa* D -(a)Fv 1-7, C.P.
9. (a:) Ft D —(3jc)-Ft 8, Trans., D.N.

Similarly, the truth of ‘ —(3 a*)—Fx D (x)Fxf is demonstrated by the following:

— * 1. —(3 a)—Ft
-*2. — Fy
3. (3a)-Fa 2, EG
4. -Ft/ D (3 a*)—Ft 2-3, C.P.
5. Fy 4, 1, M.T., D.N.
6. (a) Ft 5, UG
7. —(3a:)—Fa* D (a)Fa* 1-6, C.P.

Given the logical truths established by the two preceding demonstrations,

we conjoin them to obtain the equivalence ‘(a) Fa ~ — (3 a*)—Fx\ a logical
truth already noted in Section 4.1. Since our proof of this equivalence does not
depend upon any peculiarities of the propositional function ‘Fx\ the equiva¬
lence holds for any propositional function. Also, since our proof does not
depend upon any peculiarities of the variable V, the equivalence holds not
only for any propositional function, but also for any individual variable! The
equivalence form (v)(<\>i’) = —(3*>)—is thus seen to be logically true, and
can be added to the other logical equivalences in our list of Rules of Inference.
It permits us (under the Rule of Replacement) validly to interchange (^)(Or)
and — (3p)—(Or) wherever they may occur. This connection between the two
quantifiers, by way of negation, will now be adopted as an additional rule of
inference and may be used in constructing formal proofs of validity and
demonstrations of logical truth. When it is so used, the letters ‘QN’ (for
quantifier negation) should be written to indicate the principle that is appealed
to. It should be obvious that the forms
110 Quantification Theory [Ch. 4

—(*')(<!>*') = (3v)~~ (4>v)

(v)^(<Pv) = ~~(3 v)(<t>v)'
= (3/^)(<I>^) \.

are all logically equivalent to each other and to the form QN, and that they
are, therefore, logically true.
Some fairly obvious logical truths are simply stated and are easily demon¬
strated with our present symbolic apparatus. A logically true biconditional for
any propositional functions TV and ‘Gx’ is

[(jr)F;f(*)Gx] = (x)(Fx9Gx)

which asserts that everything has the attribute F and everything has the
attribute G if, and only if, everything has both the attributes F and G. The
demonstrations of the two implications involved may be written side by side:

* 1.
2.
{x)Fx'(x)Gx
\x)Fx I, Simp.
rL 2.
(x)(Fx-Gx)
Fy-Gy 1, UI
3. (x)Gx 1, Simp. 3. Fy 2, Simp.
4. Fy 2, UI 4. Gy 2, Simp.
5. Gy 3, UI 5. (x)Fx 3, UG
6. FyC.y 4, 5, Conj. 6. (x)Gx 4, UG
7. (x)(Fx-Gx) 6, UG 7. (x )Fx*(x)Gx 5, 6, Conj.
8. [(jc)Fr'(jc)Gar] D (x)(FxmGx) 8. (x)(Fx*Gx) I) l(x)Fx'(x)Gx]
1-7, C.P. 1-7, C.P.

Another logical truth is in the form of a conditional rather than a bicon¬

ditional. Written as

[(.t)Frv (jc)Gx] J (x)(FxvGx)

it asserts that if either everything is an F or everything is a G, then everything

is either an F or a G. Its demonstration involves making several assumptions of
limited scope, and can be written as follows:

—*1. (x)Fx\(x)Gx
r 2. (x)Fx
3. Fy 2, UI
4. Fy v Gy 3, Add.
5. (x)(Fx v Gx) 4, UG
6. (x)Fx D (x)(Fx v Gx) 2-5, C.P.
7. (x)Gx
8. Gy 7, UI
9. Fy v Gy 8, Add.
Sec. 4.7] Logical Truths Involving Quantifiers 111

| 10. (x)(FxvGx)_ 9, UG
11. (x)Gx D (xXFxv Gx) 7-10, C.P.
12. [(x)Fx D (x)(Fx v Gx)] • [(x)Gx D (x)(Fx v Gx)] 6, 11, Conj,
13. (x)(Fx v Gx) v (xXFx v Gx) 12, 1. C.D.
14. (x)(FxvGx) 13, Taut.
15. [(x)Fx v (x)Gx] D (x)(Fx v G.v) 1-14, C.P.

I he converse of this conditional is not logically true, however. The converse

states that if everything is either F or C, then either everything is F or
everything is C. That this converse is not always true can he seen by replacing
G by , for (x)(bx v ~~}<x) is true for any predicate ‘F\ whereas there are
few predicates for which \x)Fx v (x)-Fx' holds. Another logical truth that is
conditional in form is

(3x)(Fx-Gx) D [(3a*)Et•(3xK»jc]

its demonstration is perfectly straightforward and can be left as an exercise for

the reader. That its converse is not true in general can be seen by again
replacing XV by ‘—F\ For most predicates 'F\ the proposition
\3x)Fxm(3x)~~Fx’ is true (e.g. ‘something is round and something is not
round’). However, for any *F\ the proposition '(3x)(Fx • —Fx)' is logically false.
It has already been observed that propositional functions can contain
propositions and/or other propositional functions as components. Examples of
such propositional functions are

Fx*Ga, Fx v Fy, Gy v (z)llz, Gw D Fz,...

When such propositional functions as these are quantified, to obtain

(x)[Fx-Ga], (x)[Fx v Fy], (3y)[Gi/ v (z)Uz\, (x)[Gw D Esj,...

we have propositions and/or propositional functions lying within the scopes of

quantifiers, although the quantifiers have no real affect on those expressions.
When a quantifier with respect to a given variable is prefixed to an expression,
its only effect is to bind previously free occurrences of that variable. In the
expressions written above, the propositions Xki and *(z)Hz\ and the proposi¬
tional functions *Fy and ‘Gm’—although lying within the scopes of the
quantifiers \x)\ ‘(3t/)\ \x)' and \z)\ respectively—are not really affected by
them. Wherever we have an expression that contains a quantifier on the
variable jut within whose scope lies either a proposition or a propositional
function not containing any free occurrence of g, the entire expression is
logically equivalent to another expression in which the scope of the quantifier
on fi does not extend over that proposition or propositional function. An
example or two will make this clear. In the following, let \V be either a
proposition or a propositional function containing no free occurrence of the
112 Quantification Theory [Ch. 4

variable V, and let ‘Fx' be any propositional function containing at least one
free occurrence of the variable V. Our first logical equivalence here is between
the universal quantification of ‘Fx •£)’ and the conjunction of the universal
quantification of ‘Fx’ with *Q\ which is more briefly expressed as

(x)(Fx-Q)=[(x)Fx-Q]

The demonstration of this equivalence can be written as

1. (x)(Fx-Q) -* 1. (x)Fx-Q
2. Fx-Q 1, UI 2. (x)Fx 1, Simp.
3. Fx 2, Simp. 3. Fx 2, UI
4. (x)Fx 3, UG 4. p 1, Simp.
5. (J 2, Simp. 5. Fx'Q 3, 4, Conj.
6. (x)Fx-Q 4, 5, Conj 6. (x)(Fx-Q) 5, UG
T"(x)(Fx-Q) 5 [(x)Fx-Q] 7. [(x)Fx-(J] D (x)(Fx-Q)
1-6, C.P. 1-6, C.P.

Another logical equivalence holds between the universal quantification of

D Fx* and the conditional statement whose antecedent is *Q’ and whose
consequent is the universal quantification of ‘Fx’. The first asserts that given
any individual x, Q implies that x has K and is equivalent to Q implies that
given any individual x, x has F. Our symbolic expression of this equivalence is

(x)(P 3 Fx) = [Q D (x)Fx]

Its demonstration is easily constructed:

|->1. (x)(^DFx) 1. Q 3 (x)Fx

I 2. y D Fx . 1, UI 2. Q
3. (x)Fx 1, 2, M.P.
2, 3, M.P 4. Fx 3, UI
4, UG 5. Q D Fx 2-4, C.P.
| 6, (J D (x)Fx 3-5, Off. 6. (x)(<? 3 Fx) 5, UG
7. (xXP 3 Fx) D 7. [p D JxjFx} D
K? 3 (x)Fx] 1-6, C.P. (*)(P D Fx) 1-6, C.P.

The same pattern of equivalence holds for the existential quantification of

D Fx' and the conditional statement \) 3 (3x)Fx\ The first asserts that
is at least one individual x such that {) implies that x has F, and is
equivalent to Q implies that there is at least one individual x such that x has F,
which is asserted by the second. Its demonstration is very easily constructed
and will be left as an exercise. .
However, the pattern of equivalence is different when *Q' occurs as conse¬
quent, rather than antecedent. Although the universal quantification of
‘Fx D Qf implies ‘(x)Fx D Q\ it is not implied by the latter. There is an
Sec. 4.7] Logical Truths Involving QuantNtars 113

equivalence, however, between given any x, if x has F, then Q and if there is

at least one x such that x has b, then p, which was established infonnallv as (E)
on page 96 and is expressed symbolically as

(x)(FxDQ)=[(lx)FtDQ]

Although the existential quantification of 4F.v D £)' is implied by \3xWx D Q\

it does not imply the latter. There is an equivalence, however, between there is
at least one x such that if x has F, then Q and if given any .v, x has b\ then
which is expressed symbolically as

(lx)(Fx D {))=.[(x)Fx D <;)J

The first of these two logical equivalences supplies an alternative method of

symbolizing one of the propositions discussed in Section 4.4:

If something is wrong with the house, then everyone in the house

complains.

The translation given there abbreviates to

(3*)Wx 3 (y)(Py D Cij)

which, as we have just remarked, is logically equivalent to

(x)[Wx D (y)(Py D Ci/)]

We shall conclude our discussion of logically true propositions involving

quantifiers by turning our attention to four logically true propositions that are
neither equivalences nor conditionals. They correspond, in a sense, to our
quantification rules:

1- («/)[(*)Fx D Fy]
2- (y)[Fy D (3x)Fx]
3. (3y)[Fy D (x)Fx]
4. (3y)[(3x)Fx D Fy]

The first of these corresponds to UI, and says that given any individual i/, if
everything is an F, then y is an F. Its demonstration is almost trivially obvious,
proceeding:

r* 1. (x)Fx
2. Fz 1, UI
3. (x)Fx D Fz 1-2, C.P.
4. (y)[(x)Fx D Fy] 3, UG
114 Quantification Theory [Ch. 4

The second corresponds to EG, and says that given any thing t/, if y is an F,
then something is an F. The third and fourth correspond to UG and El,
respectively. They are not so immediately obvious, »biit are logically true and
quite easily demonstrated. An intuitive explanation can be given by reference
to the ancient Athenian general and statesman Aristides, who was often called
‘the just*. So outstanding was Aristides for his rectitude that the Athenians had
a saying:

If anyone is just, Aristides is just.

With respect to any attribute, there is always some individual i/, such that if
anything has that attribute, y has it. That is what is asserted by the fourth
proposition listed above, which corresponds to El. The matter can be put
another way. If we turn our attention not to the attribute of being just, but to
its reverse, the attribute of being corruptible, then the sense of the first
Athenian saying is also expressible as

If Aristides is corruptible, then everyone is corruptible.

Again generalizing, we may observe that, with respect to any attribute, there
is always some individual y, such that if y has that attribute, everything has it.
That is what is asserted by the third proposition listed above, which corre¬
sponds to UG. Its demonstration proceeds:

p. 1. ~(3y)[Fi, D (x)Fx]
2. (y)~[% ^ (x)Fx] 1, QN
3. ~lFtJ => (x)Fx] 2, UI
4. F\1—(x)Fx 3, Impl.
5. Fy 4, Simp.
6. (x)Fx 5, UG
7. (x)Fx v (3y)[Fy D (x)Fx] 6, Add.
8. -(x)Fx 4, Simp.
9. (3y)[Fy d (x)Fx] 7, 8, D.S.
10. ~(3{/)[/'V => (x)Fx] D (3;t)[Fy D (x)Fx] 1-9, C.P.
11. (ly)[Ftj D (x)Fx] v (3y)[Fy 3 (x)Fx\ 10, Impl., D.N.
12. (3y)[Fy 3 (x)Fxj 11, Taut.

Although we shall not prove it until Chapter 10, the methods of proof so far
assembled (techniques for ‘Natural Deduction’, as they are sometimes called)
permit the demonstration of all logically true propositions that are constructed
out of truth-functional connectives and the quantification of individual varia¬
bles. It will also be proved that only propositions that are logically true can be
demonstrated by these techniques.
 Sec. 4.7] Logical Truths Involving Quantifiers IIS
EXERCISES

Construct demonstrations for each of the following. (Each occurrence of *{)’ repre¬
sents either a proposition or a propositional function containing no free occurrence of
the variable V.)

*1. MW => (3x)Fr|

2. (ar)(Z*'x 3 Gx) 3 [(x)Fx 3 (x)Gx]
3. (x)(Fx D0 = [(3x)Fx 3 0]
4. (3x)(Fx-p) = [(3x)Fx-£>]
*5. (3x)(Fx*Gx) 3 [(3x)Fx-(3x)Gx]
6. (x)(3y)(Fx 3 Gy) 3 [(x)Fx 3 (3.,)Gy]
7. (3x)(Fx 3 £) = [(x)Fx 3 p]
8. (3x)(Fx v ()) = [(3x)Fx v y]
9. [(3x)Fx 3 (3x)Gx] 3 (3x)(Fx 3 Gx)
*10. (3i/)[(3x)Fx 3 Gy]
11. |(3x)Fx v (3x)Gx] ~ (3x)(Fx v Gx)
12. (v)(Fx v()| = [(x)Fx v ]
13. (x)(3y)(Fx-Gy) = (3y)(x)(Fx-Gy)
>4:'(3x)(()3Fx) = K)3(3x)Fx]
*15. [(3x)Fx 3 (y)Gy] — (x)(y)(Fx 3 Gy)
16. [(x)Fx 3 (y)GyJ = (y)(3x)(Fx 3 Gy)
17. ](3x)Fx 3 (3y)Gy| ~ (x)(3y)(Fx 3 Gy)
18. [(x)Fx 3 (3y)Gy] ~ (3x)(3y)(Fx 3 Gy)
19. (x)[Fx 3 (3y)Gy| ~ (3y)|(3x)Fx 3 Gy]
J.. 20. (x)(3y)(Fx v Gy) ~ (3y)(x)(Fx v Gy)
The Logic of Relations

5.1 Symbolizing Relations

Some propositions containing two or more proper names (of individuals) are
rightly understood to be truth-functional compounds of singular propositions
that have different subject terms. For example, the proposition

Lincoln and Grant were presidents.

is rightly understood to be the conjunction of the two singular propositions

Lincoln was a president and Grant was a president.

But for some other propositions that have the same verbal pattern that analysis
is wholly unsatisfactory. Thus the proposition

Lincoln and Grant were acquainted.

is definitely not a conjunction or any other truth function of

Lincoln was acquainted and Grant was acquainted.

On the contrary, dividing the proposition in this way destroys its significance
because its meaning is not thatT)oth Lincoln and Grant were (or had) ac¬
quaintances, but rather that they were acquainted with each other. The given
proposition does not assert that Lincoln and Grant both had a certain attribute,
but that they stood in a certain relation. Lincoln is not said simply to be
acquainted (whatever that might mean), but is said to be acquainted with
Grant. Other propositions that express relations between two individuals are

John loves Mary.

Plato was a student of Socrates.
Isaac was a son of Abraham.
New York is east of Chicago.
Chicago is smaller than New York.

116
Sec. 5.1] Symbolizing Relations 117
Relations such as these, that can hold between two individuals, are called
‘binary* or ‘dyadic relations. Other relations may relate three or more indi¬
viduals. For example, the propositions

Detroit is between New York and Chicago.

Helen introduced John to Mary.
America won the Philippines from Spain.

express ternary or triadic relations, while quaternary or tetradic relations are

expressed by the propositions

America bought Alaska from Russia for seven million dollars.

Jack traded his cow to the peddler for a handful of beans.
Al, Bill, Charlie, and Doug played bridge together.

Relations enter into arguments in various ways. One example of a relational

argument is

Al is older than Bill.

Bill is older than Charlie.
Therefore, Al is older than Charlie.

A slightly more complex example, which involves quantification, is

Helen likes David.

Whoever likes David likes Tom.
Helen likes only good-looking men.
Therefore, Tom is a good-looking man.

A still more complex relational inference, which involves multiple quantifica¬

tion, is

All horses are animals.

Therefore, the head of a horse is the head of an animal.

The latter is a valid inference that, as De Morgan observed, all the logic of
Aristotle will not permit one to draw. Us validation by our apparatus of
quantifiers and propositional functions, will be set forth in the next section.
Before discussing proofs of validity for relational arguments (and they will
require no methods of proof beyond those developed in the preceding chapter),
the problem of symbolizing relational propositions must be dealt with. Just as
a single predicate symbol can occur in different propositions, so a single
relation symbol can occur in different propositions. Just as we have the
predicate ‘human* common to the propositions:

Aristotle is human.
Plato is human.
Socrates is human.
118 The Logic of Relations [Ch. 5

so we have the relation word, 'taught’, common to the propositions:

Socrates taught Plato.

Plato taught Aristotle.

Also, just as we regard the three subject-predicate propositions as different

substitution instances of the propositional function *x is human’, so we can
regard the two relational propositions as different substitution instances of the
propositional function ‘x taught if. Substituting the constant 'Socrates’ for the
variable x, and the constant 'Plato’ for the variable *y\ gives us the first
proposition. Substituting ‘Plato’ for V and ‘Aristotle’ for ‘i/* gives the second.
The order of replacement is of great importance here. If V has ‘Aristotle’
substituted for it, and '{/’ has ‘Plato’ substituted for it, the result is the false
proposition'

Aristotle taught Plato.

Just as a propositional function of one variable like ‘x is human’ was

abbreviated as *Hx\ so a propositional function of two variables like ‘x taught
if is abbreviated as Txj/\ Similarly, the propositional function 1x is between y
and will be abbreviated as ‘Bxyz’, and the propositional function x traded y
to £ for w' will be abbreviated as ‘Txyzw’. Our first example of a relational
argument involves no quantifiers and is very easily symbolized. Using the
individual constants V, ‘b\ and ‘c’ to denote Al, Bill, and Charlie, and the
expression ‘Oxy to abbreviate ‘x is older than y\ we have

Gab
Obc
. ’. Oac

Our second argument is not much more difficult, since none of its propositions
contains more than a single quantifier. Using the individual constants *d\
and *t ’ to denote Helen, David, and Tom, respectively, ‘Gx’ to abbreviate ‘x is
a good-looking man’, and the symbol ‘Lxy’ to abbreviate ‘x likes y\ the
argument can be symbolized as

1. Lhd
2. (x)(Lxd D Lxt)
3. (x)(Lhx D Gx)
Gt

A proof of its validity is so easily constructed that it may well be set down now.
Referring back to the numbered premisses, the proof proceeds:
Sec. 5.1] Symbolizing Relations 119
4. Lhd D Lht 2, UI
5. Lht 4, 1, M.P.
6. Lht D G* 3, UI
7. Gt 6, 5, M.P.

Symbolizing relational propositions becomes more complicated when sev¬

eral quantifiers occur in a single proposition. Our discussion of the problem will
be simplified by confining attention at first to the. two individual constants, V
and ‘b\ and the propositional function 'x attracts ij\ abbreviated as 'Axy'. The
two statements ‘a attracts W and *b is attracted by a obviously have the same
meaning. The first statement expresses that meaning by use of the active voice,
while the second statement expresses that meaning by use of the passive voice.
Both statements translate directly into the single formula *Aah\ Similarly, the
two statements ‘b attracts a’ and \i is attracted by // are both symbolized by
the formula ‘Aba’. These two different substitution instances of 'Axy* are
logically independent of each other. Either can be true, without entailing the
truth of the other.
We are still on elementary and familiar ground when we come to symbolize

a attracts everything’
as '(x)Aax\
everything is attracted by a'
a attracts something’
as \3x)Aax*.
something is attracted by a
everything attracts a
as \x)Axa\
‘a is attracted by everything’
something attracts a’
as '{3x)Axa\
a is attracted by something’

But the problem of symbolizing becomes more complex when we dispense

entirely with individual constants and consider relational propositions that are
completely general. The simplest propositions of this kind are

1. Everything attracts everything.

2. Everything is attracted by everything.
3. Something attracts something.
4. Something is attracted by something.
5. Nothing attracts anything.
6. Nothing is attracted by anything.

which are symbolized by the following formulas:

1. (x)(y)Axy
2. (y)(x)Axy
3. (3x)(3y)Axy
120 The Logic of Relations [Ch. 5

4. (3y)(3 x)Axy
5- (x)(y)~Axy
6- (y)(*)~Axy

In their English formulations, propositions 1 and 2 are clearly equivalent to

each other, as are 3 and 4, and 5 and 6. The first two equivalences are easily
established for the corresponding symbolic translations:

1. (x)(y)Axy —-* 1. (3*X3 y)Axy

2. (y)Awy 1, UI 1—* 2- (3;,)Awy
Awv
3. 2, UI
r3- Awv
4. (x)Axv 3, UG 4. (3x)Axu 3, EG
5. (y)(x)Axy 4, UG 5. (3y)(3x)Axy 4, EG
6. (*)(y)Axy D (y)(x)Axy 6. (3y)(3 x)Axy 2, 3-5, El
1-5, C.P. 7. (3y)(3 x)Axy 1, 2-6, El
8. (3x)(3y)Axy D (3y)(3x)Axy
1-7, C.P.

These demonstrate the logical truth of conditionals rather than of equivalences,

but their converses can be established by the same patterns of argument. (The
equivalence between formulas 5 and 6 is clearly established by the same
pattern of argument that proves 1 equivalent to 2.)
When we turn to the next pair of statements,

7. Everything attracts something.

8. Something is attracted by everything.

there is no longer any logical equivalence or sameness of meaning. Sentence 7

is not entirely unambiguous, and some exceptional contexts might shift its
meaning. Its most natural interpretation, however, is not that there is some one
thing that is attracted by everything, but rather that everything attracts
something or other. We symbolize it by way of successive paraphrasings,
writing first *

(x)(x attracts something)

and then symbolizing the expression ‘x attracts something' the same way in
which we symbolized ‘a attracts something’. This gives us the formula

7. (x)(3ij)Axy

Sentence 8 is also susceptible to alternative interpretations, one of which

would make it synonymous with sentence 7, meaning that something or other
is attracted by any (given) thing. But a perfectly straightforward way of
Sec. 5.1] Symbolizing Relations 121

understanding sentence 8 is to take it as asserting that there is some one thing

that is attracted by all things. It, too, is symbolized in a stepwise fashion,
writing first

(3 y)(y is attracted by everything)

and then symbolizing the expression ‘i/ is attracted by everything’ the same
way that we symbolized 'a is attracted by everything. This gives us the
formula

8. (3 y)(x)Axy

There is a certain misleading similarity between formulas 7 and 8. They both

consist of the propositional function ‘Axi/’, to which are applied a universal
quantifier with respect to V and an existential quantifier with respect to ‘i/’.
But the order in which the quantifiers are written is different in each case, and
that makes a world of difference in their meanings. Formula 7, in which the
universal quantifier comes first, asserts that given anything whatever, there is
something or other that it attracts. But formula 8, in which the existential
quantifier comes first, asserts that there is some one thing such that everything
attracts it. Where two quantifiers are applied successively to one propositional
function, if they are both universal or both existential, their order does not
matter, as is shown by the equivalence of formulas 1 and 2, 3 and 4, and 5 and
fi. But where one is universal and the other existential, the order of general¬
ization or quantification is very important.
Although formulas 7 and 8 are not equivalent, they are not independent. The
former is validly deducible from the latter. The demonstration is easily con¬
structed as follows:

1. (3 y)(x)Axy
—> 2. (jc)A.rt>
3. Atw 2, UI
4. (3y)Auy 3, EG
5. (3y)Auy 1, 2-4, El
6. (.r)(3y)Axy 5, UG

The inference is valid only one way, however. Any attempt to derive formula
8 from 7 must inevitably run afoul of one of the restrictions on UG. The
argument

(*)(3il)Axy
. (ly)(x)Axy

is easily proved invalid by the method of Section 4.6. For a model containing
just the two individuals a and b, the given argument is logically equivalent to
the truth-functional argument

(Aaa v Aah)*(Aba v Abb)

(Aaa*Aba)v(Aab*Abb)

which is proved invalid by assigning the truth value T to ‘Aad and ‘Abb\ and
the truth value F to 'Aab' and ‘Aba .
A similar pair of inequivalent propositions may be written as

9. Everything is attracted by something.

10. Something attracts everything.

These are clearly inequivalent when the ‘something’ in 9, coming at the end, is
understood as ‘something or other’, and the ‘something’ in 10, coming at the
beginning, is understood as ‘there is at least one thing’. They are symbolized as

9. (y)(3x)Axy
10. (3x)(y)Axy

Relational propositions are sometimes formulated as though they were simple

subject-predicate assertions. For example, ‘a was struck’ is most plausibly
interpreted to assert that something struck a. Such implicit occurrences of
relations are often marked by the passive voice of a transitive verb. Our
symbolizing of propositions containing implicit relations should be guided by
consideration of the use to which they are to be put. Our purpose in symboliz¬
ing arguments is to put them into a form convenient for testing their validity.
Our goal, therefore, with respect to a given argument, is not to provide a
theoretically complete analysis, but rather to provide one complete enough for
the purpose at hand—the testing of validity. Consequently, some implicit
relations may be left implicit, while others require a more thorough analysis, as
may be made clear by an example. Consider the argument

Whoever visited the building was observed. Anyone who had observed
Andrews would have remembered him. Nobody remembered An¬
drews. Therefore, Andrews didn’t visit the building.

The first proposition of this argument contains two relations, one explicit, the
other implicit. Explicitly, we have the relation of someone visiting the building.
It is explicit because mention is made of both the visitor and what was visited
by him. Implicitly, we have the relation of someone observing someone, which
is implicit because no mention is made of the someone who does the observ¬
ing—the omission being marked by the use of the passive voice. However,
because the only other occurrence of ‘x visited the building’ is also as a unit, in
the conclusion, it need not be treated as a relation at all, but may be symbolized
as a simple predicate. On the other hand, ‘x observed y\ despite its merely
Sec. 5.1] Symbolizing Relations 123

implicit occurrence in the first premiss, must be explicitly symbolized as a

relation if the argument is to be proved valid. For its second occurrence is not
a simple repetition of the original unit. It appears instead as an explicit
relation, with the first variable quantified and the second replaced by the
proper name ‘Andrews’. Using ‘a to denote Andrew's, ‘V'v’ to abbreviate \v
visited the building’, ‘Ox;/’ to abbreviate \r observed y\ and 'Rxij to abbreviate
x remembers y\ a symbolic translation and proof of validity for the given
argument may be written as

1. (x)[Vx D (ly)Oyx\
2. (x)[Oxa D Rxa]
3. (x)—Rxa /.' . ~v«
4. Oza D Rza 2, UI
5. —Rza 3, UI
6. Oza 4, 5, M.T.
7. (y)~~Oya 6, UG
8. 7, QN
9. Va D (3i/)Oya 1, UI
10. — Va 9, 8, M.T.

Our proof of validity for this argument would not have been helped at all by
symbolizing ‘Andrews visited the building’ as a substitution instance of the
relational ‘x visited if rather than of the simpler ‘ Vx\ But our proof of validity
depended upon our symbolizing ‘was observed’ explicitly as a relation.
Most of our previous examples were illustrations of unlimited generality, in
which it was asserted that everything stood in such-and-such a relation, or
something did, or nothing did. A great many relational propositions are not so
sweeping. Most assertions are more modest, claiming not that everything stands
in such-and-such a relation, but that everything does if it satisfies certain
conditions or restrictions. Thus we may say either that

Everything is attracted by all magnets,

or that

Everything made of iron is attracted by all magnets.

The second, of course, is the more modest assertion, being less general than the
first. While the first is adequately symbolized, where \ix abbreviates ‘x is a
magnet’, as

(x)(y)[My i Ai/.xj

the second is symbolized, where ‘lx’ abbreviates x is made of iron, as

(*)[/* 3 (y)(My D Ayx)]

124 The Logic of Relations [Ch. 5

which is equivalent to (x)(y)[Ix D (My D Ayx)J as was proved on page 112.

That this symbolic translation is correct can be seen by paraphrasing the
second proposition in English as

Given anything whatever, if it is made of iron, then it is attracted by all

magnets.

Perhaps the best way to translate relational propositions into our logical
symbolism is by the kind of stepwise process that has already been illustrated.
Let us use it again, this time for propositions of limited generality. First let us
consider the proposition

Any good amateur can beat some professional.

As a first step we may write

(x){(x is a good amateur) D (x can beat some professional)}

Next, the consequent of the conditional between the braces

x can beat some professional

is symbolized as a generalization or quantified expression:

(3i/)[(y is a professional) *(x can beat y)]

Now, using the obvious abbreviations, ‘Gx\ ‘Px\ and 'Bxy for ‘x is a good
amateur’, ‘x is a professional’, and ‘x can beat y\ the given proposition is
symbolized by the formula

(x)[CJx D (ly)(Py-Bxy)]

Using the same method of paraphrasing by successive steps, we may symbolize

Some professionals can beat all amateurs,

first as

(3x)[(x is a professional)*(x can beat all amateurs)]

then as

(3x){(x is a professional)•(y)[(y is an amateur) D (x can beat i/)]}

and finally (using abbreviations) as

Sec. 5.1] Symbolizing Relations 125

{lx)[Px-(y)(Ay D Bxy)]

The same method is applicable in more complex cases, where more than one
relation is involved. We symbolize the proposition

Anyone who promises everything to everyone is certain to disappoint

somebody.

first by paraphrasing it as

(.t){[(x is a person) *(.r promises everything to everyone)]

D [x disappoints somebody]}

The second conjunct of the antecedent

x promises everything to everyone

may be further paraphrased, first as

(y)[(y is a person) D (x promises everything to ;/)]

and then as

(y)[(y is a person) D (z)(x promises s to tj)]

The consequent in our first paraphrase

x disappoints somebody

has its structure made more explicit by being rewritten as

(3ti)[(ti is a person)-(x disappoints u)]

The original proposition can now be rewritten as

(x){{(x is a person)m(y)[(y is a person) D (z)(x promises to i/)]}

D [3u)[(u is a person)*(x disappints u)]}

Using the obvious abbreviations, *Px , ‘Pxyz, Dxy for x is a person, x

promises y to z\ and x disappoints y\ we can express the proposition more
compactly in the formula

(x){{Px-(y)[Py D (z)Pxzy)} D (3u)(Pu• Dxu)}

With practice, of course, not all such intermediate steps need be written out
explicitly.
126 The Logic of Relations [Ch. 5

Quantification words such as everyone’, ‘anyone’, ‘everybody’, ‘anybody’,

and ‘whoever’, refer to all persons rather than to all things. Such quantification
words as ‘someone’ and ‘somebody’ refer to some persons rather than to some
things. It is frequently desirable to represent this reference in our symbolism.
But doing so is not always necessary for the purpose of evaluating arguments
that contain these words, however, and the choice of symbols is determined on
the same grounds on which one decides if a relational clause or phrase is to be
symbolized explicitly as a relation or as a mere attribute.
The words ‘always’, ‘never’, and ‘sometimes’ frequently have nontemporal
significance, as in the propositions

Good men always have friends.

Bad men never have friends.
Men who have no wives sometimes have friends.

which may be symbolized, using obvious abbreviations, as

(x)[(Gx*Mx) D (3y)Fxy]
(x)l(Bx'Mx) D — (3y)L'xy]
(3x){ | Mx- —(3jf)(Wy-Uxy)]-(3 z)Fxz)

However, some uses of these words are definitely temporal, and when they are,
they can be symbolized using the logical machinery already available, as can
other temporal words like ‘while’, ‘when’, ‘whenever’, and the like. An example
or two should serve to make this clear. Thus the proposition

Dick always writes Joan when they are separated.

asserts that all times when Dick and Joan are separated are times when Dick
writes Joan. This can be symbolized using ‘Tv’ for ‘jc is a time’, ‘ Wxijz for ‘x
writes y at (time) z\ and ‘Sjxyz for ‘x and y are separated at (time) z\ as

{x){Tx D [Sdjx D Wcljx]}

Perhaps the most vivid illustration of the adaptability of the present notation is
in symbolizing the following remark, usually attributed to Lincoln:

You can fool sorpe of the people all of the time, and all of the people
some of the time, but you cannot fool all of the people all of the time.

The first conjunct: ‘You can fool some of the people all of the time’ is ambigu¬
ous. It may be taken to mean either that there is at least one person who can
always befooled or that, for any time, there is at least one person (or other) who
can befooled at that time. Adopting, the first interpretation and using Tx’ for ‘x
Sec. 5.1] Symbolizing Relations 127

is a person, Tx for x is a time , and 'Fxy* for ‘you can fool .v at (or during) y\
the above may be symbolized as

{(3x)[Px-(y)(Ty D Fxy)]-(3y)[Ty(x)(Px D Fxy)]}-(3i,)(3x)(Ty-Px-^Fxy)

The actual testing of relational arguments presents no new problems once

the translations into logical symbolism are effected. The latter is the more
troublesome part, and so a number of exercises are provided for the student to
do before going on.

EXERCISES

I. Using the following ‘vocabulary’, translate the given formulas into idiomatic
English sentences:

Ax-x is silver Axt/-x helps tj

Bx-x is blissful Bxy-x belongs to y
Cx-x is a cloud Bxtjz-x borrows tj from
Dx-x is a dog Cxy-x can command 1/
Ex-x is smoke Dxy-x is done at (or bv) 1/
Fx-x is fire Exy-x shears y
Cx-x is glass Fxy-x is fair for y
Hx-x is home Cxy-x gathers y
/x-x is ill llxy-x hears y
Jx-x is work Ixy-x lives in y
Kx-x is a lining Jxij-x is jack of 1/
Lx-x is a lamb Kxy-x knows //
Xlx-x is moss Lxy-x likes y
Nx-x is good Mxy-x is master of y
Ox-x is a fool Nxy-x loses y
Px-x is a person Oxty-x is judged by y
Qx-x is a place Pxtyz-x blows // to z
Rx-x rolls Qxy-x keeps company with y
Sx-x is a stone Rxy-x is like y
Tx-x is a trade Sxy-x says y
Ux-x is a house Txy-x should throw y
Vx-x is a woman Txyz-x tempers y to ^
Wx-x is wind Uxty-x comes U> y
Xx-x is a time Vxy-x ventures y
Yx-x is a day Wxy-x is al (.or in) y
Zx-x waits Xxy-x is parent of y
g-Cod

Formulav
*1. (x)[Dx 3 (3t/)(Yy-Byx)]
2. (x)[(3ij)(Py-Fxy) 3 (z)(Pz 3 1-xz)]
3. (x)[(Rx-Sx) 3 (y)(My 3 -Cxy)]
4. (*)[(?*-Axx) 3 (Anx)]
128 The Logic of Relations [Ch. 5

*5. (x)[(Px-Zx) 3 (y)((/y*)J

6. (x)[I/x 3 (y)«?y 3 -fly*)]
7. (x)[(Px-~Nxg) 3 (y)(-~Nxy)]
8. (x)[(Px'~~Cxx) 3 (y)(Py 3 ~C*y)]
9. (x){Cx 3 (3y)[(AyKy)‘Byx]}
*10. (x)[Px 3 (y)lQxy 3 Oxy)]
11. (*){Ox 3 [(3y)(Ey Wyx) 3 (3z)(Fz'Wzx)l}
12. (x){[Px‘(y)(Ty 3 Jxy)] 3 (z)(Tz 3 -Ate)}
13. (*){[P*C/y)-/xi,]] 3(s)(Ss 3 ~Txz)}
14. (x){[Px'(y)(Lxy 3 Sxy)] 3 (3z)(Hxz-~~Lxz)}
*15. (*){[W**(y)[Py 3 ~~(3z)(Nz‘Pxzy)]\ 3 lx)
16. (*){[P*-(y)(~V*y)] 3 (z)(~C;*z)}
17. (*){Vx 3 (y)[Xy 3 (3z)[(Jz-Bzx)—Dzy]]}
18. (*){[E*-(3y)(Py-Ey*)] 3 (z)(Wz 3 Tgzx)}
19. (*){P* 3(3y)[Py(3S)(/tey)]}
*20. (x){Px 3 (3i/)[P?/-(3^)(~«x^y)|}
21. (x){Px 3 (y)[Py 3 {^)(~Bxsy)l>
22. (x){Px 3 (y)[Py 3 (3^)(~Bx^y)]}
23. (x)[(Nx’Dx) 3 (y)(Lxy 3 My*)]
24. (*)[P* 3 (3u)(Py • Xyx)] • (3u)[Pu■ (v)(Pv 3 ~Xuo)]
25. (x)([Qx-(y){[(Py-Wyx)-(z)(^Kyz^ D By}) 3
(u) {[(Pu • Wux) • (v)(Kiid)] 3 Ou} }

II. Symbolize the following sentences, in each case using the indicated symbols:

*1. Dead men tell no tales. (Dx: x is dead. Mx: x is a man. Tx: x is a tale. Txy: x
tells y.)
2. A lawyer who pleads his own case has a fool for a client. (Lx: x is a lawyer.
Fx: x is a fool. Fxy: x pleads y's case. Cxy: x is a client of y.)
3. A dead lion is more dangerous than a live dog. (Lx: x is a lion. A.r: x is alive.
Dx: x is a dog. Dxy: x is more dangerous than y.)
4. Uneasy lies the head that wears a crown. (Ux: x lies uneasy. IIx: x is a head.
Cx: x is a crown. Wxy: x wears y.)
*5. If a boy tells only lies, none of them will be believed. (Bx: x is a boy. Lx: x
is a lie. Txy: x tells y. Bxy: x believes y.)
6. Anyone who consults a psychiatrist ought to have his head examined. (Px: x
is a personvSx: x is a psychiatrist. Ox: x ought to have his head examined.
Cxy: x consults y.)
7. No one ever learns anything unless he teaches it to himself. (Px: x is a person.
Lxy : x learns y. Txyz: x teaches y to z.)
8. Delilah wore a ring on every finger and had a finger in every pie. (cl: Delilah.
Rx: x is a ring. Fxy : x is a finger of y. Oxy : x is on y. Px: x is a pie. Ixy: x
is in y.)
9. Any man who hates children and dogs cannot be all bad. (Mx: x is a man. Cx:
x is a child. Djc: x is a dog. Bx: x is all bad. Hxy: x hates y.)
*10v Anyone who accomplishes anything will be envied by everyone. (Px: x is a
person. Axy: x accomplishes y. Exy: x envies y.)
fl. To catch a fish, one must ha'vp some bait. (Px: x is a person. Fx: x is a fish. Bx:
x is bait. Cxy: x catches y. Hxy: x has y.)
Sec. 5.1] Symbolizing Relations 129

12. Every student does some problems, but no student does all of them. (Sx: x
is a student. Px: x is a problem. Dxy: x does y.)
13. Any contestant who answers all the questions put to him will win any prize
he chooses. (Cx: x is a contestant. ()x: x is a question. Px: x is a prize. Ax{/: x
answers y. Pxy: x is put to y. Wxy: x wins y. Cxy: x chooses y.)
14. Every son has a father, but not every father has a son. (Px: x is a person. Six:
x is male. Pxy: x is a parent of y.)
*15. A person is maintaining a nuisance if he has a dog that barks at everyone
who visits its owner. (Px: x is a person. Nx: x is a nuisance. Mxy : x maintains
tj. Dx: a is a dog. Bxy: x barks at tj. Vxy: x visits y. Hxy: x has y.)
16. A doctor has no scruples who treats a patient who has no ailment. (Dx: x is
a doctor. Sx: x is a scruple. Ilxy: x has y. Px: x is a patient. Ax: x is an
ailment. Txy: x treats y.)
17. A doctor who treats a person who has every ailment has a job for which no
one would envy him. (Dx: x is a doctor. Px: x is a person. Txy: x treats y. Ax:
x is an ailment. Hxy: x has y. Jx: x is a job. Fxyz: x envies y his z.)
18. If a farmer keeps only hens, none of them will lay eggs that are worth
setting. (Fx: x is a farmer. Kxy: x keeps //. IIx: x is a hen. Fx: x is an egg. Lxy :
x lays y. Wx: x is worth setting.)

In symbolizing the following, use only the abbreviations: Px: x is a person. Sx:
x is a store, Bxyz: x buys y from

19. Everyone buys something from some store (or other).

*20. There is a store from which everyone buys something (or other).
21. Some people make all their purchases from a single store.
22. No one buys everything that it sells from any store.
23. No one buys things from every store.
24. No store has everyone for a customer.
*25. No store makes all its sales to a single person.

In symbolizing the following, use only the abbreviations: Cx: x is a charity; Mx:
x is money; Px: x is a person; Bxy: v belongs to i/; Dxyz: x donates »/ to z.

26. Nobody donates to every charity.

27. Nobody donates money to every charity.
28. Nobody donates all of his money to charity.
29. Nobody donates all of his money to any single charity.
*30. Nobody donates all of his belongings to any single charity.
31. Nobody gives all of his donations to any single charity.
32. No charity receives all of its money from any single person.
33. No charity receives all of his money from any single person.
34. No charity receives all of his donations from any single person.
*35. No charity receives all of his donations from anv one donor.
36. No charity receives only money as donations.
37. Somebody gives money to charity.
38. Somebody donates all of his money to charity.
39. At least one person donates all of his belongings to a single charity.
*40. At least one person gives all of his donations to a single charity.
41. Some charities receive donations from everybody.
130 The Logic o! Relations [Ch. 5

42. Some charities receive donations from every donor.

43. Some donations are not given to a charity.
44. Some donors donate to every charity.
45. Every charity receives donations from at leatft one donor.

III. Symbolize each of the following:

*1. Whoso sheddeth man’s blood, by man shall his blood be shed. (Genesis 9:6)
2. His hand will be against every man, and every man’s hand against him.
(Genesis 16:12)
3. The sun shall not smite thee by day, nor the moon by night. (Psalm 121:6)
4. A wise son maketh a glad father. (Proverbs 10:1)
*5. He that spareth his rod hateth his son. (Proverbs 13:24)
6. The borrower is servant to the lender. (Proverbs 22:7)
7. Whoso diggeth a pit shall fall therein: and he that rolleth a stone, it will
return upon him. (Proverbs 26:27)
8. The fathers have eaten sour grapes, and the children’s teeth are set on edge.
(Ezekiel 18:2)
9. The foxes have holes, and the birds of the air have nests; but the Son of man
hath not where to lay his head. (Matthew 8:20)
10. ... the good that I would, I do not; but the evil which 1 would not, that l
do. (Romans 7:19)

5.2 Arguments Involving Relations

No new principles need be introduced to deal with relational arguments.

The original list of nineteen Rules of Inference, together with the strengthened
method of Conditional Proof, and our quantification rules plus Quantifier
Negation, enable us (if we have sufficient ingenuity) to construct a formal proof
of validity for every valid argument in which only individual variables are
quantified and only truth-functional connectives occur.
However, a certain change of technique is advisable in working with
arguments involving relations. In most of our previous sample proofs and
demonstrations, UI and El were used to instantiate with respect to a variable
different from any quantified in tfte premiss, and UG and EG were used to
quantify with respect to a variable different from any that occurred free in the
premiss. Our inferences were for the most part of the following forms:

(3 x)Fx
(x)Fx [— Fx Fy
7~F^' :() ’ :.(y)Fy' (3w)Fw

■■p

But our statement of the quantification rules does not require that /x and v be
different variables; they may well be the same. On the whole, it is simpler
Sec. 5.ZJ Arguments Involving Relations 131

(wherever it is legitimate) to instantiate with respect to the same variable that

had been quantified, and to quantify with respect to the same variable that had
been free in the premiss. Thus the above inferences may also take the following
forms:

(3jc )Fx

(x)Fx - Fx Fx Fij
■'■Fx' I p ’ .■.(x)Fx’ .\&y)Fy

■ P

In this way instantiation is accomplished by simply dropping a quantifier, and

generalization is accomplished by simply adding a quantifier. Of course, our
restrictions on the quantification rules must still be observed. For example,
where we have two premisses, ‘(3.r)Fx' and ‘(3x)^Fx\ we can begin our use of
El with respect to one by simply dropping the quantifier. When that has been
done, however, if within the scope of that assumption l'x\ El is subsequently
used on the other, a new variable must be used instead of \v\ for the lat ter will
already have a free occurrence in the proof under construction. Of course, we
remain perfectly free to use UI to instantiate with respect to any variable or
constant we choose. The preceding remarks can be illustrated by constructing
a proof of validity for the argument

There is a man whom all men despise.

Therefore, at least one man despises himself.

Its symbolic translation and proof, using 'Mx' and 'Dxif to abbreviate \v is a
man’ and ‘jc despises // may be written as follows:

1. (3x)[Mx •(y)(My D Dyx)] /.' . (3x)(Mx • Dxx )

XIx • D Dyx)
r2-
3. 3 Dyx) 2, Simp.
4. Xlx D Dxx 3, UI
5. Mx 2, Simp.
6. Dxx 4, 5, M.P.
7. Mx • Dxx 5, 6, Conj.
8. (3 x)(Mx-Dxx) 7, EG
9. (3x)(Mx*Dxx) 1, 2-8, El

In the foregoing proof, the only use of a quantification rule accompanied by a

change of variable was the use of UI in going from line 3 to line 4. This was
done because we needed the expression 'Dxx thus obtained.
Another formal proof serves to establish the validity of the third specimen
argument stated at the beginning of the present chapter. Its premiss, ‘All horses
132 The Logic of Relations [Ch. 5

are animals* will be symbolized as \x)(Ex D Ax)’, where ‘Ex’ and ‘Ax’ abbre¬
viate *x is a horse’ and ‘x is an animal*, respectively. In its conclusion

The head of a horse is the head of an animal.

the word ‘the’ has the same sense that it does in such propositions as The whale
is a mammal* or The burnt child dreads the fire*. We may paraphrase it,
therefore, first as

All heads of horses are heads of animals,

then as

(x)[(x is a head of a horse) D (x is a head of an animal)]

and finally, writing ‘Hxy’ for *x is a head of y\ we may express the conclusion
by the formula

(x)[(3y)(Ey-Hxy) D (3y)(AyHxy)]

Once it is symbolized, the argument is easily proved valid by the techniques

already available:

1. (x)(Ex D Ax) /:. (x)[(3y)(Ey'IIxy) D (3y)(Ay-Uxy)]

—* 2. (3 ij)(Eyllxy)
3. Ey • Hxy
4. % 3, Simp.
5. Ey D Ay 1, UI
6. Ay 5, 4, M.P.
7. Hxy 3, Simp.
8. Ay’Hxy 6, 7, Conj.
9. (1y)(AyHxy) 8, EG
10. (3y)(AyHxy) 2, 3-9, El
rt i a n n
11. (ly)[EyUxy) D (3y)(AyHxy)
12. (x)[(3y)(EyHxy) D (3y){AyHxy)\ 11, UG

The first specimen argument presented in this chapter, which dealt with the
relation of being older than, raises a new problem, which will be discussed in
the following section.

EXERCISES

I. Construct a formal proof of validity for each of the following arguments:

*1. (3x)(y)[(3z)Ayz 3 Ayx] ' . ' 2. (x)[(3y)Byx 3 (z)Bxz]
(y)(3z)Ayz (y)(z)(B)/z 3 Bzy)
•‘•(3 x)(y)Ayx .
 € 5.2] Arguments Involving Relations 133
3. (x)(Cax 3 Dxb) 6. (x){Kx 3 f(3t/)/^rrf/ D (32)L2x]J
(3x)Dxh 3 (3y)Dby (x)[(32)Lsx 3 Lxx]
(3x)Cax 3 (ly)Dby ~~(3x)Lxx
/. (x)(Kx D (y) ^ Lxy)
4. (x)[Ex 3 (y)(Fy 3 C;jcf/>] 7. (x)[Mx 3 (r/)0Viy 3 Oxi/)]
(3x)[Ex-(3y) ~~Gxy] (■x)[Px 3 (y)(Oxy 3 Qy)]
.\(3x )~~Fx (3x)(Afx -Z^x) 3 (i/)(iV{/ 3 Qy)
‘5. (3x)[Ux-(y)(lyDjxy)] 8. (x)[(flx—Sr) 3 (3y)(Txy-Uy)]
.\(x)(IIx 3 lx) 3 (3y)(IyJyy) (3x)[Vx-Rx'{y)(Txy 3 Vy)]
(x)(Vx 3 -Sx)
:.(3x)(Vx-VxJ
9. (x)(Wx D Xx)
(x)[(Yx-Xx) 3 Zx]
(x)(3y)(YyAyx)
(x)(y)[(Ayx-Zy) 3 Zx]
(^)[(?/)(At/x 3 Wjy) 3 Zx]
10. (x){[Bxt(3y)[CyDyx’(3z)(Ex'Fxz)]\ 3 (Bu^Cxwx)
(x)(y)(IIxy 3 Dy/x)
(x)(y)(Fxy 3 Ei/x)
(x)(/x 3 Ex)
.‘.(x){Bx 3 \[(3y)(CyHxy)t(3z)(lztFxz.)] 3 (3w)(3u;K.,xuJt/])

II. Construct a formal proof of validity for each of the following arguments:

1. Whoever supports Iekes will vote for Jones. Anderson will vote for no one hut
a friend of Harris. No friend of Kelly has Jones for a friend. Therefore, if Harris
is a friend of Kelly, Anderson will not support Iekes. (Sxy: x supports y. Vxy: x
votes for y. Fxy: x is a friend of y. a: Anderson i. Iekes. /: Jones, h: Harris, k:
Kelly.)
2. All circles are figures. Therefore, all who draw circles draw figures. (Ox: x is a
circle. Fx: x is a figure. Dxy: x draws y.)
3. Any friend of A1 is a friend of Bill. Therefore, anyone who knows a friend of
Al knows a friend of Bill. (Px: x is a person. Fxy: x is a friend of y. Kxy: x knows
y. a: Al. b: Bill.)
*4. Only a fool would lie about one of Bill’s fraternity brothers to him. A classmate
of Bill’s lied about Al to him. Therefore, if none of Bill’s classmates are fools,
then Al is not a fraternity brother of Bill. (Fx: x is a fool. Lxyz: x lies about y
to z. Cxy: x is a classmate of y. Bxy: x is a fraternity brother of y. a: Al. b: Bill.)
5. Whoever belongs to the Country Club is wealthier than any member of the
Elks Lodge. Not everyone who belongs to the Country Club is wealthier than
anyone who does not belong. Therefore, not everyone belongs either to the
Country Club or the Elks Lodge. (Cx: x belongs to the Country Club. Ex: x
belongs to the Elks Lodge. Px: x is a person. Wxy: x is wealthier than y.)
*6. It is a crime to sell an unregistered gun. All the weapons that Red owns were
purchased by him from either Lefty or Moe. So if one of Red’s weapons is an
unregistered gun, then if Red never bought anything from Moe, Lefty is a
criminal. (Rx: x is registered. Gx: x is a gun. Cx: x is a criminal. Wx: x is a
weapon. Oxy: x owns y. Sxyz: x sells y to 2. r: Red. I: Lefty, m: Moe.)
7. Everything on my desk is a masterpiece. Anyone who writes a masterpiece is
a genius. Someone very obscure wrote some of the novels on my desk.
 Therefore, some very obscure person is a genius. (Dx: x is on my desk. Mx: x
is a masterpiece. Px: or is a person. Gx: x is a genius. Ox: x is very obscure. Nx:
x is a novel. Wxy: x wrote tj.) •
8. There is a professor who is liked by every student who likes at least one
professor. Every student likes some professor or other. Therefore, there is a
professor who is liked by all students. (Px: x is a professor. Sx: x is a student.
Lxy: x likes y.)
*9. No one respects a person who does not respect himself. No one will hire a
person he docs not respect. Therefore, a person who respects no one will never
be hired by anybody. (Px: x is a person. Rxy: x respects y. Hxy: x hires y.)
10. Whoever donates to the United Fund gives all of his donations to the United
Fund. Therefore, if Jones is a person who does not give all of his donations to
the United Fund, then he does not donate anything at all to the United Fund.
(Px: or is a person. Dxyz: x donates y to 5. f: The United Fund. /: Jones.)
11. Any book that is approved by all critics is read by every literary person.
Anyone who reads anything will talk about it. A critic will approve any book
written by any person who flatters him. Therefore, if someone flatters every
critic, then any book he writes will be talked about by all literary persons. (Bx:
x is a book. Cx: x is a critic. Lx: x is literary. Px: x is a person. Axy: x approves y.
Rxy: x reads y. Txy: x talks about if. Fxy: x flatters y. Wxy: x writes y.)
12. A work of art that tells a story can be understood by everyone. Some religious
works of art have been created by great artists. Every religious work of art tells
an inspirational story. Therefore, if some people admire only what they cannot
understand, then some great artists’ creations will not be admired by every¬
one. (Ax: x is a great artist. Px: x is a person. Sx: x is a story, lx: x is inspira¬
tional. Rx: x is religious. Wx: x is a work of art. Cxy: x creates y. Axy : x admires
y. Txy: x tells y. Uxy: x can understand y.)

5.3 Some Attributes of Relations

There are many interesting attributes that relations themselves may possess.
We shall consider only a few of the more familiar ones, and our discussion will
be confined to attributes of binary or dyadic relations.
Binary relations may be characterized as symmetrical, asymmetrical, or
nonsymmetrical. Various symmetrical relations are designated by the phrases:
‘is next to’, ‘is married to’, and lias the same weight as. A symmetrical relation
is a relation such that if one thing has that relation to a second, then the second
must have that relation to the first. A propositional function 'Rxy’ designates a
symmetrical relation if and only if

(x)(y)(Rxy D Ryx)

On the other hand, an asymmetrical relation is a relation such that if one thing
has that relation to a second, then the second cannot have that relation to the
first. Various asymmetrical relations afe designated by the phrases ‘is north of’,
‘is older than’, and ‘weighs more than’. A propositional function 'Rxy9 desig-
Sec. 5.3] Some Attributes of deletions 135

nates an asymmetrical relation if and only if

(x)(y)(Rxy D —Ryx)

Not all relations are either symmetrical or asymmetrical, however. If one

individual loves a second, or is a brother of a second, or weighs no more than a
second, it does not follow that the second loves the first, or is a brother to the
first (possibly being a sister instead), or weighs no more than the first. Nor does
it follow that the second does not love the first, or is not a brother to him, or
does weigh more than the first. Such relations as these are nonstjmmet rival, and
are defined as those which are neither symmetrical nor asvfnrnetrical.
Binary relations may also be characterized as transitive, intransitive, or
nontransitive. Various transitive relations are designated by the phrases: ‘is
north of’, ‘is an ancestor of’, and ‘weighs the same as’. A transitive relation is
a relation such that if one thing has it to a second, and the second to a third,
then the first must have it to the third. A propositional function 'Rxtf desig¬
nates a transitive relation if and only if

(x)(y)(z)[(RxyRyz) D Rxz]

An intransitive relation, on the other hand, is a relation such that if one thing
has it to a second, and the second to a third, then the first cannot have it to the
third. Some intransitive relations are designated by the phrases ‘is mother of',
‘is father of’, and ‘weighs exactly two pounds more than'. A propositional
function ‘Rxif designates an intransitive relation if and only if

(x)(y){z)[(Rxtj-Rtjz) D —Rxz]

Not all relations are either transitive or intransitive. We define a nontransitive

relation as one that is neither transitive nor intransitive. Kxamples of non¬
transitive relations are designated by loves’, ‘is diseriminably different from’,
and ‘has a different weight than’.
Finally, relations may be reflexive, irreflexive, or nonreflexive. Various defi¬
nitions of these properties have been proposed by different authors, and there
seems to be no standard terminology established. It is convenient to distinguish
between reflexivity and total reflexivitv. A relation is totally reflexive if every¬
thing has that relation to itself. For example, the phrase ‘is identical with’
designates the totally reflexive relation of identity. A propositional function
'Rxtf designates a totally reflexive relation ii and only if

(x)Rxx

On the other hand, a relation is said to be reflexive if any thing a has that
relation to itself if there is something b such that either Rab or Rba. Obvious
examples of reflexive relations are designated by the phrases ‘has the same color
136 The Logic of Relations [Ch. 5

hair as’, ‘is the same age as\ and ‘is a contemporary of\ A propositional
function ‘Rxy’ designates a reflexive relation if and only if

(x)[(iy)(Rxy v Ryx) d Rxx)

It is obvious that all totally reflexive relations are reflexive.

An irreflexive relation is a relation that nothing has to itself. A propositional
function ‘Rxtf designates an irreflexive relation if and only if

(x)~~Rxx

Examples of irreflexive relations are common indeed; the phrases: ‘is north of’,
‘is married to’, and ‘is parent of’ all designate irreflexive relations. Relations
that are neither reflexive nor irreflexive are said to be nonrejlexive. The phrases
‘loves’, ‘hates’, and ‘criticizes’ designate nonreflexive relations.
Relations may have various combinations of the attributes described. The
relation of weighing more than is asymmetrical, transitive, and irreflexive,
while the relation of having the same weight as is symmetrical, transitive, and
reflexive. However, some attributes entail the presence of others. For example,
all asymmetrical relations must be irreflexive, as can easily be demonstrated.
Let ‘Rjci/’ designate any asymmetrical relation. Then, by definition:

l- {x)(y)(Rxy D —Ryx)

From this premiss we can deduce that R is irreflexive, that is, that (x)^Rxx:

2. (y)(Rxy D ^Ryx) 1, UI
3. Rxx D ~^Rxx 2, UI
4. '■'-Rxxv ~~Rxx 3, Imp).
5. ~~Rxx 4, Taut.
6. (x)~~Rxx 5, UG

Other logical connections among these attributes of relations are easily stated
and proved, but our present interest lies in another direction.
At this point, it should be remarked that the method of proving invalidity
that was presented in Chapter 4 (pages 78-81 and pages 105-107), which is
always applicable to nonrelational arguments, is, not always applicable to
relational arguments. Some sets of relational premisses, although perfectly
consistent, cannot all be true for any finite model. Thus premisses asserting that
the relation R is transitive, asymmetrical, and irreflexive, and further that
(x)(3y)Rxy, cannot all be true for any finite model. Consider any model that is
a nonempty set. Since it is nonempty, it contains at least one object, which we
shall call ‘a’. From (x)(3y)Rxy we know that (3y)Ray. By irreflexiveness of R,
~~Raa, so y cannot be a. Hence the model must contain another object, fc, such
that Rab. Now again, from (jc)(3tj)Rxy we know that (3y)Rby. By irreflexive-
Sec. 5.3] Some Attributes of Relations 137

ness, y cannot be b. Since Rab, by asymmetry ~~Rba, so y cannot be a either.

Hence the model must contain still another object e: such that Rbc. By transi¬
tivity, Rac also. Again from (x)(3y)Rxy, we know that (3tj)Rcy. But by
asymmetry and irreflexiveness, y cannot be a or b or c, so the model must
contain still another object, d, such that Red. It is clear that this pattern of
argument can be repeated indefinitely to show that however large a finite
model we consider, the premisses require a still larger model. Yet the premisses
are perfectly consistent, being true of the natural number 0, 1, 2, 3,... where
7ixy’ is interpreted to mean x is smaller than y. Hence the method of proving
invalidity discussed in Chapter 4 is not always applicable to arguments
involving relations.
The relevance of these various attributes of relations to relational arguments
is easily seen. An argument to which one of them is relevant might be stated
thus:

Tom has the same weight as Dick.

Dick has the same weight as Harry.
The relation of having the same weight as is transitive.
Therefore, Tom has the same weight as Harry.

When it is translated into our symbolism as

Wtd
Wdh
(x)(y)(z)\(\VxyWyz) D Wxz\
' Wth

the method of proving its validity is immediately obvious. We said that the
argument ‘might’ be stated in the way indicated. But such a way of stating the
argument would be the exception rather than the rule. The ordinary way of
propounding such an argument would be to state only the first two premisses
and the conclusion, on the grounds that everyone knows that having the same
weight as is a transitive relation. Relational arguments are often used, and
many of them depend essentially on the transitivity, or symmetry, or one of the
other attributes of the relations involved. But that the relation in question has
the relevant attribute is seldcftn— if ever—stated explicitly as a premiss. The
reason is easy to see. In most discussions, a large body of propositions can be
presumed to be common knowlege. The majority of speakers and writers save
themselves trouble by not repeating well-known and perhaps trivially true
propositions that their hearers or readers can perfectly well be expected to
supply for themselves. An argument that is incompletely expressed, part of it
being ‘understood’, is an enthymeme.
Because it is incomplete, an enthymeme must have its suppressed premiss or
premisses taken into account when the question of its validity arises. Where a
138 The Logic of Relations [Ch. 5

necessary premiss is missing, the inference is technically invalid. But where the
unexpressed premiss is easily supplied and obviously true, in all fairness it
ought to be included as part of the argument in apy appraisal of it. In such a
case, one assumes that the maker of the argument did have more ‘in mind* than
he stated explicitly. In most cases there is no difficulty in supplying the tacit
premiss that the speaker intended but did not express. Thus the first specimen
argument stated at the beginning of this chapter:

A1 is older than Bill.

Bill is older than Charlie.
Therefore, A1 is older than Charlie.

ought to be counted as valid, since it becomes so when the trivially true

proposition that being older than is a transitive relation, is added as an auxiliary
premiss. When the indicated missing premiss is supplied, a formal proof of the
argument’s validity is very easily set down.
Of course, premisses other than relational ones are often left unexpressed.
For example, in the argument

Any horse can outrun any dog. Some greyhounds can outrun any rabbit.
Therefore, any horse can outrun any rabbit.

not only is the needed premiss about the transitivity of being able to outrun left
unexpressed, but also left unexpressed is the nonrelational premiss that all
greyhounds are dogs. When these are added—and they are certainly not
debatable issues—the validity of the argument can be demonstrated as follows:

1. (x)[Hx D (y)(Dy D Oxy)] | premisses

2. (1y)[Oy(z)(Rz D Oyz)] J /.'. (x)[Hx D (z)(Rz D Oxz)|
3. (x) (y)(z)[(OxyOyz) D Oxz]
additional premisses
4. (y) (Oy D Dy)
5. Hx
—> 6. Rz
r-*7. Gy(z)(Rz D Oyz) -
8. Gy 7, Simp.
9. Gy D Dy 4, UI
10. Dy 9, 8, M.P.
11. Hx D (y)(Dy D Oxy) I, UI
12. (y)(Dy D Oxy) II, 5, M.P.
13. Dy D Oxy 12, UI
14. Oxy 13, 10, M.P.
15. (z)(Rz D Oyz) 7, Simp.
16. Rz D Oyz 15, UI
17. Oyz 16, 6, M.P.
18. Oxy Oyz 14, 17, Conj.
Sec. 5.3] Some Attributes of Relations 139
19. (y)(z)[(OxyOyz) D Oxz] 3, UI
20. (z)[(Oxy'Oyz) D Oxz] 19, UI
21. (OxyOyz) D Oxz 20, UI
22. Oxz 21, 18, M.P.
23. Oxz 2, 7-22, El
24. Rz D Oxz 6-23, C.P.
25. (z)(Rz D Oxz) 24, UG
26. IIx D (z)(Rz D Oxz) 5-25, C.P.
27. (x)[Hx D (z)(Rz D 0x2)1 26, UG

Missing premisses are not always so easily noticed and supplied as in the
present example. When it is not so obvious which needed premisses are missing
from an enthymematically expressed argument, it is a good policy in beginning
a proof of the argument’s validity to leave a little space just below the given
premisses. Additional premisses can be written in that space when need arises.
The only point to be stressed is that no statement that is as doubtful or
debatable as the argument’s own conclusion is to be admitted as a supplemen¬
tary premiss, for in a valid argument that is enthymematically stated, only the
sheerest platitudes should be left unexpressed for the hearer or reader to fill in
for himself.

EXERCISES

Prove the validity of the following cnlhymemcs—adding obviously true premisses

where necessary:
1. A Cadillac is more expensive than am low-priced car. Therefore, no Cadillac is
a low-priced car. (Cx: x is a Cadillac. Lx: x is a low-priced car. Mxy: x is more
expensive than y.)
2. Alice is Betty’s mother. Betty is Charlene’s mother. Therefore, if Charlene loves
only her mother, then she does not love Alice, (a: Alice, b: Betty. c: Charlene.
Mxy: x is mother of y. Lxy: x loves «/.)
*3. Any man on the first team can outrun every man on the second team. Therefore,
no man on the second team can outrun any man on the first team. (Lx: x is a man
on the first team. Sx: x is a man on the second team. Oxy: x can outrun y.)
4. Every boy at the party danced with every girl who was there. Therefore, every
girl at the party danced with every boy who was there. (Bx: x is a boy. Gy: x is
a girl. Px: x was at the party. Dxtj: x danced with y.)
*5. Anyone is unfortunate who bears the same name as a person who commits a
crime. Therefore, anyone who commits a burglary is unfortunate. (Px: x is a
person. Ux: x is unfortunate. Cy: x is a crime. Bx: x is a burglary. Cxy: x commits y.
Nxy: x bears the same name as y.)
6. All the watches sold by Kubitz are made in Switzerland. Anything made in a
foreign country has a tariff paid on it. Anything on which a tariff was paid costs its
purchaser extra. Therefore, it costs extra to buy a watch from Kubitz. (Wx: y is a
watch. Ty: x has a tariff paid on it. Fx: x is a foreign country. Cxy : x costs y extra.
Mxy : x is made in y. Bxyz: x buys y from s. s: Switzerland, k: Kubitz.)
140 The Logic of Relations [Ch. 5

*7. Vacant lots provide no income to their owners. Any owner of real estate must pay
taxes on it. Therefore, any owner of a vacant lot must pay taxes on something
that provides no income to him. (Vx: x is a vacant dot. Rx: x is real estate. Ixy: x
provides income to y. Txy: x pays taxes on y. Oxy: x owns y.)
8. All admirals wear uniforms having gold buttons. Therefore, some naval officers
wear clothes that have metal buttons. (Ax: x is an admiral. Ux: x is a uniform. Gx:
x is gold. Bx: x is a button. Nx: x is a naval officer. Cx: x is clothing. Mx: x is metal.
Wxy: x wears y. Hxy: x has y.)
9. Whenever Charlie moved to Boston, it was after he had met Al. Whenever
Charlie got married, it was before he ever saw Dave. Therefore, if Charlie moved
to Boston and subsequently got married, then he met Al before he ever saw Dave.
(Tx: x is a time. Ax: Charlie met Al at (time) x. Bx: Charlie moved to Boston at
(time) x. Mx: Charlie got married at (time) x. Dx: Charlie saw Dave at (time) x.
Pxy: x precedes y.)
10. A fish that chases every shiner will be hooked by an angler who uses a shiner for
bait. A greedy fish will chase every shiner. So if all anglers are sportsmen, then no
pike that is not hooked by a sportsman who uses minnows for bait is greedy. (Fx:
x is a fish. Sx: x is a shiner. Cxy : x chases y. Hxy : x hooks y. Ax: x is an angler. Bxy :
x uses y for bait. Gx: x is greedy. Px: x is a pike. Rx: x is a sportsman. Mx: x is
a minnow.)

5.4 Identity and the Definite Description

The notion of identity is a familiar one. Perhaps the most natural occasion
for its use is in the process of identification, as in a police line-up when a
witness identifies a suspect and asserts that

The man on the right is the man who snatched my purse.

Other identifications are common, as in a geography class when it is asserted

that

Mt. Everest is the tallest mountain in the world.

j*

or when it is asserted in a literature class that

Scott is the author of Waver ly.

Each of these propositions asserts that a relation obtains between the things
referred to by its two terms. The relation asserted to hold is that of identity. In
each of the preceding propositions, at least one term was a definite description,
which is a phrase of the form ‘the so-and-so’. In identifications, however, both
terms may be proper names. Just as the two propositions

Brutus killed Caesar.

Sec. 5.4] Identity and the Definite Description 141
and

Booth killed Lincoln.

assert the relation of killing to hold between the individuals referred to bv the
proper names that appear in the propositions, so the propositions

Lewis Carroll was Charles Lutwidge Dodgson.

and

Mark Twain was Samuel Clemens.

assert the relation of identity to hold between the individuals referred to by the
proper names that appear in them.
The usual notation for the relation of identity is the ordinary equals-sign ‘ = \
It is intuitively obvious that the relation of identity is transitive, symmetrical,
and totally reflexive. In our symbolic notation, we write

(*X!/)(-)U(* = ?/)•(«/ = ~)1 3 (X = -)l

(*)(!/) l(* = y)D(y = a) |
(x)(x = .v)

All of these are immediate consequences of the definition of identity that is

contained in Leibniz's principle of the Identity of Indiscernibles:

v = y if and only if every attribute of x is an attribute of y, and conversely.

This principle permits us to infer, from the premisses v = /i and any formula
containing an occurrence of the symbol v, as conclusion any formula that
results from replacing any occurrences of r in the second premiss by the symbol
ju.1 Any inference of this pattern is valid, and in a proof, it should have the
letters ‘Id/ written beside it. A specimen deduction or two will make this clear.
The argument

O. Henry was William Sidney Porter.

O. Henry was a writer.
Therefore, William Sidney Porter was a writer.

may be symbolized and proved valid by the following, in which we use the
letters 7i’ and *p to abbreviate the proper names O. Henry’ and ‘William

Mlere we are using the letters mu and nu to represent any individual symbols, constants or
variables.
142 The Logic of Relations [Ch. 5

Sidney Porter’, and the symbol ‘Wx’ for *x was a writer’:

1. h = p
2. Wh /.*. Wp
3. Wp 1, 2, Id.

Another illustration is provided by the argument

George Eliot wrote The Mill on the Floss.

George Eliot is Mary Ann Evans.
Mary Ann Evans is a woman.
Therefore, a woman wrote The Mill on the Floss.

Using the symbols ‘g\ ‘/’, ‘m’, ‘Wxy\ ‘Wx’ to abbreviate ‘George Eliot’, ‘The
Mill on the Floss\ ‘Mary Ann Evans’, ‘x wrote y\ and ‘x is a woman’, we may
formulate the given argument and demonstrate its validity as follows:

1. Wgf
2. g = m
3. Wm /.'.(3x)(Wx-Wxf)
4. Wg 2, 3, Id.
5. Wg-Wgf 4,1, Conj.
6. (lx)(Wx-Wxf) 5, EG

An alternative proof for the second argument would be the following:

4. Wmf 1,2, Id.

5. Wm- Wmf 3,4, Conj.
6. (3x)(Wx-Wxf) 5. EG

A third illustration is provided by the argument:

Only a bald inan wears a wig. Kaplan is a man who wears a wig. This man
is not bald. Therefore, thisT man is not Kaplan.

Using the symbols 't\ ‘k’, ‘Mx\ ‘Bx\ ‘Wx’, to abbreviate ‘this man’, ‘Kaplan’, ‘x
is a man’, ‘x is bald’, and ‘.r wears a wig", we may symbolize the present
argument and prove its validity as follows:

1. (x)[(Mx- Wx) D fix]

2. Mk-Wk
3. /.'. ~~(t = k)
4. (Mk-Wk) D Bk 1,UI
5. Bk 4, 2, M.P.
6. ~(t = k) 3,5, Id.
Sec. 5.4]

This last proof serves to show that we use the principle of Identity not only to
infer from <t>/x and v = p, but also to infer ^~(r = /u) from Or and For
completeness and convenience, we include in our statement of the principle of
Identity its symmetry and its total reflexiveness, even though its symmetry is
readily derived from the others. Our formulation is

Id. Oju, <l>ju

v = fi —Or r = f.L p
- T~ » —-;-r, —-, and -
. . Or . . —(r = n) . .n = v .'. p = p

Now that the symbol for identity is available, we can define still another
attribute of binary relations that is of importance in characterizing ‘partial
order’ in Set Theory, as we shall see in Chapter 7. A propositional function
‘Rxi/ designates an antm/mmetrical relation if and only if

(x)(y)[(RxyRyx) D x = i/j

An important use for the identity symbol is in the formulation of certain

common types of exceptive statements. If we wish to symbolize the proposition

A1 is on the team and can outrun anyone else on it.

using V for ‘Al\ Tx’ for \v is on the team', and ‘Oxtf for \v can outrun if, we
cannot simply write down

Toa(x)(Tx D Oax)

for that would entail

Oaa

which is false because being able to outrun is an irreflexive relation. The

preceding formula does not translate the given proposition, but rather the
necessarily false one:

Al is on the team and can outrun anyone on it.

In this second proposition, the important word ‘else’ is missing. The first
proposition does not assert that Al can outrun everyone on the team, but
everyone else on it, that is, everyone on it who is other than or not identical
with Al. The proper translation of the first proposition is

Ta-(x){[Tx-~(x = «)] D Oa.r}

If we adopt the convention of abbreviating —(v = ft) as r yt /i, the preceding

144 The Logic of Relations [Ch. 5

formula may be written

Ta*(x)[(Tx9x ^ a) D Oax]> '

Making similar use of the identity sign, we may symbolize the propositions

Only John loves Mary.

and

John loves only Mary,

as

Ljm • (x)(x ^ j D —Lxrn)

and
Ljm • (x)(x ^ m D ~~Ljx)

A similar technique can be used in symbolizing the notion expressed by the

English phrases ‘at most’ and ‘no more than’. Thus the statement

There is at most one opening.

is interpreted not as asserting that there is an opening, but that there is no more
than one. This statement may be symbolized as

(*)('/)[(£>*■ Oy) Dx = y]

Similarly, the statement

No more than two visitors are permitted.

is interpreted as leaving open the question of whether there are any at all. This
may be symbolized as

(x)(y)(z)l(Vx-Vy-Vz) D (x = zvy = zvx = y)]

The identity sign is also useful in symbolizing the concept of at least. It is not
needed for ‘at least one’, because the existential quantifier, by itself, suffices.
The statement

There is at least one applicant,

is symbolized as

(3x)Ax
Sec. 5.4] Identity and the Definite Description 145
But to symbolize

There are at least two applicants,

we use the identity sign, writing

(3x)(3i/)[Ax-Ai/-.y ^ i/l

Putting together the notations for ‘at least one’ and ‘at most one’, we have a
method for symbolizing definite numerical propositions. Thus the statement

There is one book on my desk.

meaning exactly one, is symbolized, using 'Bx' for \v is a book’ and 'Dx for \v
is on my desk’, as

(3x){Bx• l)x• (i/)[(By• /){/) D if = x]}

And the statement

Every state elects two Senators.

meaning exactly two, is symbolized, using \Sx' for \v is a state’, Wx’ for ‘.v is a
Senator’, and 'Ext/ for ‘x elects y\ as

(x){Sx D (3y)(3z)[NymNz*ExymIixzmy /
(ic)\[\-wExw) D (to = y v w — £)]]}

Finally, the (presumably false) statement

Cerberus has three heads.

is symbolized, using 7/x’ for ‘x is a head of CJerberus’, as

(3x)(3y)(3z){Hx-HyIIz-x ^ y-y ^ z-x ^ z’ _

(ir)[Hw D (w = xv w = y v w = ~)|}

The present notation is adequate for symbolizing arithmetical statements about

individuals, but to symbolize propositions of pure arithmetic requires an
expanded logical apparatus like that suggested in the following section.
Very rarely does an individual have two different proper names v and g,
which would make v = g a significant and informative statement. However,
individuals are often referred to by means of descriptive phrases rather than by
their proper names. Thus when a crime has been committed, and the police
have not yet learned who committed it, the newspapers are not silenced for
146 The Logic of Relations [Ch. 5

want of a name—the individual in question is referred to as ‘the kidnapper of

the Lindbergh baby*, or ‘the driver of the missing car’.
The word ‘the’ has a variety of uses. In one senjse, it has the force of ‘all’ or
‘any’, as in ‘The whale is a mammal’. But in another sense, it serves to indicate
existence and uniqueness, as in the phrases

The author of Waverley,

The man who shot Lincoln,
The largest city in Illinois,

which refer to Scott, Booth, and Chicago, respectively. One fairly standard
notation for this sense of the word ‘the’ involves an upside-down iota. The
three phrases above are (partially) symbolized as

(ix)(x wrote Waverley),

(7x)(x is a man** shot Lincoln),
(ix)(x is a city in Illinois *x is larger than any other city in Illinois).

In general, a formula like *(ix)(x wrote Waverleyf is read as ‘the x that wrote
Waverley, and is treated as a proper name. Thus we can replace the individual
variable p in the propositional function typ by (w)(<$>v) to obtain 'k(ij')(<l)J') as a
substitution instance.
Normally, a definite description functions in argument the same wav that a
proper name does. The principle of Identity permits us to infer the conclusion
‘Scott wrote Mannion from the premisses ‘Scott is the author of Waverley and
‘the author of Waverley wrote Marmion . The proposition ‘Something is larger
than Detroit’ follows validly by EG from the premiss ‘The largest city in
Illinois is larger than Detroit’. Of course, El instantiates only with a variable
and UG generalizes only from a variable. Therefore, there is no difference,
with respect to these two principles, between a proper name and a definite
description. But with respect to the principle of Universal Instantiation, certain
differences and difficulties arise.
Where the letter *F' designates an attribute, the symbol complex ‘(?x)(Fx)’
refers to the individual having the^ittribute F if there is one and only one such
individual. But what if no individual, or more than one individual, has this
attribute? In such a case, since there is no unique individual to be referred to
by the expression \ix)(Fx)\ that expression does not refer. The problem of
interpreting phrases that purport to refer but do not really do so can be
handled in the following fashion, attributable to Russell.2 An explicit definition
of a symbol is given by presenting another symbol that is equivalent to it in
meaning. Thus the symbol ‘bachelor’ is defined explicitly by equating it to the

“See ‘On Denoting/ Mind, n.s., vol. 14 (1905). Russell’s paper, together with alternative
treatments attributed to G. Frege, Keith* Donpellau. and Hilary Putnam is reprinted in Part 3 of
Contemporary Philosophical Jjogic, 1. M. Copi and J. A. Gould eds., St. Martin’s Press, New York,
1978.
Sec. 5.4] Identity and the Definite Description 147

phrase ‘unmarried man’. An alternative method of explaining the meaning of a

symbol is to give a contextual rather than an explicit definition for it. A
contextual definition of a symbol does not explain the symbol's meaning in
isolation, but rather explains the meaning of any statement or context in which
that symbol occurs. We do not give an explanation of the word ‘the' yor the iota
symbol) in isolation, but instead, we present a method for interpreting any
sentence (or formula) in which it appears. A contextual definition is also called
a ‘definition in use’. Russell's analysis of the definite description consists of a
contextual definition, or definition in use, of the word ‘the’—in the sense in
which it signifies existence and uniqueness. Consider the proposition

The author of Waverletj was a genius.

It seems to assert three things: first

There is an individual who wrote Waver Icy,

second

only that individual wrote Waverletj,

and finally

that individual was a genius.

The three parts of its meaning can be symbolized—without using the iota—as
follows:

(3jt)(x wrote Waverletj)

(!/)(;/ wrote Waverletj D tj — x)

and

x was a genius.

Putting these three parts together, we obtain

{lx){(x wrote Waverletj)-(y)(y wrote Waverley Di/-- x)-(x was a genius)}

Here we have a symbolic translation of the given statement that contains

neither the troublesome word ‘the’ nor any synonym for it. In general, any
statement of the form

The so-and-so is such-and-such,

148 The Logic of Relations [Ch. 5

or any formula such as

is regarded as logically equivalent to

(3x){(x is a so-and-so)'(y)(y is a so-and-so D y = x)*(x is a such-and-such)}

or as

(3x){<J>x-(i/)(<l>t/ D (/ =

Incidentally, when a property is expressed in the superlative form as ‘best’,

‘fastest’, ‘heaviest’, or the like, any proposition containing it can be expressed
by using only the comparative forms ‘better’, ‘faster’, ‘heavier’, or the like.
Thus the statement

The largest ocean is to the west of America.

can be symbolized by using ‘Ox’ for *x is an ocean’, ‘ Wx' for ‘x is to the west of
America’, and ‘Lxt/’ for ‘x is larger than {/’, as

(3*){Ox*(y)[(Oy*y / x) D Lxy)• Wx)

A definite description is ordinarily used only when it is believed to refer. One

normally uses the words ‘the so-and-so’ only when he believes that there is one
and only one so-and-so. But beliefs are often mistaken, and one sometimes uses
such a phrase even when it does not refer. When it, does not, any sentence
affirming that the-so-and-so has such-and-such an attribute or stands in this or
that relation, is false. Thus, even though it might be true that everything has
mass, it is false that

The immortal man has mass.

for this sentence affirms the existence of exactly one immortal man, when there
is none. And unless it occurs in a context that makes clear either that some
particular mountain is being referred to or that mountains in general are under
discussion, the statement

The mountain has mass.

is false. This is so because it affirms that there is only one mountain, whereas in
fact there are many. These remarks should serve to make clear that a phrase of
the form ‘the so-and-so’ or a symbol such as ‘(7x)(Fx)’ cannot be instantiated by
the principle of Universal Instantiation alone. In order to deduce the conclu-
Sec. 5.4] Identity and the Definite Description 149

sion *G(rx)(I<x) from \x)Gx\ we require the additional premiss that there is
exactly one thing which is an F. Where that premiss is missing, the inference is
invalid. But where it is present, the argument is easily proved valid as follows:

1. (x)Gx
2. (3*)[Fx-(y)(Fy D if = x)] /.'. G{ix)(Fx)
3. Fx-(y)(Fy D tj = x)
4. Gx 1,UI
5. Fx*(y)(Ftj Di/ = x)-Gx 3,4, Conj.
6. (3x){Fx*(y)(Fy D y = x)-Gx) 5, EG
7. (3 x){Ff(y)(Fy D y = a) - C } 2, 3-6, El
8. G('ix)(Fx) 7—definition

Any proposition of the form ^(ix)^*) is false either if there is no x that is

(I> or if there is more than one of them. In such a proposition, the descriptive
phrase (7x)(<I>x) is said to have a primary occurrence. But a proposition con¬
taining a descriptive phrase may be part of a larger (truth functional) context,
in which the descriptive phrase is said to have a secondary occurrence. A
proposition in which (?jc)(<1>x) has a secondary occurrence may be true even if
there is no x that is <1> or there is more than one of them. One of the simplest
examples, — might mean either that

(3x) {<l>x •((/)(<!> t/ D !/ = x)' —y\'x] (1)

or that

~(3x){<1>a-(t/)(«!>!/ J y = x)-*x) (2)

If either there i5 no .v that is or there is more than one of them, (1) is false and
(2) is true. To make otherwise ambiguous propositions definite and unambigu¬
ous, Russell proposed using formulas of the form (ix)(<l>x) as a kind of ‘scope
indicator’. But the rules governing this use are quite complicated, and we shall
have no use for them here. Indeed, we shall make no further use of formulas of
the form (ix)(<bx) and shall use instead the explicit symbolic translation that
involves quantifiers and the identity symbol.

EXEHCISKS

Prove the validity of the following arguments—using only the identity symbol in
addition to the indicated abbreviations:
1. The architect who designed Tappan Hall designs only oHiee buildings. Therefore,
Tappan Hall is an office building. (Ax: x is an architect, t: Tappan Ilall. Dxy: x
designed y. Ox: x is an office building.)
*2. The Professor of Greek at Siwash is very learned. Therefore, all Professors of
Greek at Siwash are very learned. (Px: x is a Professor of Greek. Sx: x is at Siwash.
Lx: x is very learned.)
150 The Logic of Relations [Ch. 5

3. The smallest state is in New England. All states in New England are primarily
industrial. Therefore, the smallest state is primarily industrial. (Sx: x is a state. Nx:
x is in New England, tx: x is primarily industrial. $xy: x is smaller than y.)
*4. The fastest running person is a Scandinavian. Therefore, anyone who is not a
Scandinavian can be outrun by someone (or other). (Sx: x is a Scandinavian. Px: x
is a person. Fxy: x can run faster than y.)
5. All entrants will win. There will be, at most, one winner. There is at least one
entrant. Therefore, there is exactly one entrant. (Ex: x is an entrant. Wx: x will
win.)
*6. Any fish can swim faster than any smaller one. Therefore, if there is a largest fish,
then there is a fastest fish. (Ex: x is a fish. Lxy: x is larger than tj. Sxy: x can swim
faster than #/.)
7. Adams and Brown were the only men at the banquet who drank. All the men at
the banquet who brought liquor drank. Adams did not bring any liquor. If any
man at the banquet drank, then some man at the banquet who drank must have
brought liquor. All who drank became ill. Therefore, the man at the banquet who
brought liquor became ill. (a: Adams, b: Brown. Mr: x was a man at the banquet.
Dx: x drank. Lx: x brought liquor, lx: x became ill.)
8. Anyone who has climbed Mt. Blanc is braver than anyone who has not. Of all
those on our team, only the youngest member has climbed Mt. Blanc. Everyone on
our team is a veteran. Therefore, the bravest member of our team is a veteran.
(Cx: x has climbed Mt. Blanc. Tx: x is on our team. Vx: x is a veteran. Bxy: x is
braver than y. Oxy: x is older than y.)
9. There is exactly one penny in my right hand. There is exactly one penny in my left
hand. Nothing is in both my hands. Therefore, there are exactly two pennies in mv
hands. (Px: x is a penny. Rx: x is in my right hand. Lx: x is in my left hand.)
10. All accompanists were bagpipers. All bagpipers were in the cabin. At most two
individuals were in the cabin. There were at least two accompanists. Therefore,
there were exactly two bagpipers. (Ax: x was an accompanist. Bx: x was a
bagpiper. C.v: x was in the cabin.)

5.5 Predicate Variables and Attributes of Attributes

In all of the preceding discussion, quantification has been confined to

individual variables. With the exception of the Greek letters phi and psi, all
attribute symbols and relation symbols introduced thus far have been constants.
The letters 4W\ ‘S’, and 7T in our usage of 4W.v\ ‘Sxy’, and ‘Bxyz as abbrevia¬
tions for \r is wise’, \v is a son of y\ and \t is between y and z\ have designated
definite attributes of, or relations between, individuals. However, attribute
variables and relation variables can be introduced also, and they too may be
quantified.
If we set aside the capital letters ‘F\ ‘CT, 7/\ as attribute or relation
variables and refer to them indifferently as 'predicate variables’, then the .same
quantification techniques that are already familiar will permit symbolizing a
greater variety of statements. We shall be able to symbolize statements about
all or some attributes or relations that things may have or stand in. The
Sec. 5.5] Predicate Variables and Attributes of Attributes 151

expression

Fx

consisting of a predicate variable and an individual variable juxtaposed in that

order may be regarded as a propositional function of two variables. By
instantiation with respect to both variables, we obtain such singular proposi¬
tions as ‘Socrates is mortal’ and ‘Plato is wise’, expressed symbolically as \Y/.y'
and ‘Wp\ By instantiation with respect to the predicate variable and general¬
ization with respect to the individual variable, we obtain such familiar singly-
general propositions as

(x)\tx (everything is mortal)

(3.r) Wx (something is wise)

YVe have seen all these before. A new kind of proposition, however, although
still singly general, is obtained by instantiating with respect to the individual
variable and generalizing with respect to the predicate variable. I lere we have

(F)Fs (Socrates has all attributes)

(3F)Fp (Plato has some attribute)

Finally, by generalizing with respect to both variables, the following doubly-

general propositions are obtained:

(1) (x)(F)Fx (5) (F)(x)Fx

(2) (x)(3F)Fx (6) (F)(3x)Fx
(3) (3x)(F)Fx (7) (3F)(x)Fx
(4) (3x)(3 F)Fx {«) (3F)(3x)Fx

Of these, (J) and (5) are clearly equivalent, (1) stating that

Every thing has every attribute,

and (5) stating that

Every attribute belongs to every thing.

Propositions (4) and (8) are also equivalent to each other, with (4) stating that

Some thing has some attribute,

and (8) stating that

Some attribute belongs to some thing.

152 The Logic of Relations [Ch. 5

The remaining propositions, however, are all distinct. They may be expressed
in English as

(2) Every thing has some attribute (or other).

(3) There is a thing that has every attribute.
(6) Every attribute belongs to some thing (or other).
(7) There is an attribute that belongs to every thing.

There are no equivalences here, but proposition (3) logically entails proposition
(6), and proposition (7) logically entails proposition (2). These implications can
be formally established by means of our familiar quantification rules, allowing
the symbols 7*' and V of the rules to denote predicate variables as well as
individual variables—with the same restrictions on their application, of course.
When predicate variables are introduced and their quantification permitted,
a formal, that is, completely symbolic definition, can be given for the identity
symbol. The definition is

(x = y)=df(F)(Fx = Fy)

From this definition we have

(*)(«/)[(* = !/) = (t’)(h'x=.Fi/)]

following as a logical consequence. And from the latter all the attributes of the
identity relation can be deduced.
For the most part, our preceding discussion has concerned only attributes of
things. But things are not the only entities that have attributes. In Section 5.3
we discussed various attributes that can be attributed to, or predicated of,
relations. Attributes, themselves, can have attributes. Thus the attribute of
being honest has the attribute of being desirable; a virtue (which is an attribute)
may, itself, have the attribute of being rare; and the attribute of being incon¬
siderate is common.
Once it has been noted that some attributes can be predicated of other
attributes, the temptation arises to"predicate certain attributes of themselves.
For example, the attribute of being abstract seems, itself, to be abstract,
whereas the attribute of being concrete seems not to be itself concrete. Any
attribute that can be truly predicated of itself will be said to be a predicable
attribute. Thus being predicable is an attribute that belongs to all those and
only those attributes that can be truly predicated of themselves. On the other
hand, any attribute that cannot be truly predicated of itself will be said to be
an impredicable attribute. Thus being impredieable is an attribute that belongs
to all those and only those attributes that cannot be truly predicated of
themselves.
If we now ask whether the attribute of being impredicable can be truly
predicated of itself or not, we are led to the following unhappy conclusion. If
Sec. 5.5] Predicate Variables and Attributes of Attributes 153

the attribute of being impredicable can be truly predicated of itself, then it has
the attribute of being impredicable. It follows then by definition that the
attribute cannot be truly predicated of itself. On the other hand, if the attribute
of being impredicable cannot be truly predicated of itself, then like all other
attributes that cannot be truly predicated of themselves, it has the attribute of
being impredicable. This means that it can be truly predicated of itself. We
have thus been led to a contradiction.
The contradiction can be derived more clearly by symbolizing the attribute
of being impredicable as 7\ and defining it formally as

IF = df-FF

This definition has the following general proposition as an immediate logical

consequence:

(F)(IF= ^FF)

From the latter, by the principle of Universal Instantiation, we can instantiate

with respect to 7* itself to obtain

11= -11

which is an explicit contradiction.*

Several methods have been proposed for avoiding such contradictions. One
of the best known is Russell’s (simple) Theory of Logical Types,1 which can be
given the following rough formulation. According to Russell, entities are
divided into a hierarchy of different logical types, the lowest of which consists
of all individuals. The next level of logical types consists of all attributes of
individuals, the next of all attributes of attributes of individuals, the next of all
attributes of attributes of attributes of individuals, and so on. Relations and
their attributes give us additional hierarchies, but they will be ignored in the
present discussion. The essential point to the theory of types is not merely the

3 See Russell’s Principles of Mathematics, Cambridge University Press, Cambridge, England,

1903, pp. 79-80, 97-98, and 102. See also Carnap's Logical Syntax of Language, l larcourt Brace
Jovanovich, New York, 1937, p. 211.
4 Russell first formulated his theory of logical lypes in Appendix B of his Principles of Mathe¬
matics. A more complex version of the theory, designed to cope with certain other problems as
well, will be presented in Appendix C of the present volume. See also Russell’s ‘Mathematical
Logic as Based on the Theory of Types,’ American Journal of Mathematics, vol. 30 (1908), pp.
222-262, reprinted in Bertrand Russell: Logic and Knowledge. Essays 1901-1950, Robert Charles
Marsh, ed., George Allen & Unwin Ltd., London, 1956, and Chapter 2 of the Introduction,
Principia Mathematica, 1st ed., Whitehead and Russell, Cambridge University Press, Cambridge,
England, 1910-1913. The interested reader should also consult Irving M. Copi, The Theory of
Logical Types, Routledge and Kegan Paul, London, 1971. The best-known alternative resolution of
the contradiction is the kind of axiomatic set theory first published by E. Zermelo in ‘Untersu-
chungen fiber die Grundlagen der Mengenlehre I\ Mathematische Annalen, vol. 65 (1908), pp.
261-281. Zermelo’s theory, as modified by Fraenkel and others, is discussed in Chapter 7.
154 The Logic of Relations [Ch. 5

division of all entities into different logical types, but the restriction that any
attribute that may significantly be predicated of an entity of one logical type
cannot significantly be predicated of any entity of any other logical type. For
some attributes, the theory of types seems perfectly obvious. Thus an individual
thing may be orange in color, but it clearly does not make sense to say of any
attribute that it is orange in color. An attribute may have many instances, but it
does not make sense either to affirm or to deny of an individual thing that it
has many instances.
The primary motivation for accepting the theory of logical types, however,
is not its naturalness or obviousness, but the fact that it permits the avoidance
of such contradictions as that of the alleged attribute ‘impredicable*. According
to this theory, the type of an attribute must be higher than the type of any
entity of which it can significantly be predicated. Consequently, it does not
make sense either to affirm or to deny of any attribute that it belongs to itself.
Such expressions as ‘FF* and ‘^FF* must be dismissed as meaningless. Conse¬
quently, no such attribute as ‘impredicable’ can be defined, and the contradic¬
tion vanishes. The version of the theory of types sketched above—sometimes
called the ‘Simple Theory of Types’—suffices to rule out all contradictions of
this kind. It also has a certain consonance with common sense. However, there
are alternative solutions or ways to avoid the contradictions, so the theory of
logical types cannot be regarded as the solution. It is widely accepted, how¬
ever, and we shall follow it in introducing a new type of symbol to represent
attributes of attributes of individuals.
Some words have more than a single meaning, of course, and may in one
sense designate an attribute of individuals, and in another sense designate an
attribute of attributes of individuals. Thus the word ‘rare’ in one sense desig¬
nates an attribute of attributes of individuals. An attribute of individuals is rare
if it is exemplified by only a few individuals and, in this sense, it cannot be
affirmed or denied of individuals, themselves. On the other hand, there is a
different sense of the word ‘rare’, in which it designates an attribute of an
individual piece of meat that has been cooked for only a short while, and in this
sense it cannot significantly be affirmed or denied of any attribute. To avoid
ambiguity, we symbolize attributes of attributes of individuals by boldface
italic capital letters ‘A\ ‘B\ ‘C\7.., to prevent their being confused with
attributes of individuals. With this additional symbolic apparatus, we can
translate into our notation such propositions as ‘Unpunctuality is a fault*, and
‘Honesty is a virtue*. Here we use ‘Ux, *Hx\ ‘FF*, and ‘VF\ to abbreviate ‘x is
unpunctual’, ‘x is honest', ‘F is a fault’, and ‘F is a virtue*, and symbolize the two
stated propositions as ‘FU\ and ‘VI/’. More complex propositions can also be
symbolized. The propositions

All useful attributes are desirable.

and

Some desirable attributes are not useful.

Sec. 5.5] Predicate Variables and Attributes of Attributes 155

can be symbolized, using the symbols UF' for ‘F is useful’ and DF’ for ‘F is
desirable’, as

(F)(UF D DF)
and
(3 F)(DF-—UF)

Finally, the proposition

Tom has all of his mother's good qualities.

can be symbolized, using the additional symbols 7' for Tom' and 'Mxy' for \v
is mother of y\ either as

(F){(3x)\Mxt-(y)(Myt D y = x)-Fx-GFj D Ft}

or as

(3x){Mxt-(y)(Myt D 1/ = x)-(F)|(FvGF) D Ft]}

KXRKCISES

I. Symbolize the following propositions:

1. Nothing has all attributes.

2. Some attributes belong to nothing.
*3. No two things have all their attributes in common.
4. Any two things have some common attribute.
5. Napoleon had all the attributes of a great general, (n: Napoleon. Cx: x is a
great general.)
*6. David has all of his father’s virtues and none of his faults, (d: David. Fxtj: x
is father of y. FF: F is a fault. Vi': F is a virtue.)
7. Jones and Smith share all their good qualities but have no bad qualities in
common. (/: Jones, s: Smith. GF: F is a good quality. BF: F is a bad quality.)
8. Nothing that possesses every rare attribute has any ordinary attribute. (RF: i
is rare. OF: F is ordinary.)
*9. A man who possesses all virtues is a virtuous man, but there are virtuous men
who do not possess all virtues. ( Mx: x is a man. V.v: x is a virtuous individual.
VF: F is a virtue.)
10. Everyone has some unusual attribute or other. It would be an unusual person
who had no unusual attributes. (Px: x is a person. Ux: x is an unusual
individual. UF: F is an unusual attribute.)

II. Prove the following:

1. l(x)(F)Fx] = [(F)(x)Fx]
*2. l(3x)(3F)Fx] == l(3F)(3x)Fx]
3. [(3x)(F)Fx] D [(F)(lx)Fx\
4. [(3F)(x)Fx] D [(x)(3F)Fx]
5. l(3H)(x)(3y)iixi/] 3 [(x)(3!/)(3fl)«x!/]
196 The Logic ol Relations (Ch. S

*6. Every (binary) relation that is transitive and irreflexive, is asymmetric.

7. All intransitive binary relations are irreflexive.
8. Any binary relation that every individual his to. some individual or other is
totally reflexive if it is both symmetric and transitive.
9. From the premiss \x)(y)[(x = y)~ (F)(Fx = Fy)]’ (which is true by defini¬
tion), deduce that the identity relation is symmetric, without using the
principle Id.
*10. From the same premiss as in 9 and under the same restriction, deduce that
the identity relation is totally reflexive.
11. From the same premiss as in 9 and under the same restriction, deduce that
the identity relation is transitive.
12. If circles are ellipses, then circles have all properties of ellipses.
13. All (binary) relations that are both symmetric and transitive are reflexive.
Deductive Syeteme
Science is built up with facts, us a house is with stones. But a collection of facts is no more
a science than a heap of stones is a house.
Ilmri hlinear/'

6.1 Definition and Deduction

In the preceding chapters, a number of principles of logic have been set

forth. These principles embody some knowledge about logic, but they do not
constitute a science of logic, for science is organized knowledge. No mere list or
catalog of truths is ever said to constitute a system of knowledge or a science.
We have scientific knowledge only when the propositions setting forth what we
know are organized in a systematic way, to display their interrelations. If a
system of logic or a science of logical principles is to be achieved, those
principles must be arranged or organized in a systematic fashion. This task will
be attempted, on a limited scale, in the following chapters. But first, it will be
of interest to consider the general questions of what interrelations are impor¬
tant, and how propositions may be organized into a system or science.
All knowledge that’we possess can be formulated in propositions, and these
propositions consist of terms. In any science, some propositions can be deduced
from or proved on the basis of other propositions. For example, Galileo’s laws
of falling bodies and Kepler’s laws of planetary motion are all derivable from
Newton’s more general laws of gravitation and motion. The discovery of these
deductive interrelationships was an exciting phase in the development of the
science of physics. Thus one important relationship among the propositions of
a science is deducibility. Propositions that embody knowledge about a subject
become a science of that subject when they are arranged or ordered by
displaying some of them as conclusions deduced from others.
In any science, some of the terms involved in its propositions can be defined
on the basis of other terms. For example, in physics again, density is defined as
mass per unit volume. Acceleration is defined as the time rate of change in
velocity, and velocity is in turn defined as the tune rate of change of position.
This definition of some terms by means of others also serves to reveal interre¬
lations among the propositions. It shows their concern with a common subject
matter and integrates the concepts of the science just as deductioas integrate its
laws or statements. Propositions that embody knowledge are helped to become

157
158 Deductive Systems [Ch. 6

a science when some of the words or symbols they contain are defined in terms
of their other symbols.
The recognition that definition and deduction are important to science may
suggest a specious ideal for scientific systems. It may be imagined that in an
ideal science, all propositions should be proved by deducing them from others
and all terms should be defined. But this would be ‘ideal’ only in the pejorative
sense that this goal is impossible to realize. Terms can be defined only by means
of other terms, whose meanings are presupposed and must be understood
beforehand if the definitions are to explain the meanings of the terms being
defined. Deductions can establish their conclusions only on the basis of prem¬
isses that must already have been verified if the conclusions are really to be
established by the proofs. Hence, if all terms or symbols of a system are to be
defined within the system, there must be either infinite sequences of definitions,
or circular definitions, as in a pocket dictionary that defines the word ‘big’ as
meaning large, and the word ‘large* as meaning big. Circular definitions are
obviously worthless as explanations, and infinite sequences of definitions are
worthless also. No term can really be explained until the end of the sequence is
reached and an infinite sequence has no end. Similarly, to prove all proposi¬
tions, there must be either infinite regressions of proofs or circular proofs. And
these are equally objectionable.
It must be admitted that within a system of propositions that constitutes a
science, not all propositions can be proved and not all terms can be defined. It
is not that there is some particular proposition that cannot be proved or some
particular term that cannot be defined, but rather that they cannot all be
proved or defined without a vicious regression or circularity. The ideal of
science, then, cannot be a system in which every proposition is proved and
every term defined. The ideal is rather one in which a minimum number of
propositions suffices for the deduction of all the rest and a minimum number of
terms suffices for the definition of all the others. This ideal of knowledge is
described as a deductive system.

6.2 Euclidean Geometry

Euclidean Geometry is the oldest example of systematized knowledge or

science. Of historical interest and importance in its own right, it has the
advantage (for our purpose) of being an example with which the reader has
already come into contact in high school.
It is generally recognized that geometry, as a science, was originated and
developed by the Greeks. Among the most important contributors to its
development were the mathematicians Pythagoras and Euclid. And yet,
geometrical truths were known to the. Egyptians thousands of years earlier, as
attested by their pyramids that were already ancient in the time of Pythagoras
(6th century b.c.). Records reveal that the Babylonians, even earlier, were
familiar with various principles of geometry. If geometrical knowledge already
Sec. 6.2] Euclidean Geometry 159

existed before their time, in what sense did the Greeks originate the science of
geometry? The answer has already been indicated. Before Pythagoras, man's
geometrical knowledge consisted of a collection or catalog of almost wholly
isolated facts. Geometrical truths constituted a mere list of useful empirical
rules-of-thumb for surveying land or constructing biidges or buildings, and
there was no system to their knowledge of geometrical truths. Bv introducing
order into the subject, the Greeks transformed it from a mere body of isolated
bits of knowledge into a science.
System was introduced into geometry by the deduction of some of its
propositions from others. The propositions of geometry were ordered by listing
first those that could be used as premisses in the demonstrations of propositions
that were put later. This systematization of geometry was begun by Pythagoras
and continued by his successors. It culminated in the Elements of Euclid (c.3(K)
b.c.), in which all geometrical propositions were arranged in order, beginning
with Axioms, Definitions, and Postulates, and continuing with Theorems
deduced from the initial propositions. Geometry was cast by the (-reeks into
the form of a deductive system. Theirs was the first deductive system ever
devised. So great was the achievement that it has served as a model for
scientific thought down to the present time. Even today the most advanced
sciences are those which most nearly approximate the form of a deductive
system. These are the sciences that have achieved a relatively small number of
very general principles from which a relatively large number of other laws and
special cases have been derived. Parts of physics have actually been formulated
as deductive systems and similar attempts have been made, with somewhat less
impressive results, in parts of biology and psychology. Perhaps the boldest
attempt in this direction was that of Spinoza, whose most important work, the
Ethics, was written in ‘geometrical’ form. Starting with axioms and definitions,
Spinoza attempted to'deduce the rest of his metaphysical and ethical doctrines
as theorems that were provable on the basis of those initial assumptions.
Euclid begins his geometry with definitions of some of the terms used in its
development. Thus Definition I reads: ‘A point is that which has no parts’, and
Definition 2 reads: ‘A line is length without breadth’.1 Euclid does not attempt
to define all his terms, of course. The first two definitions define the terms
point’ and ‘line’, respectively. The words used in these definitions, such as
parts’, ‘length’, and ‘breadth’, are not, themselves, defined but are among the
undefined terms of the system for Euclid. As more new terms are introduced,
their definitions make use of previously defined terms as well as the original
undefined ones. Thus Definition 4: ‘A straight line is ... [a line] ... which lies
evenly between its extreme points’, makes use not only of such undefined terms
as ‘evenly’ and ‘between’, but also the previously defined terms ‘point’ and
‘line’.
The use of defined terms is, from the point of view of logic, a matter of
convenience only. Theoretically, every proposition that contains defined terms
1 These and the following are quoted from the Todhunter edition of The Elements of Euclid, No.
891, Everyman’s Library, London and New York, 1933.
160 Deductive Systems [Ch. 6

can be translated into one that contains only undefined ones. This can be done
by replacing each occurrence of a defined term by the sequence of undefined
terms that were used to define it. For example, Postulate 1: ‘Let it be granted
that a straight line may be drawn from any one point to any other point’. This
postulate contains the defined terms ‘straight*, ‘line*, and ‘point*, and can be
expressed, without using those defined terms, as: ‘Let it be granted that a length
without breadth which lies evenly between its extreme parts, which (them¬
selves) have no parts, may be drawn from any one thing which has no parts to
any other thing which has no parts*. This version of the Postulate is extremely
awkward, however. Although they are theoretically eliminable, in actual
practice a considerable saving of space, time, and effort is effected by using
relatively brief, defined terms to replace lengthy sequences or combinations of
undefined ones.
In setting up his deductive system of geometry, Euclid divided his unproved
propositions into two groups, one called ‘Axioms’, the other called ‘Postulates*.
He gave, however, no reason for making this division, and there seems to be no
very clear basis for distinguishing between them. Possibly, he felt that some
were more general than others or psychologically more obvious. The contem¬
porary practice is to draw no such distinction but to regard all the unproved,
initial propositions of a deductive system as having the same standing. Now, it
is customary to refer to them all, indifferently, as ‘axioms’ or as ‘postulates’,
without attaching any difference in meaning to those two terms.
Every deductive system, to avoid falling into circularity or into a vicious
regression, must contain some axioms (or postulates) that are assumed but are
not proved within the system. They need not be precarious assumptions or mere
assumptions. They may be very carefully and convincingly established—but
they are not proved within the system itself. Any argument intended to
establish the truth of the axioms is definitely outside the system, or extra-
systematic.
The older conception of Euclidean geometry held that all of its theorems
followed logically from its axioms and were, therefore, just as true as the
axioms. Euclidean geometry also held that the axioms were self-evident. It is
in this tradition that any statement was regarded as ‘axiomatic’ when its truth is
beyond all doubt, since it was sel£evident and did not require any proof. It
should be clear from what has already been said, however, that we are not
using the word ‘axiom’ in that sense. No claim is made that the axioms of any
system are self-evidently true. Any proposition of a deductive system is an
axiom of that system, if it is assumed, rather than proved, in that system. This
modern point of view has arisen largely as a consequence of the historical
development of geometry and physics.
The self-evident truth of the Euclidean axioms (and postulates) was long
believed. It was not believed quite whole-heartedly, however. Most of the
axioms, such as Axiom 9, ‘The whole is greater than its part’, were not
questioned. Although there was* no doubt about the truth of Axiom 12 (the
famous ‘parallel Postulate*), there was considerable scepticism about its ‘self-
Sec. 6.2] Euclidean Geometry 161

evidence . Axiom 12 reads: If a straight line meet two straight lines, so as to

make the two interior angles on the same side of it, taken together, less than
two right angles, these straight lines, being continually produced, shall at
length meet on that side on which are the angles which are less than two right
angles’.2 Proclus, a fifth century a.d. commentator, wrote of it: This ought
even to be struck out of the Postulates altogether; for it is a theorem involving
many difficulties... .3 That is, although its truth was not questioned, its
self-evidence was denied, which was deemed sufficient reason to demote it
from its exalted position as an axiom to the more humble status of a mere
theorem.
The history of mathematics is filled with attempts to prove tire proposition in
question as a theorem, either by deducing it from the remaining axioms of
Euclid or from those axioms supplemented by some more nearly ‘self-evident’
additional assumption. The latter kind of attempt was pretty uniformly un¬
successful, because every additional or alternative assumption that proved
strong enough to permit the deduction of the parallel postulate turned out to
be no more self-evident than Euclid’s own hypothesis. The first kind of attempt
failed also. It was just not possible to deduce the parallel postulate from the
others. The most fruitful attempt was that of the Italian mathematician
Gerolamo Saccheri (1667-1733), who replaced the parallel postulate by alter¬
native, contrary assumptions, and then sought to derive a contradiction from
them, together with Euclid’s other axioms. Had he succeeded in doing so, he
would have obtained a reductio ad absurdum proof of the parallel postulate.
He derived many theorems that he regarded as absurd because they were so
different from common sense or ordinary geometrical intuition. He believed
that he had succeeded thereby in demonstrating the parallel postulate and in
Vindicating Euclid’. Although ‘absurd’ in the sense that they violated ordinary
geometrical intuitions; his derived theorems were not ‘absurd’ in the logical or
mathematical sense of being self-contradictory. Instead of proving the parallel
postulate, Saccheri (unknowingly) did something more important: He was the
first to set up and develop a system of non-Euclidean geometry.
The parallel postulate is, in fact, independent of the other Euclidean
postulates—although it was not proved to be so until the modem period. It is
independent of the other postulates in the sense that neither it, nor its denial, is
deducible from them. Alternative systems of ‘geometry’, non-Euclidean ge¬
ometries, were subsequently developed, notably, by Lobachevsky and Rie-
mann. These were long regarded as ingenious fictions, mere mathematical
playthings, in contrast with the Euclidean geometry, which was thought to be
‘true’ of the real space about us. But subsequent physical and astronomical
research along lines suggested by Einstein’s theory of relativity have tended to
show that—to the extent that the question is significant—‘real’ or physical

2Listed as Postulate 5 by Sir Thomas L. Heath in The 'Thirteen Books of Euclid’s Elements,
Cambridge University Press, Cambridge, England, 1926. For an interesting discussion of the
history of the parallel postulate, the reader is referred to pp. 202 il. of Volume 1 of that work.
3Ibid., p. 202.
162 Deductive Systems [Ch. 6

space is more probably non-Euclidean than Euclidean. In any event, the truth
or falsehood of its axioms is a purely external property of any deductive system.
The truth of its propositions is an extrasystematic consideration. It is no doubt
important, to the extent that a deductive system is ordered knowledge. When we
concentrate our attention on the system as such, however, its order is its more
important characteristic.
From the purely mathematical or logical point of view, a deductive system
can be regarded as a vast and complex argument. Its premisses are the axioms,
and its conclusion is the conjunction of all the theorems that are deduced. As
with any other argument, the logical question does not concern the truth or
falsehood of the premisses, but deals with the validity of the inference. Granted
the truth of the axioms, does the truth of the theorems necessarily follow? That
is the question with which the logician and the mathematician are concerned.
The answer is, of course, yes—if the demonstrations of the theorems are all
valid arguments. Hence the most important aspect of any deductive system is
the cogency with which its theorems are proved. During the rigorous devel¬
opment of deductive systems in abstraction from the extrasystematic explana¬
tion of their undefined terms, the question of truth or falsehood is obviously
irrelevant.

6.3 Formal Deductive Systems

There are serious errors in the system of geometry set forth by Euclid in his
Elements. Indeed, a mistake occurs in his very first proof. The flaw in his proof,
paradoxically enough, was the result of his knowing too much about his subject.
He did not appeal to his explicitly stated axioms, alone, as premisses, but
depended, also, upon what might be called his geometrical intuition.4 Where a
chain of argument involves familiar notions, there is always the danger of
assuming more than the explicitly stated premisses warrant. This is particularly
serious in the development of a deductive system, for any attempted system¬
atization that appeals to unacknowledged assumptions in the derivations of its
theorems, thereby fails to achieve its aim. In a deductive system, the theorems
must be deduced rigorously from the stated postulates. If they are not, however
true they may be, the result falls short of the goal of systematization.
Since lapses from rigor are most often occasioned by too great familiarity
with the subject matter, mathematicians have found it helpful to minimize or
eliminate such familiarity in the interest of achieving greater rigor. In the case
of geometry, that end is accomplished by abstracting from the meanings of
such geometrical words as ‘point*, ‘line', and plane*, and-developing the

4 Euclid’s proof and a brief discussion of his,mistake can be found on pp. 241-243 of vol. 1 of
The Thirteen Books of Euclid’s Elements, by Heath, op. cit. An example of how the same type of
error can lead to conclusions that are false or even self-contradictory can be found on pp. 77-78,
W. W. Rouse Ball, Mathematical Recreations and Essays, Macmillan Publishing Co., Inc., New
York, 1940.
Sec. 6.3] Formal Deductive Systems 163

theorems as purely formal consequences of the postulates. The familiar

geometrical words, with all their associations and suggestions, are replaced by
arbitrary symbols. Instead of deductive systems explicitly and avowedly
concerned with geometrical entities, mathematicians today develop formal
deductive systems whose primitive or undefined terms include arbitrary,
uninterpreted symbols—usually letters of the Greek or Latin alphabets. Since
the undefined terms of a formal deductive system include arbitrary symbols, its
postulates are not propositions at all, but mere formulas. The theorems also are
formulas. The deductive system for Class Algebra in Section 7.2 is a formal
deductive system in the sense just explained.
Deductive relationships can exist, of course, among mere fdrmulas as well as
among propositions. Thus the formula all F's are H's is logically deductible
from the formulas ‘all F’s are GV and ‘all C’s are H’s. Because the postulates
and theorems of a formal deductive system are formulas rather than proposi¬
tions, the proofs of theorems can proceed unhampered by familiar associations
and unconscious assumptions. Moreover, because the formulas are not propo¬
sitions, the question of their truth is strictly irrelevant and does not arise.
More than rigor is gained by the formal development of deductive systems.
Since some of the symbols of a formal deductive system are arbitrary, un¬
interpreted symbols, it may be possible to give them different, alternative
interpretations. And since the theorems are formal consequences of the axioms,
any interpretation of the arbitrary symbols that makes the axioms true will
necessarily make the theorems true, also. The additional advantage of gener¬
ality is thus gained. An example may help to make this clear. With some
knowledge about astronomy, a deductive system for that subject can be set up.
To avoid the errors into which familiarity with the subject matter may lead in
deducing theorems from the axioms chosen, the system may be developed
formally. Instead of taking, say, ‘stars’ and ‘planets’ among the undefined terms,
one may take ‘A’s and ‘B’s. The axioms and theorems will contain these
symbols, and when the system is developed, all its formulas may be interpreted
by letting the symbol ‘A’ designate stars and the symbol designate planets.
Now, if the axioms are true when so interpreted, the theorems must be true
also, and the formal system with this interpretation will constitute a science or
deductive system of astronomy. It may be possible on the other hand, to find
different interpretations of the symbols ‘A’ and 7*’ that also make the axioms
true (and hence the theorems also). The formulas of the system might be made
into different, but equally true, statements by letting the symbol ‘A’ designate
atomic nuclei and the symbol *B’ designate electrons. If this could be done
(and at one stage in the history of atomic physics it seemed highly plausible),
the original formal system with this second interpretation would constitute a
science or deductive system of atomic physics. Hence developing a deductive
system formally, that is, not interpreting its undefined terms until after its
theorems have all been derived, not only helps achieve rigor in its develop¬
ment, but also achieves greater generality because of the possibility of finding
alternative interpretations for and applications of the deductive system that is
164 Deductive Systems [Ch. 6

formulated. This kind of advantage is often realized in pure mathematics. For

example, different interpretations of its arbitrary primitive symbols will
transform the same formal deductive system into the theory of real numbers, on
the one hand, or into the theory of points on a straight line, on the other. That
fact provides the theoretical foundation for the branch of mathematics called
Analytical Geometry. In Section 7.2, we note that the formal deductive system
for Class Algebra is also a Propositional Calculus.
As the term is being us^d here, a formal deductive system is a deductive
system that consists of axioms and theorems, some of whose undefined or
primitive terms are arbitrary symbols, the interpretation of which is com¬
pletely extfaSysfcematic. In addition to those special undefined terms, and others
defined by means of them, the formulas (axioms and theorems) of the system
contain only such logical terms as ‘if ... then ...’, ‘and’, ‘or’, ‘not’, ‘all’, ‘are’,
and the like, and possibly (unless the system is intended for arithmetic itself)
such arithmetical terms as ‘sum’ and ‘product’, and numerical symbols.

6.4 Attributes of Formal Deductive Systems

Usually, though not always, a formal deductive system is set up with some
particular interpretation ‘in mind’. That is, the investigator has some knowl¬
edge about a certain subject, and wishes to set up a system adequate for its
expression. When the formal system has been constructed, the question natu¬
rally arises whether or not it is adequate for the formulation of all the propo¬
sitions it is intended to express. If it is, it may be saicj^jfo be ‘expressively
complete’ with respect to that subject matter. We are here discussing what can
be said in the system, not what can be proved. With respect to a given subject
matter, a formal deductive system is ‘expressively complete’ when it is possible
to assign meanings to its undefined terms in such a way that every proposition
about that subject matter can be expressed as a formula of the system. Whether
the true propositions can be proved as theorems or not is another question that
will be discussed later in this section.
A system is said to be inconsistent if two formulas—one of which is the
denial or contradictory of the other-»can both be proved as theorems within it.
A system is consistent if it contains no formula such that both the formula and
its negation are provable as theorems within it. As was remarked in Section 3.7,
a contradiction logically entails any proposition whatever. Hence a derivative
definition or criterion for consistency can be formulated as follows*. Any system
is consistent if it contains (that is, can express) a formula that is not provable as
a theorem within the system. This is known as the ‘Post criterion for consist¬
ency’, having been enunciated by the American mathematician and logician,
E. L. Post. Consistency is of fundamental importance. An inconsistent deduc¬
tive system is worthless, for all of its formulas are provable as theorems,
including those which are explicit negations of others. When the undefined
terms are assigned meanings, these contradictory formulas become contradic¬
tory propositions and cannot possibly all be true. Since they cannot possibly be
Sec. 6.4] Attributes of Formal Deductive Systems 165

true, they cannot serve as a systematization of knowledge—for knowledge is

expressed in true propositions only.
If one succeeds in deriving both a formula and its negation as theorems of a
system, that proves the system to be inconsistent. But if one tries and does not
succeed in deriving both a formula and its negation as theorems, that does not
prove that the system is consistent. It may onlv reflect a lack of ingenuity in
making proofs on the part of the investigator. How, then, can the consistency
of a deductive system be established? One method of proving the consistency of
a formal deductive system is to find an interpretation of the system in which all
its axioms and theorems are true propositions. Since its theorems are logical
consequences of its axioms, any interpretation that makes its Axioms true will
make its theorems true, also. Hence it is sufficient for the purpose of proving
that a system is consistent to find an interpretation that makes all of its axioms
true.
The axioms of a deductive system are said to be independent (or to exhibit
independence) if no one of them can be derived as a theorem from the others. A
deductive system that is not consistent is logically objectionable and utterly
worthless. There is no logical objection to a deductive system whose axioms are
not independent. However, it is often felt that making more assumptions than
necessary for the development of a system is extravagant and inelegant and
should be avoided. When a formula need not be assumed as an axiom but can
be proved as a theorem from the other axioms that are assumed, aesthetic and
prudential considerations dictate that it ought to be proved and not assumed
for the sake of ‘economy’. A set of axioms that are not independent is said to
be ‘redundant’. A redundant set of axioms is aesthetically inelegant, but it is not
logically ‘bad’.
If one of the axioms of a system can be derived from the remaining ones, the
set of axioms is thereby shown to be redundant. But if one tries and is not able
to derive any of the axioms from the remaining ones, the axioms are not
thereby shown to be independent. The failure to find a demonstration may be
due only to the investigator’s lack of ingenuity. To prove any particular axiom
independent of the others, it suffices to find an interpretation that makes the
axiom in question false and the remaining ones all true. Such an interpretation
will prove that the axiom in question is not derivable as a theorem from the
others, for if it were, it would be made true by any assignment of meanings that
made the others true. If such an interpretation can be found for each axiom,
this will prove the set of axioms to be independent.
The notion of deductive completeness is very important. The term ‘com¬
pleteness’ is used in various senses. In the least precise sense of the term, we can
say that a deductive system is complete if all the desired formulas can be
proved within it. We may have an extrasystematic criterion for the truth of
propositions about the subject matter for which we constructed the deductive
system. If we have, then we may call that system complete when all of its
formulas that become true propositions on the intended interpretation are
provable formulas or theorems of the system. (In any sense of the term, an
inconsistent system will be complete, but in view of the worthlessness of
166 Deductive Systems [Ch. 6

inconsistent systems, we shall confine our attention here to consistent systems.)

There is another conception of completeness that can be explained as
follows. Any formal deductive system will have a certain collection of special
undefined or primitive terms. Any terms that are definable within the system
are theoretically eliminable, that is, they are replaceable in any formula in
which they occur. They can be so replaced by the sequence of undefined terms
by means of which they were first defined. We shall ignore defined terms for
the present. All formulas that contain no terms other than these special
undefined terms (and logical terms) are expressible within the system. We may
speak of the totality of undefined terms as the base of the system. The formulas
expressible in the system are all formulas constructed on that base. In general,
the totality of formulas constructed on the base of a given system can be
divided into three groups. First is the group of all formulas that are provable as
theorems within the system. Second is the group of all formulas whose nega¬
tions are provable within the system and third, the group of all formulas such
that neither they nor their negations are provable within the system. For
consistent systems, the first and second groups are disjoint or mutually exclu¬
sive, that is, they have no formulas in common. Any system whose third group
is empty, that is, contains no formulas at all, is said to be deductively complete.
An alternative way of phrasing this sense of completeness is to say that every
formula of the system is such that either it or its negation is provable as a
theorem.
Another definition of completeness’, that is entailed by, but not equivalent
to, the preceding one, is that a deductive system is complete when every
formula constructed on its base is either a theorem or else its addition as an
axiom would make the system inconsistent.
An example of an incomplete deductive system would be Euclidean geom¬
etry minus the parallel postulate. For the parallel postulate is, itself, a formula
constructible on the base of the Euclidean system. Yet, neither it, nor its
negation, is deducible from the other postulates. It is clear that, although
completeness is an important attribute, an incomplete deductive system may
be very interesting and valuable. By investigating the incomplete system of
Euclidean geometry without the parallel postulate, we can discover those
properties possessed by space independently of the question of whether it is
Euclidean or non-Euclidean. Perhaps a more cautious and less misleading
formulation of the same point is to say that by investigating the incomplete
system, we can discover the common features of Euclidean and non-Euclidean
geometries. Yet, for many purposes, a complete system is to be preferred.

6.5 Logistic Systems

The most important attribute that a deductive system should possess is that
of rigor. A system has rigor when no- formula is asserted to be a theorem unless
it is logically entailed by the axioms. For the sake of rigor, arbitrary rather than
Sec. 8.5] Logistic Systems 167

familiar symbols are taken as undefined or primitive terms and the system is
developed formally. Listing clearly all the undefined terms and explicitly
stating all the axioms used as premisses for the theorems will help specify
precisely which formulas are to be esteemed as theorems and which are not.
With the increased emphasis on rigor that characterizes the modern period,
critical mathematicians have seen that this procedure is not enough. To achieve
rigor, more is required.
A system is rigorous only when its theorems are proved logically, or derived
logically from its axioms. However clearly its axioms are stated, a formal
system will lack rigor unless the concept of logical proof or logical derivation is
also specified precisely. All deductive systems of the sort that have been
mentioned, even formal deductive systems that contain logical terms in
addition to their own special uninterpreted symbols, depend upon ‘ordinary
logic’ for their development. They assume logic, in the sense that their
theorems are supposed to follow logically from their axioms. But they do not
specify what this ‘logic’ is. Hence all earlier deductive systems, for geometry,
or physics, or psychology, or the like, contain concealed assumptions that are
not explicitly stated. These hidden assumptions are the rules or principles of
logic to which one appeals in constructing proofs or derivations of theorems.
Hence all those deductive systems fall short of complete rigor, since not all of
their presuppositions are acknowledged. Therefore, their developments are not
entirely rigorous, but are more or less loose.
The question naturally arises: llow can this looseness be eli ninated and
greater rigor be achieved? The answer is obvious enough. A deductive system
will be developed more rigorously when it is specified not only what axioms are
assumed as premisses in deriving the theorems, but also what principles of
inference are to be used in the derivations. The axioms must be supplemented
by a list of valid argument forms or principles of valid inference.
The demand for rigor and for system does not stop even here, however. For
the .sake of rigor, in addition to its own special axioms, a deductive system must
specify explicitly what forms of inference are to be accepted as valid. It would
be unsystematic—and probably impossible—simply to list or catalog all re¬
quired rules of logic or valid modes of inference. A deductive system of logic,
itself, must be set up. Such a deductive system will have deduction, itself, as its
subject matter. A system of this type, often referred to as a logistic system, must
differ from the ordinary, less formal varieties in several important respects.
Since the subject matter is deduction itself, the logical terms ‘if ... then ...’,
and’, ‘or’, ‘not’, and so on, cannot occur in it with their ordinary meanings
simply assumed. Uninterpreted symbols must be used in their stead. The logical
principles or rules of inference that it assumes for the sake of deducing logical
theorems from logical axioms must be few in number and explicitly stated.
A second fundamental difference between logistic systems and other formal
deductive systems is that in the latter, the concept of a significant or ‘well
formed’ formula need not be specified. It is absolutely required in a logistic
system, however. In an ordinary (nonformal) deductive system, it will be
168 Deductive Systems [Ch. 6

obvious which sequences of its words are significant propositions of English (or
of whatever the natural language is in which the system is expressed). In a
formal but nonlogistieal deductive system, the sequences of *ts symbols are
easily divided into those which ‘make sense’ and those which do not, for they
will contain such ordinary logical words as ‘if ... then ..‘and’, ‘or’, or ‘not’,
by whose disposition in it the sequence can be recognized as significant or
otherwise. An example will make this clear. In a formal deductive system that
contains ‘A’, ‘JB\ and ‘C’ as uninterpreted primitive symbols, the sequence of
symbols ‘If any A is a B, then* it is a C’ is clearly a complete and ‘significant’
formula that may or may not be provable as a theorem. The sequence of
symbols ‘If any A is a B’ is obviously incomplete, while the sequence ‘And or or
A B not not if’ is clearly nonsense. These are recognized as ‘complete’ or ‘well
forjned’, as ‘incomplete’ or ‘ill formed’ by the presence in them of some
symbols whose meanings are understood. In a logistic system, however, all
symbols are uninterpreted. There are no familiar words within its formulas (or
sequences of symbols) to indicate which are ‘well formed’ and which are not.
Where the symbols ‘A*, ‘B’, and ‘ D’ are uninterpreted, there must be some
method of distinguishing between a well formed formula like ‘A D —B’ and
one like ‘AB% D —’, which is not well formed. By our knowledge of the normal
interpretations of these symbols, we can recognize the difference and classify
them correctly, but for the rigorous development of our system, we must be
able to do this in abstraction from the (intended) meanings of the symbols
involved.
The matter may be expressed in the following terms. As ordinarily con¬
ceived, a nonformal deductive system (interpreted, like Euclidean geometry) is
an arrangement or organization of propositions about some specified subject
matter. Consisting of propositions, it is a language in which the subject matter
may be discussed. Understanding the language, we can divide all sequences of
its words into those which are meaningful statements and those which are
meaningless or nonsensical. This division is effected in terms of meanings and is
thus done nonfonnally. In a logistic system, the situation is different because
prior to the extrasystematic assignment of meanings or interpretation, all
sequences of symbols are without meaning. Yet we want, prior to and inde¬
pendent of its interpretation, a comparable division of all of its formulas into
two groups. When meanings are assigned to the primitive symbols of a logistic
system, some of its formulas will express propositions, while others will not. We
may informally characterize a formula which on the intended interpretation
becomes a significant statement as a ‘well formed formula’ (customarily ab¬
breviated Any formulas which on the intended interpretation do not
become significant statements, are not well formed formulas. In a logistic
system, there must be a purely formal criterion for distinguishing well formed
formulas from all others. To characterize the criterion as ‘purely formal’ is to
say that it is syntactical rather than semantical, pertaining to the formal
characteristics and arrangements of • the symbols in abstraction from their
meanings. Thus a logistic system must contain only uninterpreted symbols and
Sec. 6.5] Logistic Systems 169

must provide a criterion for dividing sequences of these symbols into two
groups. The first group will contain all well formed formulas, the second group
will contain all others. Of the well formed formulas, some will be designated as
Axioms (or Postulates) of the logistic system.
All well formed formulas that are not axioms are divided into two groups,
those which are theorems and those which are not. The former are those which
are derivable from the axioms or postulates, within the system. Although
uninterpreted, the well formed formulas of a logistic system constitute a
‘language in which derivations or proofs can be set down. Some well formed
formulas are assumed as axioms, and other well formed formulas are derived
from them as theorems. It might be proposed to define ‘theorem' as any wff
that is the conclusion of a valid argument whose premisses include only axioms
of the system. This proposed definition of ‘theorem* will be acceptable only if
the concept of a valid argument within the logistic system can be formally
defined. Because all wffs of the system are uninterpreted, the ordinary con¬
ception of validity cannot be used to characterize arguments within the system.
This is the case because the usual concept of validity is semantical. An
argument is regarded as valid if and only if the truth of its premisses entails the
truth of its conclusion. Consequently, a purely formal or syntactical criterion of
validity must be provided for arguments that are expressed within the system.
‘Valid* arguments within the system may have not only postulates or already
established theorems as premisses, with new theorems as conclusions. They
may also have as premisses any wffs, even those which are neither postulates
nor theorems, and they may have as conclusions wffs that are -not theorems. Of
course, it is desired that any argument within the system which is syntactically
‘valid* will become, when given its intended or normal’ interpretation, a
semantically valid argument.
Any logistic system; then, will contain the following elements: (1) a list of
primitive symbols which, together with any symbols defined in terms of them,
are the only symbols that occur within the system; (2) a purely formal or
syntactical criterion for dividing sequences of these symbols into formulas that
are well formed (wffs) and those that are not; (3) a list of wffs assumed as
postulates or axioms; (4) a purely formal or syntactical criterion for dividing
sequences of well formed formulas into ‘valid* and ‘invalid arguments; and (5),
derivatively from (3) and (4), a purely formal criterion for distinguishing
between theorems and nontheorems of the system.
Different logistic systems may be constructed as systematic theories of
different parts of logic. The simplest logistic systems are those which formalize
the logic of truth-functional compound statements. These systems are called
propositional calculi or, less frequently, sentential calculi. One particular
propositional calculus will be presented and discussed in Chapter 8.
Set Theory

7.1 The Algebra of Classes

The four traditional types of subject-predicate propositions discussed in

Section 4.1 can be understood as being about classes. On this interpretation,
these categorical propositions (as they have been traditionally called) are
understood to affirm or to deny that one class is included in another, either in
whole or in part. Thus the A proposition ‘All humans are mortal’ is taken to
assert that the whole of the class of humans is included in the class of mortal
beings, and the I proposition ‘Some humans are mortal’ asserts that part of the
class of humans is included in the class of mortals. Their respective negations,
the O and E propositions, simply deny these inclusions.1
The notion of a class is too basic to be defined in terms of more fundamental
concepts. A variety of synonyms are available: collection, aggregate, totality,
set, and so on. Although mentioning them may help to make the notion of class
more intuitive, they can scarcely provide a useful definition. In the following
sections, where axiomatic treatments are set forth, ‘class’ is an undefined or
primitive term.
On the class interpretation of categorical propositions, both their subject and
predicate terms designate classes. What has been said so far may suggest that
the relation of class inclusion is fundamental. But in developing an algebraic
treatment of classes, statements about them are most conveniently expressed as
equations and inequalities. To periflit this kind of formulation, three aerations
on classes are taken as fundamental, and in terms of these operations the
relation of class inclusion can readily be defined.
In this section and the next, we use lower-case Greek letters to symbolize
classes. Given the classes a, [1, y,... additional classes can be defined in terms
of them. We can define the class of all objects that belong either to a or to /?;
this class is called the sum or union of a and /? and symbolized by *a U /S’. We
can define the class of all objects that belong to both a and /?; this class is called
the product or intersection of a and /? and is symbolized by ‘a fl [V or simply

'See Chapters 5 and 6, Irving M. Copi, Introduction to Logic, Macmillan Publishing Co., Inc.,
New York and London, 1978.

170
Sec. 7.1] The Algebra of Classes 171

*afi . And we define the class of all objects that do not belong to a; this class is
called the complement of a and is symbolized either by —« or by a.
Many statements about classes can be formulated using the ordinary equals
sign. The equation a = p asserts that all members of a, if any, are also
members of fi9 and that all members of (1, if any, are also members of a. Many
of the properties of the sum, product, and complement of classes can be
expressed by means of equations. It is clear, for example, that the operations of
forming the sum and forming the product of two classes are commutative. For
example, symbolically, we have

(a U /?)=(/? U a) and a(l = (la

They are also associative

(« U ft) U y = a U (ft U y) and (aft) y = a (fly)

Two principles of distribution also hold for class sums and products. Any
object that belongs either to a or to both (1 and y must belong either to a or
to (1 and must also belong either to a or to y, and conversely. In symbols,

a U (fly) = (a U (1 )(« U y)

Moreover, any object that belongs both to a and to either (1 or y must belong
either to both a and (1 or to both a and y. In symbols,

a(/l Uy^ = a(l U ay

Two other principles express the idempotence of U and O; they resemble

the tautology principle for statements (see page 40) and are immediate
consequences of the definitions of the sum and the product of classes:

a = a U a and a = aa

Another immediate consequence of those definitions is the principle of ab¬

sorption:

a = a U a(l

Turning now to the concept of class complement, we observe that since any
object belongs to a given class if and only if it does not belong to the class of
all objects that do not belong to the given class, the complement of the
complement of a class is the class itself. We thus have a sort of double negative
rule for complementation, which can be expressed in symbols as

a = 3
172 Set Theory [Ch. 7

An object that does not belong to the sum of two classes belongs to neither of
them and must, therefore, belong to both of their complements. And an object
that does not belong to the product of two classfes must belong to the class
complement of at least one of them. These two propositions, and their con¬
verses, which are also true, can be expressed symbolically as

a U p = ap and ap = a U p

which are versions of De Morgan's theorems applying to classes.

Two special classes are the empty class, which has no members, and the
universal class, to which all objects belong. The empty class is symbolized as ‘O'
(sometimes as ‘A’, and usually in set theory proper as ‘0') and the universal
class is symbolized as T (sometimes as ‘V*). It is clear that the empty class is
the complement of the imiversal class:

1 = 0

The following two equations are immediate consequences of the preceding

definitions:

a U a = 1 and aa = 0

Further immediate consequences are these:

a U 0 = a, «1 = a, aO = 0, and a U 1 = 1

It is easily shown that any class can be designated by infinitely many

different class-expressions. Thus the class designated by V can also be desig¬
nated by 'a(p U ft)’ (since p U ft = 1 and a 1 = a), and by
*[«(/? U /?)](y U y)\ and so on. By this ‘Law of Expansion’ we can always
introduce any class symbol we choose into a given class expression in such a
way that the original and the expanded class expressions designate the same
class.
By the principle of distribution?the class a(P U P) is the same as aP U ap.
To aid in describing the form of the latter expression, let us use the phrase
simple class term’ to refer to the class symbols V, ‘P\ Y,..., in contrast to
other class expressions such as sums and products. Now we can describe the
expression a/f U ap' as a sum of distinct products, such that in each product
only simple class terms or their complements appear and such that any simple
class term which appears anywhere in the entire expression appears exactly
once in every product.2 By means of the equations presented thus far, any class
expression can be transformed into another, perhaps simpler, expression that
designates the same class. Thus a(a U p) is equal by De Morgan's Theorem to
a U (a U P) which is equal, again by De Morgan's Theorem, to a U ap. By

2Compare these with the Boolean Expansions for statements discussed in Appendix B.
Sec. 7.1] The Algebra of Classes 173

double negation this result is equal to a U aft, which is equal by expansion to

a(ft U ft) U ol[1, which by distribution is equal to aft U a/T U aft. (Our asso¬
ciation principle, (a U ft) U y = a U (ft U y), permits us to drop parentheses
and write either simply as a U ft \J y.)
Any class, whatever, will divide the universal class into two snlxli'visions or
subclasses which are mutually exclusive and jointly exhaustive. That is, for any
class a: 1 = a U a and a a = 0. Any two classes will divide the universal class
into four subclasses that are exclusive and exhaustive. Thus for any classes <v
and ft: 1 = aft U aft U aft U aft, and the product of any two of those four
products is the empty class. Similarly, any n classes will divicje the universal
class into 2W subclasses which are exclusive and exhaustive.
The notations introduced thus far permit symbolizing the A and E subject-
predicate propositions. The E proposition, No a is ft, asserts that the classes a
and ft have no members in common, which means that their product is empty.
The E proposition is, therefore, symbolized as

aft = 0

The A proposition, All a is ft, asserts that there is nothing which belongs to a
but not to ft, which means that the product of a and the complement of ft is
empty. The A proposition is, therefore, symbolized as

aft = 0

To symbolize the I and O categorical propositions, we must introduce the

inequality sign where *a rfc ft9 asserts that either a contains an object that
is not a member of ft Qr ft contains an object that is not a member of <v. The /
proposition: Some a is ft, asserts that there is at least one member of <v which is
also a member of ft, that is, that the product of a and ft is not empty. In
symbols, the I proposition appears as

aft ^ 0

The O proposition: Some a is not ft, asserts that there is at least one member of
a that is not a member of ft, that is. the product of a and ft is not empty. In
symbols, the O proposition is expressed as

aft ^ 0

When formulated in our class notation, it is completely obvious that the A and
O propositions are contradictories, as are the E and I propositions.
Some of the traditional ‘immediate inferences’ involving categorical propo¬
sitions are already contained in the notation of the class algebra. Thus every
proposition has exactly the same symbolization as its obverse, for example,
'aft = O’ symbolizes both All a is ft and No a is non-ft. And conversion, where
174 Set Theory [Ch. 7

it is valid, is an immediate consequence of the principle of commutation, for

example, Some a is ft and Some ft is a are symbolized by *aft ^ 0’ and
*fia ^ O’, respectively, which are obviously equivalent, since aft = fta by
commutation.
When we turn to the ‘mediate inferences' of the traditional categorical
syllogism, we can divide all categorical syllogisms into two kinds: those which
contain only universal propositions (A and E), and those which contain at least
one existential proposition (I or O). It is easily shown that all valid syllogisms of
the first kind have the form

aft = 0, fty = 0 ay = 0

The validity of this form can be derived within the algebra of classes by
appealing to the results already set forth in this section. Since yO = 0, and
aft = 0 is a premiss, we have y(aft) = 0, which by association and commuta¬
tion yields (ay)ft = 0. Now a0 = 0, and fty = 0 is a premiss, so we have
a(fty) = which by association and commutation yields (ay)ft = 0. Hence
(ay)ft_U (ay)ft = 0, which by distribution yields (ay)(ft U ft) = 0. Since
ft U ft = 1 and (ay )1 = ay, we have ay = 0, the syllogism’s conclusion.
It can also be shown that all valid syllogisms of the second kind have the
form

aft ^0, fty = 0 ay ^ 0

To establish the validity of this form, we first observe that since a0 = 0, if

aft 0 then a ^ 0 and ft ^ 0. Since a() = 0, and fty — 0 as a premiss,
a(fty) = 0. By association, we have (aft)y = 0. Now aft = (aft)l and y U
y = 1, hence aft = (aft)(y U y), and by distribution, aft = (aft)y U (aft)y.
But (aft)y U 0 = (aft)y, and we have already shown that (aft)y = 0, hence
aft = (aft)y. Since aft ^ 0 is a premiss, we know that (aft)y ^ 0. By associa¬
tion and commutation, we obtain (ay)/? ^ 0 from which it follows that ay ^ 0,
which is the syllogism’s conclusion. Hence the algebra of classes is adequate not
only to validate immediate inferences involving categorical propositions, but is
capable of validating categoricaljsyllogisms also.
The symbol 4 C * is introduced to symbolize class inclusion. The expression
‘a C ft* asserts that all members of a, if any, are also members of ft, and is used
as an alternative formulation of the A proposition, All a is ft. It can be^defined
in terms of the symbols already introduced in various ways: either as aft =0 or
as aft = a or as a U ft = ft or as a U ft = 1, all of which are obviously
equivalent. The relation C is reflexive and transitive (see pages 135-136) and
has the (transposition) property that if a C ft then ft C a. The latter is an
immediate consequence of double negation and commutation when4a C ft9 is
rewritten as 4aft = O’ and ‘/i C a’ is rewritten as 4fta = O’. Its reflexiveness is
obvious when ‘a C a* is rewritten as 4«a = O', and its transitivity has already
been established in our algebraic proof of validity for categorical syllogisms
containing only universal propositions.
Sec. 7.2]
Axioms for Ciacs Algebra 17S

Axioms for Class Algebra

The algebra of classes can be set up as a formal deductive system. Such a

system is called a Boolean Algebra, and a vast number of alternative postulate
sets for Boolean Algebra have been devised. One of them can be set forth as
follows.
Special undefined primitive symbols:

c, pi, u, —, /}, y,
Axioms:

Ax. 1. If a and ft are in C, then a U ft is in C.

Ax. 2. If or and ft are in C, then tx PI ft is in C.
Ax. 3. There is an entity 0 in C such that n U 0 = or for any ix in C.
Ax. 4. There is an entity 1 in C such that a Pi 1 = a for any « in C.
Ax 5. If or and ft are in C, then a U ft = ft U <v.
Ax. 6. If or and ft are in C, then a Pi ft = ft Pi a.
Ax. 7. If or, /^, y are in C, then a U (ft PI y) = (or U ft) Pi (or U y).
Ax. 8 . If or, ft, y are in C, then or PI (ft U y) = (or Pi ft) U (or Pi y).
Ax. 9. If there are unique entities 0 and 1 satisfying Axioms 3 and 4, then
for every or in C there is an —or in C such that

a U —a = 1 and <x Pi — nr = 0

Ax. 10. There is an a in C and a ft in C such that or ^ ft.

The present system3 is a formal deductive system rather than a logistic system.
On its intended interpretation, of course, C is the collection of all classes, 0 and
1 are the empty and universal classes, respectively, and the symbols U, PI, and
— represent class addition, multiplication, and complementation, respectively.
The reader who is interested in deducing some theorems from these axioms
will find the following fairly easy to derive:

Th • 4- There is at most one entity 0 in C such that nr U 0 = nr.

*T ii . 2. There is at most one entity 1 in C such that a Pi 1 = nr.
Th . 3. a U or = nr.
Th . 4. a Pi a = a.
Th . 5. a U 1 = 1.
Th . 6. a PI 0 = 0.
*T H. 7. 0*1.
Th . 8. If a = —ft then ft = —a.

3From E. V. Huntington’s ‘Sets of Independent Postulates for the Algebra of Logic’, Transac¬
tions of the American Mathenuitical Society, vol. 5 (1904), p. 288.
176 Set Theory

Th. 9. a =-a.
Th. 10. If a H p ^ 0, then a ^ 0.
Th. 11. a = (a fl ft) U (a H — /?).
*T H . 12. a U (|8 U y) = (a U /i) U y,
Th. 13. a fl (]8 fl y) = (a fl /}) PI y.
Th. 14. 0 = -1.
Th. 15. a U (a fl p) = a.
Th. 16. a ^ —a.
Th. 17. -(a n /?)=-« u -p.
Th. 18. -(a U (i) = -a PI —/?.
Th. 19. If a fl ~P — 0 and P fl —y = 0, then a fl —y = 0.
*T H . 20. If (x H p ^ 0 and p fl —y = 0, then a fl y ^ 0.

The methods of proof proceed largely by the substitution of equals for

equals. For example, Th. 1 is proved by considering any entities 0X and 02 in
C such that a U 0t = a and a U 02 = a. Since a is any member of C, we have
both Oj U 02 = 0t and 02 U (^ = ()2. Since Oj U 02 = ()2 U 01 by Ax. 5, we
obtain, by substitution, first U 02 = 02 and then = 02, which establishes
the theorem.
Since a Boolean Algebra is a formal deductive system in its own right, it is
susceptible of various interpretations. One of them, of course, is the algebra of
classes. But we can give our Boolean Algebra a propositional rather than a class
interpretation. Suppose we interpret ‘C’ as the collection of all propositions,
and ‘a’, */?’, Y, • • • as symbolizing propositions, and interpret ‘RYU’, and ‘ *
as symbolizing conjunction, (weak) disjunction, and negation. Then if we
further interpret the equals sign as symbolizing material equivalence, all
axioms and theorems of the Boolean Algebra become logically true proposi¬
tions of the propositional calculus. Hence we can say that the propositional
calculus is a Boolean Algebra.4 5

7.3 Zermelo-Fraenkel Set Theory |ZF|—'The First Six Axioms

Abstract set theory was developed in the nineteenth century by Georg

Cantor (1845-1918) in connection with his mathematical researches/1 Unfor¬
tunately, Cantor’s set theory gave rise to a number of contradictions, some of
which will be discussed later in this section and also in Sections 7.4 and 7.7.
The occurrence of these contradictions motivated mathematicians and logi¬
cians to search for an improved version of set theory that would be consistent.
It was suspected that some unrecognized assumption was responsible for the

4 For more details see Robert R. Stoll, Set Theory and Logic, W. II. Freeman and Company,
Publishers, San Francisco and London, 1963, pp. 273-274; or Robert R. Stoll, Sets, Logic, and
Axiomatic Theory, 2nd ed., W. H. Freeman and Company, Publishers, San Francisco and London,
1974, pp. 201-202.
5The mathematical background is described at length by P. G. B. Jourdain in his Introduction
to Georg Cantor, Contributions to the Founding of the Theory of Transfinite Numbers, P. E. B.
Jourdain, tr., Dover Publications, Inc., New York, 1915.
Sec. 7.3] Zermelo-Fraenkel Set Theory (ZF)-The First Six Axioms 177

contradictions. So it was natural to attempt to construct a deductive system for

set theory, in which every assumption would be explicitly stated as an axiom.
Cantor conceived a set to be ‘any collection into a whole ... of definite and
separate objects ... The objects are called the elements [members] of the set’.6
Of the ideas involved in this conception the following seems clear. First, what
is meant by definite is that there is some criterion for membership in the set,
so that any object can be cognized, at least in theory, to be a member or not to
be a member of the set in question. Second, what is meant by ‘separate’ (or
‘distinct) is that any member can be recognized, at least in theory, as different
from any other member, so that no member gets counted more than once: thus
the set containing just the objects 1, 2, and 2-1, is to be counted as having just
two members rather than three. Finally, that a set is ‘a collection into a whole'
indicates that sets themselves are objects and, therefore, are eligible to be
members of other sets.
The fundamental relation in abstract set theory is that of a membership. The
symbol for ‘is a member of’ or ‘belongs to’ is the Greek letter ‘£ ’. To assert that
a is a member of the set A, we write *a £ A’. And to deny that b belongs to A,
we write ‘b ^ A’. If we regard the nation, France, as the set of all its citizens,
and Pierre is a citizen of France, it is true that

Pierre £ France

And since France is a member of the United Nations,

France £ U.N.

But because the U.N. has only nations as members, it is clear that

Pierre ^ U.N.

which shows that £ is not a transitive relation.

It is obvious that different attributes may belong to exactly the same things.
Thus (apart from the plucked chicken of Diogenes) the attribute of being a
featherless biped belongs to all and only those things that have the (different)
attribute of being a rational animal. A set, however, is determined by its
members. Thus sets that contain exactly the same members are the same set.
Thus the set of all featherless bipeds, l\ is equal to or identical with, the set of
all rational animals, R. We write this as

F = R

And to deny that the set A is the same as the set B, we write

A 5* B
8Op. cit., p. 85.
178 Set Theory [Ch. 7

In Chapter 5 Leibniz's principle of the Identity of Indiscemibles was

stated as

x = y if and only if every attribute of jc is an

attribute of y, and conversely

When quantification over predicate variables was introduced, the identity

relation was defined, in accord with Leibniz’s principle:

x = y = <M(F)(Fx=Fy)

From this definition, it was remarked, there follow all of the characteristics of
the identity relation that are embodied in the rules of inference that were
labeled ‘Principle of Identity’. This principle is included in the underlying logic
that is used in set theory in deriving conclusions from premisses, and in
deriving theorems from axioms in axiomatic set theory.
The difference between sets and attributes was remarked above. Although
two attributes may belong to exactly the same objects and yet be distinct, this
is not true of sets. If the set A contains exactly the same members as the set B,
then A = B. This is a fact about sets. It can be regarded as the most basic fact
about sets. And so it is stated as the first axiom of the ZF system: not as a logical
truth about the identity relation but as an assumption about sets.

Z F -1 A = B = df (x)(x £ A = r £ B)

This is the Principle or Axiom of Extensionality and it states that a set is

determined or defined by its members.
A set C, all of whose members are elements of a set D, is called a subset of
D and if C is a subset of D, then C is said to be included in D. The subset
relation ‘C’ is defined in our notation by

C C D = di(x)(x £ C D x £ D)

This definition implies that every set is a subset of itself; in other words, set
inclusion is a reflexive relation. It is obvious that if C C D and D C E, then
C C E. Thus set inclusion is a transitive relation. If C C D and D C C, then by
the Principle of Extensionality, C = D. Where CCD but C ^ D, C is said to
be a proper subset of D.
In the case of sets containing only a small number of members, a set may be
symbolized by putting braces around an enumeration of the names of those
members. Thus the set containing just the two objects a and b can be symbol¬
ized as ‘{a, by or as ‘{fc, a}\ The order in which the names of the members is
written is indifferent, since sets containing the same members are identical. A
set containing just two members is called a ‘pair set’ or ‘doubleton’.
Sec- 7 31 Zermelo-Fraenkel Set Theory (ZFJ-The First Six Axioms 179

For sets containing many members, another notation is available. The new
notation involves, in addition to braces, a variable (such as V), a colon (repre¬
senting the concept of such that), and a propositional function that expresses
some condition that is satisfied or fulfilled by all those objects and only those
objects that belong to the set. Thus the set of all citizens of France would be
symbolized by ‘{x: x is a citizen of France}’. More generally, where Fx is
satisfied by just those objects belonging to the set S, S = {.v: Fx}. So for any
object y, y £ {x: Fx} = Fy, This more powerful way of symbolizing sets can
be seen to subsume the first method, for the set {a, b, c} can also be symbolized
as 4{x: x = a v x = b v x = c}\ This notation is in the spirit of Bertrand
Russell's remark that ‘... a class [set] may be defined as all th6 terms [objects]
satisfying some propositional function.’7
Cantor apparently believed—for a time at least8—that any condition on
objects could determine or define a set, the set containing or comprehending
just those objects satisfying that condition. Although Cantor did not formulate
this belief as a principle, it could be stated as

(3S)(x)(xe S = q>x)

where V represents any predicate or condition and there is no free occurrence

of ‘S’ in q)X. Unfortunately, this unrestricted principle of comprehension leads
to a contradiction, the one generally known as the Bussell paradox. If we take
as <px the propositional function x ^ x, we have the set R = {x: x $ x}, whence
tj £ R = y $ y, and putting R for y yields R £ R e= R $ R, which is a contra¬
diction.9
To avoid that contradiction, some restriction on the comprehension princi¬
ple is needed. The one most commonly used today was first proposed in 1908
by Ernst Zermelo (1871-1953),10 and the axiomatization of set theory proposed
by him then was later (1922) augmented by Abraham A. Fraenkel11 (b. 1891) to
produce the system usually referred to as ‘ZF\12 In ZF, instead of assuming
that any condition on any objects defines the set of all such objects, it is stated
that given anv set A, any condition on members of A defines the subset of A
that contains just those members of A which satisfy that condition. This more
restricted principle of comprehension (called the Aussonderung Axiom by

7 Bertrand Russell, The Principles of Mathematics, Cambridge University Press, Cambridge,

England, 1903; 2nd cd., W. \V. Norton & Company, Inc., New York, 1938.
8See Cantor’s letter to Dedekind in Jean van Heijenoort, ed., Prom Frege to Godel, Harvard
University Press, Cambridge, Mass., 1967, pp. 113-117.
9Cf. pp. 159 ff. supra, and exercise 1 on p. 343.
10Ernst Zermelo, 'Investigations in the Foundations of Set Theory 1\ Stefan Bauer-Mengel-
berg, tr., in Jean van Heijenoort, From Frege to Godel, op. cit., pp. 199-215.
11 Abraham A. Fraenkel, ‘Zu den Grundlagen der Cantor-Zermeloschen Mengenlehre’, Math.
Annalen,~vo\. 86, (1922), pp. 230-237.
12 Because of Thoralf A. Skolem’s contribution to the theory it has been suggested that \ .. it
would seem historically more appropriate to call it Zermelo-Fraenkel-Skolem set theory’. See
Patrick Suppes, Axiomatic Set 'theory, D. Van Nostrarid Company, New York, 1960, p. 8.
180 Set Theory [Ch. 7

Zermelo) can be stated as

ZF-2 (3S)(x)(x £ S = x £ Au<px)

where, again, it is understood that the variable ‘S’ has no free occurrence in the
propositional function <px. Thus, given any set A and any condition <p, there is
a set S which is that subset of A which contains just those members of A that
satisfy the condition <p. (Since the expression ‘qpjt* in the Axiom ranges over all
kinds of propositional functions, the Axiom is more than a single axiom and is
usually referred to as an ‘Axiom Schema*.) This Aussonderung principle has
been referred to, also, as Abstraction,13 as Comprehension,14 as Separation,15 as
Specification,16 and as the Axiom of Subsets.17 How the Axiom of Separation
avoids the Russell paradox can be seen as follows. Given any set A, if in the
Axiom of Separation we take, as cpx, the propositional function x ^ x, we have
the set R = {x: x £ A mx $ x}, whence y £R=y £ Amy y. Now putting R
for y yields R £ R = R £ A *R ^ R, which is not a contradiction but can be
simplified to R £ RmR A.18 From this last formula it immediately follows
that R A. But since A was any set whatever, and R’s existence follows from
the Axiom of Separation, it has been shown that, given any set, there is
something that is not a member of it. Halmos has described this result
somewhat whimsically as the conclusion “that nothing contains evert/thing, or,
more spectacularly, there is no universe. ‘Universe* here is used in the sense of
‘universe of discourse,* meaning in any particular discussion, a set that contains
all the objects that enter into that discussion.”19 Here we have a fundamental
difference between the Algebra of Classes, in which there is posited a universal
class (1 or V), and ZF set theory, in which—as has just been proved—there is
no universal set.
For the Axiom of Separation to produce a set S, there must already be given
a set A of which the new set S is a subset. As Halmos remarked, “All the basic
principles of set theory, except only the axiom of extension, are designed to
make new sets out of old ones.’*20 If we were to make a minimal assumption
that there is at least one set A (as we assumed in Chapter 4 that there is at least
one individual), then we can prq^e by the Axiom of Separation that there is an
empty set, that is, a set that has no members. We already have at hand the
propositional function x ^ x that nothing satisfies, so by Separation, given

13Patrick Suppes, Axiomatic Set Theory, D. Van Nostrand Company, New York, 1960, p. 5.
14 Abraham A. Fraenkcl and Yehoshua Bar-Hillel, Foundations of Set Theory, North-Holland
Publishing Company, Amsterdam, 1958, pp. 137 ff.
15Emst Zermelo, op. cit., p. 202; Patrick Suppes, op. cit., p. 6; and Paul J. Cohen, Set Theory
and the Continuum Hypothesis, W. A. Benjamin, Inc., Menlo Park, Calif., 1966, p. 55.
16Paul R. Halmos, Naive Set Theory, D. Van Nostrand Company, New York, 1960, p. 4.
17 Abraham A. Fraenkel, Abstract Set Theory, North-Holland Publishing Company, Amsterdam,
1966, p. 16.
18 See exercise 5 on p. 343.
19 Halmos, op. cit., pp. 6 ff.
20Halmos, op. cit., p. 4.
Sec. 7.3] Zermelo-Fraenkel Set Theory (ZF)-The First Six Axioms 181
the set A, we have

(3B)(x)(x £ B = x £ A -x ^ x)

Thus B = {x-. x £ Amx ^ x}. Since no y is such that y £ A*y ^ y, it follows

that for every y,y £ B, and by Extensionality there cannot be two empty sets.
It is customary to symbolize the empty set as ‘ 0 ’ (sometimes as ‘A’ or as ‘0’).
Objections have been raised against the very idea of an empty set. Zermelo
referred to it as a ‘fictitious’ set in his original formulation. If a set is deter¬
mined or defined by its members, no set would seem to be defined if there are
no members to define it. As Russell observed long ago: \ .. there is a grave
logical difficulty in trying to interpret extensionally a class [set] which has no
extension.*21
He went on to say both that ‘Great difficulties are associated with the
null-class [empty set].. ,’22 and that \ .. with the strictly extensional view of
classes... a class which has no terms [members] fails to be anything at all: what
is merely and solely a collection of terms cannot subsist when all the terms are
removed.*23 Yet there are reasons for countenancing an empty set. For one
thing, we may not know when what would otherwise be a perfectly acceptable
set might be empty. That is so even in mathematics, for it is not yet known
whether there are any members in the set of all integers n greater than 2, such
that there exist integers .r, t/, ^ all greater than 0 where xn + yn = zn. For
another, it is convenient to have a set that is the product or intersection of two
sets—even where it might not be known whether the two sets have any
members in common. And then, il there are any sets at all, the Axiom of
Separation entails the existence of an empty set. Near the very end of his
Principles of Mathematics, Russell finally concluded: ‘It is quite essential ...
that there should be a null-class .. ' 2A
Instead of making the minimal assumption that there is at least one set, from
which Separation gives us the empty set, the ZF system makes a much stronger
existence assumption as its Axiom VII. That stronger assumption, however,
appears to presuppose the existence of the empty set, as will be seen. Hence it
is really simpler to postulate the existence of the empty set directly. I hat is the
procedure followed by Zermelo in his original formulation, as (part of) Axiom
II, directly following Extensionality, which was his first axiom. We assume
then, as another axiom (the third, in the sequence presented here) the Empty
Set Axiom:

ZF-3 There is a set 0 such that for any object x, x ^ 0. It is an immediate

consequence that for any set S, 0 C S.

21 Bertrand Russell, Principles of Mathematics, Cambridge University Press, Cambridge,

England, 1903, p. 32.
22 Ibid., p. 73.
23Ibid., p. 74.
24 Ibid., p. 514.
182 Set Theory [Ch. 7

We can now see another fundamental difference between the Algebra of

Classes and set theory. In the former, every class'has an absolute complement.
But in set theory this cannot be true, for the absolute complement of 0 would
contain everything and we have already shown that there can be no such set.
At this point, it might be well to avow the general practice of those who
develop one or another variant of ZF set theory. The objects we talk about
here are limited to sets exclusively. Mathematicians work with sets all the time:
sets of points, sets of lines, sets of numbers of various kinds. In fact, all of the
familiar mathematical objects such as numbers and functions can be construed
as sets, so for mathematical purposes no other kinds of objects need be
considered.25 Apart from truth-functional operators, quantifiers, the special
relational symbols ‘£\ ‘=\ V * the colon, and our general notation ‘cpx’
for conditions on x, all of our symbols here, V, *b\ V, ... as well as ‘A\ 7J\ *C\
... will range over sets exclusively. There are no individuals in ZF set theory in
its usual formulations. To allow for individuals would complicate matters more
than is necessary for the mathematical purposes for which set theory is
typically developed.26
Another useful principle is the Axiom of Pairing:

Z F - 4 Given any sets a and b, there is a set S having as members just a and b.
In symbols: (3S)(x)(x ^ S ~ x — a \ x — b).

So for any sets x and y, there is the pair set (x, i/}, which is sometimes called
the unordered pair—in contrast with another concept to be introduced in the
next section. No distinctness assumption was made in the Pairing Axiom. Thus,
given any set a, we can form the unordered pair {a, a) which can also be
written as ‘{a}/ and is called a unit set or singleton.
Russell puzzled over the question whether {a} was the same as a, itself,
initially being inclined to say that it appeared to be ‘self-evident’ that ‘A whole
composed of one term only is that one term.’27 But he subsequently decided
that ‘it is necessary to distinguish a single term from the class [set] whose only
member it is’.28 He was persuaded by an argument he attributed to Frege: Let
u be a set having more than ope member; let {u} be the set whose only
member is u; then {u} has one member, u has many members, hence u ^ {w}.
Had Russell not also had misgivings about the empty set, he might have been
equally well persuaded by the following argument suggested by Stoll. Because
0 has no members and {0 } has a member, it follows that 0 ^ {0 }, which
proves that u ^ {m}.
Apart from the nonexistence of a universal set and the absence of absolute
complements, we want our sets to conform to the rules governing classes set

25 Even points can be construed as sets, as explained in Alfred North Whitehead, Process and
Reality, Cambridge University Press. .Cambridge, England, 1929, Part 4.
26 A set theory containing individuals .can be found in Suppes, op. cit.
27B. Russell, Principles of Mathematics, op. cit., p. 77. See also pp. 68, 73.
28Ibid., p. 106. See also pp. 514 and 517.
Sec. 7.3]
Zermelo-Fraenkel Set Theory (ZF}-The First six Axioms 1SS
forth in the preceding sections. It is easy to see that any two sets a and b have
an intersection or product a n b. This is assured by the Separation Axiom. If
we let x £ b be the yx in that axiom, we have

(3S)(x)(x e s = X e a-x e b).

Since S is unique (by Extensionality), we can introduce the connective ‘fl’ by

the following definition: '

a D b = df [ x: x £ a-x £ b)

This process can be used repeatedly to obtain (a Pi b) P c if c is also a set,

[(a P b) P c] P d if d is also a set, and so on. It is easilv shown that
a P 0 = 0, and that P is commutative, associative, and idempotent. There
is a more general intersection principle than the preceding would indicate. If
A is a nonempty collection of sets a„ a2, a3, ..., there is a set B that is the
intersection of all the members of A. This is given by the Separation Axiom, if
we let ai be any particular set in A (there is one because A is nonempty) and let
<px be the condition that x belongs to every set in A:

(3J3)(x)[x £ B = x £ a{-(a)(a £ A D x £ «)]

Here B is in no way dependent on the choice of one a- rather than another;

indeed B = {x: (a)(a £ A D x £ a)}. Since B is unique (by Extensionality), we
can define the general intersection symbol ‘fV bv the following:

.HA = df (x: (a){a £ A D x £ ri)}

Other notations for HA are ‘Qa-’, ‘fl {a £ A}\ PI {x: x £ A}\ and It
is obvious that PI {a, h) = aPfc.
Our definition of HA was conditional on the set A’s being nonempty. If we
did not impose that condition, there would be no set al whose membership in
A is needed for using the Separation Axiom. So ‘PI 0' is not yet defined. We
cannot define it as {x: (a)(a £ 0 x £ a)} because every x satisfies the
condition that a £ 0 D x £ a, since a $ 0 for every a. Thus the suggested
definition would make fl 0 contain everything, whereas we have already
shown that there is no universal set in ZF. We could leave it undefined (as
division by zero is left undefined in arithmetic), or we could define H 0 to be
0 itself, in order to make fl S defined for any set S whatever.
We also want any two sets a and b to have a union or sum a U b which
contains all (and only) those things that are either members of a or members of
b. The Separation Axiom will not help us here. We could introduce a new
axiom

(3B)(x)(x 6 B = x £ (i v x £ h)
184 Set Theory [Ch. 7

where,by Extensionality, B is the unique set {xt x £ avx £ b] for which we

introduce the notation ‘a U b\ Then we could use it repeatedly to obtain
(a U b) U c if c is also a set, and [(a U b) U c] U chif d is also a set, and so on.
Were we to do so, it would be clear that a U 0 = a, that U is commutative,
associative, and idempotent, and also that fi and U are each distributive
with respect to the other. But such a ‘one at a time* method of forming unions
of given sets is not really enough, because we want a set that contains
every object that belongs to any set a{ of a collection A of sets,
{av a2, a3,..., ai9...}, regardless of whether the collection A is finite or
infinite. (A collection is infinite if there is no natural number 0, 1,2,... that is
the number of members of that collection. The concepts of finite and infinite
will be discussed at length in later sections of this chapter.)
Given a collection A of sets av a2t a3, ... we want an axiom that will
produce a set B containing anything that belongs to any member of A. This is
stated as a more general Union Axiom:

ZF-5 (A)(3B)(x)[x £B = (3a)(x £ a •a £ A)]

As usual the Extensionality Axiom guarantees the uniqueness of the set B, so we

can use any of a number of available alternative notations for it. Where
A = {alt a2, tf3,...}, the set B is symbolized as ‘UA’ or as ‘Urt/ or as
‘{jc: (3tf)(x £ ama £ A)}’ or as ‘U {«: a £ A\ or as ‘U{x: x £?A}' or as

From this general Union Axiom and the Axiom of Pairing, we can obtain
the more modest a U b from sets a and b as follows. By the Axiom of Pairing,
we have (a, b). The set U {a, b) given by the Union Axiom is
{.t: (3z)(x £ z*z £ {«, b})}. But the only z s belonging to (a, b} are a and b,
themselves, so U {«, b) = {x: x £ «vxf b}, which is the desired a U b. It is
obvious that {.t: x £ 0 } = 0, so U 0 = 0, and that U {x: x £ {</}} = a, so
U {a} = a. It is also clear that U {x: x £ {//, b}} = a U fc, so U {«, b} =
a U b, and that {a} U {b} = (a, b).
We have already remarked that there is no absolute complement of any set
in ZF. But the concept of relattve complement is available. The relative
complement A — B of B in A, whose existence and uniqueness is given by
Separation and Extensionality:

A - B = df {x: x£A-x$ B)

which is the set of all members of a that are not members of R It is obvious
that A — 0 = A, that A — A = 0 — A = 0, and that if A H B = 0 then
A — B = A. The relative complement is more useful in mathematics than an
absolute complement would be. For example, one would want the complement
of the set of even numbers to be the set of odd numbers only rather than a set
containing, in addition to odd numbers, points and circles and functions and
Sec. 7.4] Relations and Functions 185

everything else. The relative complement of B in A is sometimes written as

A\B . If we make the assumption in a particular discussion that all of the sets
b, c, ... being considered are subsets of some specified set V, then we can
symbolize the various relative complements of a in V, of b in V, of c in V, and
so on, as a , b , c , and so on. Then in the context of that discussion, the sets
0, b, c,..., a, h, c,... will obey all the laws of the Algebra of Classes presented
in Sections 7.1 and 7.2.
The notion of a subset of a set has already been defined. There are occasions
when one wants to consider the totality of subsets of a set. That the totality of
subsets of a given set E form a set is asserted by the Power Set Axiom:

Z F-6 (E)(3S)(x)(x £SeiC E)

Extensionality insures uniqueness, and we define the power set of a set E,

symbolized as ‘(PE’, by

(PE = df {.v: ,v C /•}

Where E is a finite set, it is easy to see that (Pi. contains more members
than E does. For example: (P0 = {0 }, (P{r/} = {0,{«}}, (?{«, h) =
{ 0, {a}, {b}, {a, /?}}. It is easy to prove hy mathematical induction29 that if
E contains n members, then (PE contains 2" members, which explains the origin
of the term ‘power set’. For any set E, we know that E C E and 0 C E, so for
any set E, we have both E £ (PE and 0 £ (PE. From this last, it is obvious that
D { x: x £ (PE ) = 0. Other immediate consequences of our definitions and
axioms are that E C E if and only if (PE C (PE that (PE U (PE C <P(E U E),
that(P(£ n F) = (PE n (PE, and that, although E = U (PE, in general we have
only that E C (P UE.

7.4 Relations and Functions

Before exploring further implications of the Power Set Axiom, it will be

useful to discuss relations and functions. Relations were discussed in Chapter 5,
where many of the attributes of binary relations were defined and their
interrelations examined. Because not every binary relation is symmetrical, it is
necessary to distinguish between Rah and Rba. Because Rub might be true
while Rba was false, the order of the arguments a and b is essential. We could
express this by saying that the relation R holds for (or the propositional
function Rxu is satisfied by) the ordered pair <a, b) but not the ordered pair
<b,a).
Early in the preceding section we remarked that sets can be regarded as the
extensions of predicates: subject to the restrictions of the Separation Axiom, the

29The method of proof by mathematical induction is reviewed on pp. 223 ff.

186 Sot Theory [Ch. 7

set F consists of all objects that satisfy the predicate or condition <px. By
analogy, we focus our attention on the collection of ordered pairs of objects,
written <x, t/>, between the first and second of whidh the relation Rxy holds. To
develop this notion within set theory, the idea of ‘order' must somehow be
defined in terms of sets. Although its members determine a unique set inde¬
pendently of any order in which they are arranged or in which they may be
named, it is, nevertheless, possible to define the notion of order in purely set
theoretic terms.30 The essence of the notion of order is that the ordered pair
(x, y) is the same as the ordered pair (a, b) if, and only if, both x = a and
y = b. This essence is captured by the Wiener-Kuratowski definition of
ordered pair:

<*,y> =df{{*},{*,y}}

Given any x and y, we use the Axiom of Pairing to produce the set in question
and it is unique by Extensionality. It is immediately obvious that if x = a and
y = b, then (x, y) = (a,b). To establish the converse, we assume that
(x,y) = <a, b>, that is, {{x}, {x,y}} = {{a}, {a, b}}. Since {x} £ {{*},
{x, y}}, we have {x} £ {{«}, [a, b}}. Hence {x} = {a} or {x} = {a, b). In
eithercasex = a.Since{a, b) £ {{«}, [a, b} },wehave {a, b} £ {{x}, {x,y}}.
Hence either (1) {a, b) = {x} or (2) {a, b} = (x, y). By a similar argument,
we also have either (3) {x, y} = {a} or (2) {x, y) = (a, b). If (1) and (3) are
both true, then x = a = y = b, which includes the desired result. If either (1)
or (3) is not true, then {x, y) = {a, b}, whence by our previous result that
x = a, we have that {«, y] = {«, b}. Here, if a ^ b then y = b, but if a = fo,
then y = b. So in every case y = b.
Given two sets A and B, we can form ordered pairs (x, y) with x £ A and
y £ B. The collection of all such pairs is called the Cartesian Product A X Bo f
A and B, and for each ordered pair (a, b) in it, a is called the first coordinate
and b is called the second coordinate. (An example is the following: given the
Euclidean plane with x and y axes established, if A is the set of all real numbers
identified with points on the x-axis and B is the set of all real numbers identified
with points on the y-axis, then thejjet of all ordered pairs (x, y> represents the
Cartesian plane itself—which accounts for the name). That this collection of
ordered pairs is a set, follows from the Power Set, the Union, and the Separa¬
tion Axioms. Given sets A and B, we form their union A U B, which contains
every x £ A and every y £ B. Next, we form the power set (P(A U B) which
contains every set {x} and every set {x, y) as members—along with other
members. Then, we form that set’s power set (P(P(A U B) which contains every
set {{x}, {x, y}} as a member—along with other members. Now, the Separa¬
tion Axiom tells us that there is a subset that is exactly the set wanted.

30 The method for doing so was first devised by N. Wiener and subsequently improved by
C. Kuratowski. See ‘A Simplification of‘the*Logic of Relations’ by N. Wiener, Proc. of the
Cambridge Philosophical Soc., vol. 17 (1914), pp. 387-390, and, ‘Sur la notion de l’ordre dans la
theorie des ensembles’, C. Kuratowski, Fundamenta MathematicOy vol. 2, (1921), pp. 101-171.
Sec. 7.4] Relations and Functions 187

(3S)(z)[z € S= [(* € (P(P(A U B))-(3x)(3i/)(x £ A • 1/ £ B^z = <x, i/»]]

So for any sets A and B, there is a set that contains all and only those ordered
pairs <x, t/> with x £ A and ?/ £ B. Extensionality insures uniqueness, so
‘A X B can symbolize the set in question. Much the same sort of argument,
but using unions instead of powers, will show that if <x, i/> £ A, then both x
and y belong to U U A. One point should be made clear: The Cartesian product
of A and B does not require A 76 B. The product A X A is a perfectly good
Cartesian product, that of A with itself.
We can now define a (binary) relation quite simply as any set of ordered
pairs. R is a relation if and only if

(x)(x £ R D (3m)(3i?)(x = <n, i?»)

Obvious examples of relations are 0, and the Cartesian product of any sets. A
and B, and any subset of a Cartesian product.
The domain of a relation R, (dom R), is the set of all first coordinates of R.
The range of a relation R, (ran R), is the set of all second coordinates of R, and
the field of a relation R (fid R) is the set of all coordinates of R. Civen any
relation R, the axioms permit us to specify both its domain and its range. We
first use the Union Axiom twice to produce the set UUR, and then by
Separation

(3S)(x)[x £ See [x £ U UR-(3?/X<x, ?/> € R)|]

gives us the unique dom R, and

(3T)(x)[x <E T= [x G U UR,(3:)«:, x> € R)]]

gives us the unique ran R. Examples are plentiful: dom 0 = ran 0=0,
dom (A X B) = A and ran (A X B) = B, and if R is equality of members of E,
then dom R = ran R = E.
Except for the empty relation 0, one must always specify a set within, for,
or over which, the relation is defined. If we tried to define the identity relation
X = Y without limiting its domain (or range, or field) we should get into
trouble—for the aggregate Z of all ordered pairs <x, y) such that x = y would
contain every set z = {{x}, (x, x}}, and the union of its union, U UZ would
be the universal set, which has been proved not to exist. The same problem
would arise in connection with set membership (x £ y) and set inclusion
(x C y).
The notion of a function is a familiar one. The pressure of a fixed volume of
gas is a. function of its temperature, and the cost of sending a letter by mail
from New York to San Francisco is a function of its weight. We define a
function in set theoretic terms as a many-one relation, that is, a set of ordered
pairs no two of which have the same first coordinate. If X and Y are sets, a
188 Set Theory [Ch. 7

function from (or on X) to (or into) Y is a relation, /, such that dom f — X and
such that for each x in X, there is a unique y in Y \yith <x, y> £ /. That / is a
function from X to Y is stated as ‘/:X —> Y\ and.is'defined as

/C X x Y-(x)[x € X D (3y)[t/ € Y-<x, y> € /•(*)«*, z)£fDy = z)]\

For each x in X, the unique y in Y such that <x, y> £ /is represented by ‘/(x)’;
so /(x) = y. The object y is the uaiue that the function f assumes or takes on
at or for the argument x. Another locution commonly used is that f maps x onto
y. The domain of a function / from X into Y is X, but its range can be a proper
subset of Y. If the range of f is equal to Y, then f is said to map X onto Y.
A function /from X to Y that maps distinct elements onto distinct elements
is called one-to-one. If /: X —> Y is one-to-one and onto Y, then / effects a
pairing of the elements of X with those of Y, and, as such, is called a one-to-one
correspondence between X and Y. If there exists a one-to-one correspondence
between X and Y, these sets are said to be equivalent (or equinumerons); this is
symbolized as ‘X ~ Y\
The familiar 0, 1, 2, 3,... are called natural numbers or counting numbers.
They will be defined in set theoretic terms in the following section. A finite set
is defined as a set onto which a one-to-one function maps the set of all natural
numbers that are less than some natural number n. Such a set is said to contain
n elements. Thus the one-to-one function that maps 0 onto the letter a, 1 onto
b, 2 onto c,..., 24 onto y, and 25 onto z, maps the set of all natural numbers
less than 26 onto the set of all letters of the alphabet. An infinite set is then
defined (negatively) as any set that is not finite.
No finite set is equivalent to any of its proper subsets. But the matter is
otherwise with infinite sets. Galileo remarked long ago that, although not all
numbers are perfect squares, yet ‘there are as many squares as there are
numbers, because they are just as numerous as their roots and all the numbers
are roots/31 Some equivalences are even more surprising. There are as many
whole numbers as there are rational fractions. The one-to-one correspondence
between these two sets is shown by the following enumeration of all fractions:
of two fractions, we enumerate first that fraction, the sum of whose numerator
and denominator is smaller. Of two fractions whose numerator and denomina¬
tor have the same sum, we count first the one with the smaller numerator. (Of
course, we do not repeat equal fractions in our enumeration: If 1/2 has already
been counted, we do not also include 2/4 or 3/6 or • ■ • .) With this ordering of
the fractions, the one-to-one correspondence appears as follows:

1/1 1/2 2/1 1/3 3/1 1/4 2/3 3/2 4/1 1/5 5/1 1/6 ...
0 1 2 3 4 5 6 7 8 9 10 11 ---

31 Galileo Galilei, Dialogues Concerning Two New Sciences, Northwestern University Press,
Evanston and Chicago, 1946, p. 31.
Sec. 7.4] Relations and Functions 189
An enumeration is of course a one-to-one function from the natural numbers
(0, 1, 2, 3 ...) onto the set being enumerated. Any set, all of whose members
can be mapped one-to-one onto the natural numbers, is said to be denumerable
(or countable, or enumerable). (Of course, we have not yet shown that the
collection of all natural numbers is a set. This will be discussed in the next
section.)
If there is a one-to-one correspondence between X and a proper subset of Y
but no one-to-one correspondence between X and Y, then Y is said to be larger
than X, or to have more elements than X, and X is said to be smaller than Y or
to have fewer elements than Y. Cantor proved that there are more real numbers
than there are integers by showing that in any enumeration of real numbers
there must be some real number that gets left out. (Of course, the integers can
be mapped one-to-one onto a subset of the real numbers by the function f
where /(n) = n.) With no loss in generality, we confine our attention to real
numbers in the interval from 0 to 1, and consider any enumeration of real
numbers from that interval:

■ «11 «12 .. Uy
• f,21 «22 a 23 ' ' d ‘2j

• «:n a:v> *
.. a:]j

'• "n
a jo «I3 •
.. a,,

where atj is the ;th digit of the ith decimal fraction enumerated. Given any
such array representing a one-to-one mapping of natural numbers into real
numbers, a ‘missing* real number can be specified by the following diagonal
procedure. One missing real number is the decimal fraction A — 0.dx d2 ...
dk ... whose fcth digit dk is defined to be 7 if akk ^ 5, or 3 if akk > 5. This
‘diagonal’ number is clearly in the interval from 0 to 1, but it cannot be any one
of the numbers in the given array because it differs from each of them in at
least one digit. For any k, the diagonal number differs in its fcth decimal place
from the fcth decimal fraction of the array. Hence no enumeration of real
numbers can contain all of them—which proves that the real numbers are
nondenumerable, that is, that there are more of them than there are integers.
Now, we are in a position to consider another implication of the Power Set
Axiom. The Cantor diagonal proof just described is a special form of the proof
of a more general theorem, which states that for any set A, its power set (PA is
larger than A, itself. There is obviously a one-to-one mapping from A onto the
proper subset of (PA that consists of all singletons in (PA. (For every a in A, its
singleton {a} is in (PA, and s(a) = {a} is obviously one-to-one.) Now consider
any one-to-one function / from A into (PA, and form the subset A' of (PA that
consists of just those members a of A which are not members of/(a), the
element of (PA onto which they are mapped: A — {*: x f(x)}. Now,
190 Set Theory [Ch. 7

consider any a in A. If a £ A', then a $f(a) and so /(a) ^ A', but if a £ A',
then a £ f(a) and so again,/(a) ^ A'. But a is any member of A, so no member
of A is mapped by/onto A', which is, therefore, left out of the mapping/. But
/ was any one-to-one function from A to (PA, so (PA is larger than A.
Because Cantor’s principle of Comprehension was not restricted in the way
that the ZF Axiom of Separation is, Cantor’s set theory contained the set S of
all sets. According to the result derived in the preceding paragraph from the
Power Set Axiom, it follows that the power set (PS of the set S of all sets is
larger than S. But this is impossible, because every subset of S, being a set, is
a member of S. The Cantor paradox, then, is that given any set s there is a still
larger set (Ps, yet there is a largest possible set S. Cantor was aware of this
problem and communicated it to Dedekind in 1899.32 But Cantor did not work
out a satisfactory resolution of this difficulty. In the ZF set theory, the Cantor
paradox does not arise. The problem is solved by the more limited Axiom of
Separation that takes the place of Cantor’s too generous principle of Compre¬
hension. Although larger and ever larger sets can be produced by repeated use
of the Power Set Axiom, no ‘largest’ set is ever reached.
There are other definitions for ‘finite’ and ‘infinite’. One can define an infinite
set as a set that is equivalent to one of its proper subsets. Given this definition of
infinite, a finite set can then be defined as a set that is not an infinite set. This
pair of definitions can be proved to be equivalent to those first mentioned on
page 184, in the presence of the Choice Axiom, which will be introduced in
Section 7.6.
Still another definition of finite set, attributed to Dedekind, is ‘A set S is
called finite if there exists a mapping of S into itself such that no proper subset
of S is mapped into itself.’33

7.5 Natural Numbers and the Axiom of Infinity

Numbers are useful in many ways. Early in childhood we learn to use the
cardinal numbers ‘one’, ‘two’, ‘three’, ... to count and to answer the question
‘How many?’ And we learn also that the ordinal numbers ‘first’, ‘second’, ‘third’,
... can be used in ordering a collection of objects. Our task is to develop or
define these numbers within ZF set theory. We postpone our discussion of
ordinal numbers until Section 7.7. In Section 7.6 we shall consider cardinal
numbers in general. In this section we are concerned with the ‘natural num¬
bers’: 0, 1, 2, 3, ....
Mathematicians have two ways of characterizing a collection of (mathe¬
matical) objects that they propose to study. One way is to set forth an axiom
system that will provide an ‘implicit definition’ of those objects, from which—it

32See Cantor’s letter in van Heijenoort, op. cit.. pp. 113-117.

33 See Raymond L. Wilder, Introduction to the Foundations of Mathematics, 2nd ed., John
Wiley & Sons, Inc., New York-Lomlon-Sydney, 1965, p. 66.
Sac. 7.5]
Natural Numbers and the Axiom of Infinity 191

is hoped—all of the other facts concerning those objects can be derived as

theorems. The other way is to construct (or explicitly define) the new objects in
terms of old ones that are already at hand.
During the nineteenth century mathematicians carried out what has come to
be called the arithmetization of analysis . That involved the construction of
complex numbers, real numbers, algebraic numbers, rational fractions, and
negative numbers, on the basis of—or in terms of—the natural numbers 0, 1,
2, 3, ... using notions of intuitive set theory.3*1 But the natural numbers,
themselves, were developed axiomatically, by Dedekind and subsequently by
Peano. Peano s formulation of an axiom system for the natural numbers has
long been accepted as an admirably compendious characterization (or implicit
definition) of natural numbers, from which all their essential mathematical
attributes can be derived.
Peano s system is based on three primitive or undefined concepts: zero,
number, and successor. These are not defined explicitly hi the system, but
rather are implicitly defined by the axioms that make statements about them.
It is convenient to represent the successor of x as a': What Peano intended here
can be explained informally (and, of course, extrasystematically) by saying that
the successor of a number is obtained by adding 1 to it. The five axioms are

PI. 0 is a number.36
P 2. If x is a number, then xf is a number.
P 3. If x and y are numbers and x' = if, then x = //.
P4. There is no number x such that x' = 0.
P5. If qr() and for every number a, if qx then qx\ then for every number
x, cpx.

Peano himself acknowledged that not only the natural numbers satisfy these
axioms. For example, the sequence 1, 1/2, 1/4, 1/8, ... also satisfy them, if
‘zero' is taken to denote 1, ‘number’ is taken to denote the set of positive
fractions with numerator 1 and denominator any power of 2, and ‘successor of

M A lucid and detailed account of these constructions is found in Edmund Landau, Foundations
of Analysis, Chelsea Publishing Company, New York, 1951. Or see II. B. Enderton, Elements of
Set Theory, Academic Press, Inc., New York, 1977, or Suppes, op. fit.
35H. Dcdekind, 'Was sind mid was sullen die 7.ahlen?\ Brunswick, 1888, W. W. Beinan, tr., in
Essays on the Theory of Numbers, The Open Court Publishing Company, Chicago and London,
1901. See also his ‘Letter to Kefferstein’ (1890) 11. Wang and S. Bauer-Mengelberg, tr., in van
Heijenoort, op. cit. G. Peano, Arithmetics Principiu, Nova Methodo Exposita, Turin, 1889, tr., in
van Heijenoort, op. cit.
3f,Some people find it objectionable to call zero a number, on the grounds, for example, that if
there are no apples in a box, it is misleading to say that the box contains a number of apples, since
the number is zero. One should less misleadingly say that there are no apples in the box or no
number of apples in the box. Indeed Peano’s own original formulation of his system had the
natural numbers begin with 1 rather than with 0. However, the number 0 is so useful in our
notations and computations that it is accepted in mathematics in spite of possibly misleading
locutions to which it may give rise.
192 Set Theory [Ch. 7

x' is taken to denote x/2, But systems other then the natural numbers that
satisfy the axioms can be shown to be ‘just like the natural numbers, and could
be taken to be just different expressions for them.*
It is clearly desirable to characterize the natural numbers in a more univocal
way than Peano’s. However, Peano’s axioms are very useful: If we manage to
construct the natural numbers, somehow, we can check our constructions by
verifying that they satisfy the Peano axioms, for the latter do yield all of the
strictly formal arithmetic truths that mathematicians prize.
Frege, and later independently Russell/17 proposed a brilliant ‘coastructive’
definition of number, based on the following considerations. If A and B are
equinumerous classes, that is, classes between whose members there is a
one-to-one correspondence, we say that A and B have the same number of
members. Every pair set, for example, has exactly two members. Is ‘two’, then,
the name for some characteristic shared by all pair sets? Does ‘two’ signify the
common characteristic of all pairs? It may be that all pair sets share several
characteristics, so the number ‘two’ cannot be defined in quite this way. Frege
and Russell avoided the ‘positing’ of a unique shared characteristic by defining
‘two’ to be the class (or set) of all pairs—which is unique by Extensionality. The
Frege-Russell definition of ‘zero’ is the class whose only member is the empty
set. The definition of ‘one’ is the class of all singletons and the definition of
‘three’ is the class of all triples, and so on. There is much wisdom here and deep
insight as well as ingenuity. But it cannot be incorporated into ZF set theory
directly, because if there were a set of all singletons (or all doubletons, or etc.),
then its Union would be the universal set, which has been proved not to exist.
Zermelo proposed to identify the natural numbers 0, 1, 2, 3, ... with the sets
0, { 0 }> {{0}}» {{{ 0 }}},•• ■ Later, John von Neumann38 proposed the use
of an alternative and more convenient sequence of sets to define the natural
numbers. By the Axioms of the Empty Set, Pairing, and Union, we know that
each of the following sets exist: 0, {0}, {0,(0}}, {0,{0}, {0,{0}}},
... where each set u is followed by the set u U {u}. It is convenient to define
the successor set u+ of the set u to be the sum of u and {u}. Then, 0 is defined
to be the empty set; 1 is defined to be the successor set 0+ which is 0 U { 0 }
or simply {0 }; 2 is defined to be 1+, which is {0} U {{0}} or simply
{ 0, { 0 } }, and so on. On this approach, instead of trying to define n with
Frege and Russell as the set of all n-membered sets, which we have seen to be
impossible in ZF, we define n to be a particular and special n-membered set, a
representative n-membered set, that the ZF axioms guarantee to exist:

37 G. Frege, Die Grundlagen der Arithmetik, Verlag von Wilhelm Koebner, Breslau, 1884,
J. L. Austin, tT., The Foundations of Arithmetic, Basil Blackwell, Oxford, 1950. B. Russell,
Principles of Mathematics, op. cit. There is some difference between these authors’ formulations.
Frege regarded a number as the number of a concept, whereas Russell regarded a number as the
number of the extension of a concept, that is, of the class of objects satisfying the concept.
38J. von Neumann ‘Zur Einfuhrung der transfiniten Zahlen’, Acta litterarum ac scientiarum
Regiae Universitatis Hungaricae Francisco-]osephinae, Sectio scientiarum mathematicarum 1,
1923, tr., in van Heijenoort, op. cit.
Sac. 7.5]
Natural Numbers and the Axiom ol Infinity 193
Now we have the following:

0=0
1 = {0}
2 = {0,1}
3 = {0,1,2}

n+ = {0,1, 2,..., n }

These infinitely many sets exist, but is there a set containing all of them?
Another axiom is required to insure that there is.

ZF-7 Axiom ok Infinity. There exists a set containing 0 and

containing the successor of each of its elements.

By itself, as here stated, the Axiom of Infinity does not provide us with a unique
set whose only members are 0,1,2, 3.because other objects might belong
to a set, S, guaranteed to exist by the Axiom of Infinity. We shall call any set an
inductive set if it contains 0 and the successor of each of its members. It is clear
that 0,1,2,3,... all belong to every inductive set. The Axiom of Infinity states
that there exists an inductive set, S. Every inductive subset of S contains 0, 1,
2,3,.... so the intersection of all inductive subsets of S, which we shall call to,
exists by the Separation Axiom, is unique by Extensionality, and contains just
the natural numbers.

to = df n {Sj: Sj C S• S’j is an inductive set)

= {(),], 2,3,...}

To verify that <o satisfies the Peano axioms is fairly easy. That 0 £ to follows
from the fact that to is an inductive set, as does also the fact that if « £ to, then
u+ £ to, so PI and P2 hold for to. Axiom P4, which states that no u in to is such
that u+ = 0, follows immediately from the fact that u+ always contain u,
whereas 0 is the empty set 0 and does not contain any u. Because to is the
minimal inductive set, it follows that if any subset S of to is an inductive set,
then S = to. That is, if S C to, if 0 £ S, and if x+ £ S whenever x £ S, then
S = to. This is all that is stated by P5, which is a formulation of the principle of
mathematical induction.
To prove that P3 holds for every n in to, we introduce a new concept and
then prove two lemmas concerning it (a lemma is a subsidiary or preliminary
194 Set Theory [Ch.7

theorem introduced for the purpose of proving .some further theorem). The
new concept is that of a transitive set, defined to be a set that contains every
member of any of its members. Formally,

a is a transitive set = df(x)(y)[(x £ yy € a) D x £ a]

Presumably, a transitive set is so-called because £ between its members and

their members is a transitive relation. Alternative equivalent ways in which ‘a
is a transitive set’ can be defined are

UaCfl or (x)(x £ a D x C a) or a C (Pa

Next we establish the two lemmas.

Lemma 1. If a is a transitive set, then U (a+) = a.

Proof: U(a+) = U(« U {a}) (by definition of +)
= Ua U U {a}
= Ua U a
= a (because Ufl C a for a transitive set a)

Lemma 2. Every natural number n is a transitive set.

Proof: (by induction, appealing to P5) Let T be the set of natural numbers
that are transitive sets, that is, T = {x: xfwUxCx}.
First we observe that 0 £ T, since for every x, x $ 0, whence x £ 0 D x £ T.
Next we consider any natural number n and show that n £ 7" D n+ £ T.
Suppose n £ T. By Lemma 1, U(n+) = n, and since n+ = n U (n), n £ n+,
and so U(n+) £ n+. Therefore, n+ £ T. Now, by induction, T = co, that is,
every natural number is a transitive set.
Now we can prove P3, that for any n and in in co, if n4 = ra+, then n = m. If
n+ = m+, then U(n+) = U(m+). Since both n and m are in co, both are
transitive sets by Lemma 2. And by Lemma 1, U(n+) = n and U(m+) = w,
whence n = m.
This finally verifies that the von >feumann sets satisfy all of the Peano axioms
and, therefore, have all the usual arithmetic properties of natural numbers.
They are not the philosophically more plausible Frege-Russell numbers (which
cannot be sets in ZF) but are instances of the Frege-Russell numbers. Each
number n is itself equivalent to every set containing just that number n of
members. Hence they are admirably suited for counting the number of mem¬
bers in any finite set. They have, indeed, been called ‘counter sets'.39 Because
each natural number n is a set containing all of the natural numbers less than n,
the present definition is exactly in accotd with what was stated on page 184 in

39 W. V. O. Quine, Mathematical /x>gic. Harvard University Press, Cambridge, Mass., 1947,

p. 247.
Sec. 7.6] Cardinal Numbers and the Choice Axiom 195

defining a finite set. Each natural number, then, can serve the function of a
cardinal number and be used to answer the question ‘How many?' when asked
about a finite set.40 Some of their peculiar set theoretic features are surprising.
For example, each natural number n ‘belongs to' or ‘is a member of’ every
larger number, which may seem repugnant to common sense. But to a greater
or lesser degree, every scientific advance marks some departure from the
common sense that preceded it.41

7.6 Cardinal Numbers and the Choice Axiom

In Section 7.5 the natural numbers 0, 1, 2, 3,... were given a set theoretic
definition. Each of them is the number of members in some finite set or other
and is a finite set with that number of members. The Axiom of Infinity stated in
that section guarantees the existence of the infinite set co. In Section 7.4, it was
proved that the set co has fewer members than its power set (Pio, and more
generally that any set A has fewer members than its power set (PA. 1 here are,
then, infinitely many sets, each containing a different finite number of mem¬
bers, and also infinitely many more sets, each containing infinitely many
members. Thus there is an unending ascension of sets, each larger than any that
precedes it.
Cantor symbolized the cardinal number of a set A by A , where the two
lines over the symbol for the set signify a double abstraction from the concrete
nature of the set, first, abstracting from the particular nature of its members
and secondly, abstracting from the particular order in which those members
are arranged. It is more usual today to symbolize the cardinal number of a set
A by card A\ The essential nature of a cardinal number is given by

card A = card B EE A *** B

The Freee-Russell concept of a cardinal number as a set of all equivalent or

equinumerous sets cannot be used in ZF, so we took for our finite «.rd.na
numbers representative sets from among sets equivalent to each other. It is
customary to do the same for infinite cardinal numbers. We associate with each
set, whether finite or infinite, the concept of its size or magnitude. And we
conceive a cardinal number to be a measure of the size of any set whose
cardinal number it is. From each plurality of equivalent sets we select one
representative set to serve as the cardinal number of any of them. So any pair
of bivalent sot, has the tamo cardiml number For definiteness »o should
specify which set front among a plurality of equivalent sets ts selected as the

, .. 1^ nf cardinal numbers are the finite cardinals, and sets which

^ «
Be’, The Philosophical Review, vol. 74 (1965), pp.
198 Set Theory [Ch. 7

representative, or cardinal number. This has already been done for the finite
case. It will be convenient to delay this specification for the infinite case until
Section 7.7. In that section it will also be remarked*that the cardinal numbers
are serially (totally, or simply) ordered by the relation < (is less than). That is,
for any cardinal numbers m and n, if m ^ n then either m < n or n < m,
where the relation < is asymmetric, irreflexive, and transitive (these attributes
of relations were defined on pages 134-136). This feature of cardinal numbers
is, of course, necessary if they are to fulfill their intended function, which is to
measure the size of sets.
It was remarked in Section 7.4 that the set of all natural numbers, <o, is
equivalent to various other sets: the set of all perfect squares and the set of all
rational fractions. There is a single cardinal number, card to, that is the number
of members in each of these equivalent sets. We define the symbol for it:

K0 = df card co

this is the Hebrew letter aleph with subscript zero, and is read ‘aleph-nuH\ It
is the smallest infinite cardinal number.
Because the Power Set Axiom produces ever larger sets, there are infinitely
many cardinal numbers. Each is the unique representative of sets equivalent to
each other and to itself. Infinite cardinal numbers have an arithmetic that, in
some respects, is very similar to the familiar arithmetic of finite numbers,
although, in other respects, it is very different. The definitions of the funda¬
mental arithmetic operations on cardinal numbers are the same for both finite
and infinite cardinals.
Intuitively, the sum of two numbers, m and n, is the number p if and only if
there are disjoint sets M and N with m = card M and n = card N and p = card
(M U N). We define m -f n to be card (M U N) where M and N satisfy the
conditions just stated. In other words, the arithmetic sum of two cardinal
numbers x and y is defined to be the cardinal number of the set-theoretic sum
(or union) of any two disjoint sets X and Y that have those cardinal numbers,
respectively. That addition, so defined, is independent of the particular disjoint
sets M and N, can be seen from the fact that for any other disjoint sets M' and
N\ if M~M' and N~N\ then (MU N) ~ (M’ U N’). The familiar com¬
mutation and association laws for + follow from the commutation and
association of U.
The differences between arithmetic addition for finite and infinite cardinals
can be brought out clearly by the following contrasts. For finite numbers m and
n,m + n = m only if n = 0. But for infinite cardinal numbers, that is not true.
The set of natural numbers co can be decomposed into a finite and an infinite
part. For any finite number n, co = {0, 1, 2,..., n — 1} U (n, n + 1,
n -I- 2,...}, where the first summand has cardinality n and the second
summand has cardinality K0. Hence

w + — H0 4" ft — Hq (1)
Sec. 7.6] Cardinal Numbers and the Choice Axiom 197

for any finite number n, even where n ^ 0. Another contrast is the following.
The sum of any two finite numbers different from 0 is larger than either
summand. Because the set of all odd numbers {1, 3, 5, 7,...} and the set of all
even numbers (0, 2,4, 6,...} each has the cardinality K0 and their union is the
set of all natural numbers, we have

+ «o = K0 (2)

Finally, it should be remarked that for any finite cardinal numbers m and n,
where m < n, the result of subtracting m from n, (n — m) is unique. The result
of subtracting m from n, where m and n are infinite cardinals, however, is not
unique. (1) and (2) above show that K0 — K0 can equal any cardinal number
from 0 to H0 itself.
Arithmetic addition is defined also for infinitely many summands. If {mi} is
a set of infinitely many cardinal numbers, then where {Mi} is a set of infinitely
many sets, such that mi = card and i ^ j D Mi fl Mj = 0 for every i and
/, then by definition

y mi = card (U )
i

For example, the denumerable set of all natural numbers can be decomposed
into denumerably many sets, each of which contains dentnnerably many
natural numbers, as in the following (diagonal) array:

1 2 4 7.
3 5 8.
6 9.
10...

which shows that

Mo + *o + *0 + • • • = H0-

Intuitively, the product m*Ti (or inn) of two numbers tn and ti is the result of
adding m to itself n times. An alternative (equivalent) way to define the
product of m and n is to produce sets M and N where n = card M and
n = card N and then use the Cartesian product to define

= card (M X N).

It is easily seen that this product is unique (i.e., independent of the particular
sets M and N whose Cartesian product is formed), that it is equal to 0 if, and
only if, either M = 0 or N = 0. It is also evident that the product of m and n
198 Set Theory [Gh. 7

obeys the familiar commutation and association laws for multiplication and
that, for the infinite case as well as for the finite one, multiplication is distribu¬
tive with respect to addition. » '
There is, however, a complication in the case of multiplication that involves
infinitely many factors—even where there are ‘only* R0 factors and even where
each factor is finite. Suppose 911 to be a set of infinitely many disjoint nonempty
finite sets M{ where mi = card Mit and we wish to define the product IT
Intuitively, this number is the number of distinct selection sets, ja, where each ju.
contains exactly one element from each of the sets in 911. In the finite case
where911 = {{a, b}, (c, d, e}}, that is, where Mt = (a, b) andM2 = {c, d, e),
the distinct sets ju, are {a, c}, {a, d}, {a, e}f {b, c}, {b, d}, {b, e}. There are six
such sets, which is the expected product since 2 = card M1 and 3 = card M2.
The actual process of forming distinct sets ju, is simple. Select one member from
Mv and select one member from Af2. Form their pair set, which is our first ja.
Then select one member of M, and select a different member of M2. Next, form
their pair set, the second /a, and so on. Now, the problem in the infinite case is
this: an arbitrary choice must be made from among the members of each of the
infinitely many sets A/f of the collection 9fl—unless each set Mi (that is not a
singleton) has its members already arranged alphabetically, numerically, by
size, or in any other order that permits the statement of a rule that dictates
which should be chosen first for inclusion in the first fi, which next for inclusion
in the next ja, and so on. Now for any single finite set or even for an infinite
one, there is no great problem of making an arbitrary selection. But the notion
of making an infinite sequence of arbitrary choices is somewhat fanciful. What
is needed here is another axiom, one that asserts the existence of a selection set,
ja, that contains exactly one member from each of any number of sets,
Zermelo stated this axiom in approximately the following form:

ZF-8 Axiom of Choice (Axiom der Auswahl): If 911 is a set

whose elements are all sets that are different from 0 and mutually disjoint, its
union U9TI includes at least one subset ju having one and only one element in
common with each element o/911.42

Bertrand Russell called it the ‘multiplicative axiom’, because of its essential

role in defining multiplication for infinitely many cardinal numbers as factors.
One of his ingenious illustrations to clarify its role and need concerned a
mythical millionaire who accumulated R0 pairs of shoes and R0 pairs of socks.
The questions ‘How many shoes did he have? And how many socks?’ require
different methods of answering. Russell wrote:

The reason for the difference is this: Among boots we can distinguish
right and left, and therefore we can make a selection of one out of each
pair, namely, we can choose all the right boots or all the left boots; but
with socks no such principle of selection suggests itself, and we cannot be

42Zermelo, op. cit., in van Heijenoort, op. cit.

Sec. 7.6] Cardinal Numbers and the Choice Axiom 199

sure, unless we assume the multiplicative axiom, that there is any class
consisting of one sock out of each pair. Hence the problem.
We may put the matter in another way. To prove that a class has K0
terms, it is necessary and sufficient to find some way of arranging its terms
in a progression. There is no difficulty in doing this with the boots. The
pairs are given as forming an N0, and therefore as the field of a progres¬
sion. Within each pair, take the left boot first and the right second,
keeping the order of the pair unchanged; in this way we obtain a
progression of all the boots. But with the socks we shall have to choose
arbitrarily, with each pair, which to put first; and an infinite number of
arbitrary choices is an impossibility. Unless we can find a rule for
selecting, i.e. a relation which is a selector, we do not know that a
selection is even theoretically possible.43

Clearly, if we have one subset /a, we can get another by taking the relative
complement of each set — p, in turn, and using the Choice Axiom to get
another p, and so on. Hence the new axiom guarantees the existence of all the
selection sets jut in U3H, and legitimizes the definition:
If {m*} is a set of cardinal numbers, and if {Af,} is a corresponding set of sets,
such that card = mi for each i, then

1.1 mi = card (x Af-)

i ■i
The same definition in more familiar symbols would be written:

11 m{ = df card {p: (mi)(mJ)(3Mi)(3Mi)(m. = card Mr

rrij = card Mjm(i)(j)(i ^ / D Mi Pi Mi = 0)*
(3*x* € Mi-x e p-(yX(y € Mi-y € fO 3 !/ = *)))}
Another formulation of this useful axiom is

Axiom ok Choice (Alternative version): b'or any set M there is a

is a function/ (a ‘choice function for M) such that the domain off is the set of
nonempty subsets of M and f(S) £ S for every nonempty S C. M.

Here, in the case of defining the product of cardinal numbers mit the choice
function /for the set U 9K will select an element xt from each nonempty subset
M4 of U 3R; each determines a singleton (xf}, and the union of all of them will
be a selection set /a.
There are several theorems equivalent to the Choice Axiom, one of which
plays an important role in the next section.44 Much has been written about the

43Bertrand Russell, Introduction to Mathematical Philosophy, C,eorge Allen & Unwin Ltd.,
London, 1919, p% 126. .. .
44cwcnnnM nn no 243-250. A very full discussion of various applications of the Choice
Axiom tadiffwirt^ranches'of mathematics can be found in Fraenkel and Bar-Hillel, op. cit, pp.
59-74.
200 Set Theory [Ch. 7

Choice Axiom. Efforts were made to derive it from the other axioms of set
theory. None succeeded. Some people worried that a contradiction might arise
from adding the Choice Axiom to the other axioms. Those worries proved
groundless. In 1938 Godel proved that if the other axioms of set theory are
consistent, then the result of adding the Choice Axiom to them is also consist¬
ent.45 In 1963 Paul J. Cohen proved that the Choice Axiom cannot be derived
from the other axioms.46
Exponentiation of cardinal numbers follows the same pattern. In the finite
case, just as multiplication involves the addition of equal summands, so
exponentiation involves the multiplication of equal factors. We recall that for
finite nonzero cardinal numbers a, g\ d, the term ba is the result of multiply¬
ing b by itself a times, whence ba+c =ha*bv, (b*d)a = ha • da, (ba)c =
(hc)a = ,
and where h ^ 0, bli = 1 and where a 0, 0" = 0.
It is useful at this point to introduce a new notation for certain sets of
functions:

BA = df {/: f is a function on A to B}

This notation is intended to reveal and exploit the analogy with the familiar
notation for exponentiation. That there is an analogy is shown by the following
examples. Suppose that B = {bv h2, ..., frn},son = card 6; andlet A = {r/1},
so card A = 1. In this case there are exactly n functions in BA: fx(ax) = bv
f2(cil) = b2, ... = bn. In this case, card (BA) = (card B)CArdA. Now for
the same B, let A = {av a2}\ here card A = 2. In this case there are n2
functions in BA\

/ll(«l) = bl
/12(«j) = bx fvtflx) = • = b\

fn(a2) = b\ fvi(a2l = b2 fj‘h) = b:\ •/.„(« 2)=K

'"w

II

CM

/2l(al) = b2 /2 3(fll) = b2 • /2>l) = b2

. £
II

II
CM

/2l(fl2) = bl f 23(^2) = b3
a

II

/nl(«l) = K fn s(«l) = bn • • /»„(« 1) = K

II

fn 2(«2) = b2 mfn 3(a2) = b3 ■ ■ ■ /n»(«2) = bn

In this case also card (BA) = (card B)2 = (card B)cardA. In general, for any
finite nonempty sets A and B, card (BA) = (card B)K"ArdA. So it is plausible to
extend the definition of exponentiation to cover all cardinal numbers a and b,

45 K. Godel, ‘The Consistency of the Axiom of Choice and of the Generalized Continuum
Hypothesis’, Proceedings of the National Actulemy of Sciences, U.S.A., vol. 24 (1938), pp. 556-557.
See also The Consistency of the Axiom of Choice and of the Generalized Continuum Hypothesis
with the Axioms of Set Theory, Annals of Mathematics Studies, No. 3, Princeton, N.J., 1940.
46 Paul J. Cohen, ‘The Independence of the Continuum Hypothesis’, I and II, Proceedings of the
National Academy of Sciences, V.S.A., vol. .50 (1963), pp. 1143-1158, and vol. 51 (1964), pp.
105-110. See also Paul J. Cohen, Set Theory ami the Continuum Hypothesis, W. A. Benjamin, Inc.,
Menlo Park, Calif., 1966.
 Sec. 7.6]
Cardinal Numbers and the Choice Axiom 201
where a and b can be either finite or infinite:

= df card (BA) where a = card A and b = card B.

Just as cardinal addition is based on the union of disjoint sets that have the
cardinal numbers being added, and just as cardinal multiplication is based on
the Cartesian product of sets having the cardinal numbers being multiplied, so
cardinal exponentiation is based on mappings of one set onto another, where
the range has the cardinal number that is raised to a power and the domain has
the cardinal number that is the exponent. It is easily shown that for any
nonzero infinite cardinal numbers a, b, c, d the same equalities obtain as for
finite cardinals. So we have = 1, and for any finite cardinal a > 0,

No = *V*V*V- • ••«(, = K().

If the set B in BA is equal to 2 ( = {0, 1}), we have the set of all characteristic
functions of subsets of A. The definition is this:
For any subset A' C A the characteristic function

Fr : A —> 2 is such that

WNA'D/^r) = J)
and (x)(x £ (A - A') D fA.(x) = 0).

It is useful to think of the characteristic function fA,, for each A' C A, as a

condition on members of A, such that every member x of A for which
Ja'(x) = 1 satisfies the condition, and such that every member .r of A for which
fA>(x) = 0 does not satisfy the condition. The Separation Axiom states that any
such condition fAfx) on members of A determines a unique subset A' C A.
Here we have the converse of that statement, namely, that any subset A' of A
determines a characteristic function or condition(x). It is obvious, then, that
the set of all characteristic functions of subsets of A is equivalent or equi-
numerous to the set of all subsets of A. That is, for any set A, 2A ~ (PA, and in
particular, 2" ^ (Pu. Since {A': A'Cw}= (Pto and is a set by the Power Set
Axiom, we know that 2U ~ (A': A' C to}. Cantor proved that for any set S,
S < (PS, and in particular, < (Pio. We also know, by Cantor’s diagonal proof,
that io < {jc: x is a real number such that 0 < x < 1}, so w < ,R, where R is
the set of all real numbers.
Both (Pto and R are greater than w. It is not difficult to prove that (P<o — R.
There are as many real numbers between 0 and 1 as there are real numbers
altogether so that without any loss in generality, we can confine our attention
to real numbers in that closed interval. Every such real number is representable
as a binary fraction, 0.dyd2d3..., where each di is one of the two binary digits
0 and 1. (I is 0.111 ... 1 ...» 0 is 0.000 ... 0 ..., 1/2 is represented both by
0.100 ... 0 ... and by 0.011 ... 1 ...). Each binary fraction determines a
202 Set Theory [Ch. 7

unique subset of co, namely, the subset that contains the natural number n if
and only if dn = 1. Thus the binary fraction 0.111... 1 ... determines the set
co, itself; the binary fraction 0.000 ... 0 ... determiqes the empty subset of co;
and the binary fraction 0.101... 1... determines the subset of co that contains
every natural number except 2. The binary fraction 0.0101 ... 0101 ...
determines the set of all even numbers, where only dn’s with even-numbered
subscripts are l’s, and so on. Conversely, every subset A' of to determines a
binary fraction 0.dxd2d3 ..., in which dn = 1 if, and only if, n £ A'. Thus we
have a one-to-one correspondence between (Pco and (x: x is a real number such
that 0 < x < 1} and hence between (Pco and R, so (Poo ~ R, and card
((Poo) = card R. The usual symbol for card R is K, the cardinal num¬
ber of the continuum. Since 2W ~ (Poo, card (2W) = card ((Poo), and since
card (2W) = (card 2)cardl° which is 2*o, we have established that 2N«> = N.
As for K0, it is also true for N that N + N = N"N=:N. More infinite cardinals
exist by the Power Set Axiom. Some of them will be discussed in the next
section.

7.7 Ordinal Numbers and the Axioms of Replacement and Regularity

The elements of a set may be related to each other in a variety of ways: some
may be smaller than others, some may be elements of others, some may be
subsets of others, and so on. Any such relation among its members can be said
to impose some kind of order, even if only partial, on the set. It may be useful
to consider some examples.

1. Let S be the power set (P (a, fo, c) with the ordering relation that of set
inclusion, C.
2. Let P be the set of positive integers and let the ordering relation be
divides without remainder, d.
3. Let co be the set of natural numbers with the ordering relation less than
or equal to, <.

Each of these sets is partially ordered by the relation mentioned for it,
where we define partial order in a set X to be a reflexive, antisymmetric,
and transitive relation in X. These notions were discussed and defined in
Chapter 5 (pages 135-136 and 143). Their definitions in the context of ZF
set theory are

1. R is reflexive in X = df(x)(x £ X D xRx)

2. R is antisymmetric in X = df(x)(y)((x £ Xmy £ X) D ((xRyyRx) D
* = y))
3. R is traasitive in X = df(x)(y)(z)((x € X-y £ X'Z £ X) D ((xRyyRz) D
xRz)) • _ ‘
3®®- 7-7l Ordinal Numbers and the Axioms of Replacement and Regularity 203

A little thought will confirm that C is a partial order in (P{«, b, c}, d is a

partial order in P, and < is a partial order in 10. The first two sets are only
partially ordered by the relations mentioned for them, because there are
distinct members of each set that are not related to each other by the ordering
relation. The third set is totally ordered, or simply ordered, which—as the term
is generally used—includes or entails partial order.
A very useful kind of ordered set to work with is a well ordered set, which is
defined to be a partially ordered set, every nonempty subset of which has a first
(or least, or smallest) element, that is, an element that is related by the ordering
relation to every other element in the subset. The third of the preceding
examples is a well ordered set under the ordering relation mentioned for it. But
neither of the first two examples is well ordered under the relation specified for
it. (The subset of (P (a, b, c) containing just {a} and {b} does not contain a first
element, since neither {//} C {b} nor {h} C {r/}. And the subset of P con¬
taining just 2 and 3 does not contain a first element, since neither divides the
other without remainder.) Each of the first two examples is well ordered,
however, under another relation of its members. If we consider the relation of
alphabetical precedence, tt, among a, b, and c\ avrl), birc, and ottc, we can then
define a derived alphabetical precedence relation 11 among the members of
<P{a, b, c} in the obvious way. Our definition is first, for any subset a of
{a, fo, c}, if a ^ 0 then 0 11«, and if a ^ {a, b, <;} then all (a, b, c}, and then
{ x1} II { x2} if and only if xxttx2, { xv y x} 11 { x2, f/2} if and only if either xxttx2 or
xx = x2my]7ny2- It is obvious that b, c*} is well ordered under this lexico¬
graphic ordering. And the set P of the second example is well ordered by <. In
summary, a set may be well ordered by one relation, but not by another.
In developing the theory of well ordered sets, it is convenient to work with
the narrower relation of precedence <, or ‘less than’, rather than the disjunc¬
tive ‘less than or equal to’, <. The precedence relation is irreflcxive, and
asymmetric rather than antisymmetric, but like <, < is transitive. The first
remark to make about well ordered sets is that they are connected by the
relation that well orders them, that is, for any two distinct members xx ^ x2 of
a well ordered set, either xx < x2 or x2 < xv This follows immediately from the
well ordering condition, for any doubleton {xv x2] is a nonempty subset and,
therefore, contains a least element.
Two partially ordered sets are said to be similar if there is a one-to-one
correspondence between them that preserves order. That is, if X and Y are
partially ordered sets and there is a one-to-one correspondence f from X onto
Y, such that for any and x2 in X, xx < x2 (in X) if, and only if, f(xx) < f(x2)
(in Y), then X is similar to Y, symbolized as X s Y (sometimes as X s Y). It is
obvious that X s Y entails that X ~ Y. An infinite well ordered set can be
similar to a proper subset of itself. For example, the function m = 2n is a
one-to-one correspondence between the set of all natural numbers and the set
of all even numbers which preserves order in the two sets.
One important feature of any such mapping/, from a well ordered set X onto
204 Set Theory [Ch. 7

a subset of itself, is that x < f(x), for every x in X. This is proved as follows.
Suppose that there were an x in X, such that f(x) < x. Then since X is well
ordered, there must be a least such x, call it x0. It could not be the least element
in X, because /(x0) £ X and f(x0) < x(). For any y in X, if y < x0 then y </(«/)•
But f(x0) is such a y, so f(x0) < f(f(x0)). But since f(x0) < Xq, by the order
preserving nature of /, f(f(x0)) < /(x0). -Now by transitivity, /(x0) < /(x0),
which contradicts the irreflexiveness of <. So the supposition that there is an
x in X, such that f(x) < x, must have been false.
There is an important consequence of the result just obtained. To state it we
must define a new term. If X is a well ordered set, and a £ X, then the subset
{x: x £ X#x< a) is called the initial segment determined by a, and is usually
denoted by s(a). Now we can prove that no well ordered set can be similar to
one of its initial segments. For if X were a well ordered set containing a as an
element, and/were a similarity function from X onto s(a), then we would have
f(a) £ s(a), whence f(a) < a, which was proved to be impossible.
Many truths about well ordered sets can be proved by what is called the
principle of transfinite induction. It is analogous to the principle of mathe¬
matical induction discussed in Section 8.5, and, indeed, for finite sets the two
are equivalent. In the case of infinite well ordered sets, only the new principle
is applicable. It can be stated as follows:

If S is a subset of a well ordered set X, and if an element x of X belongs to

S whenever the initial segment determined by x is included in S, then
S = X.

Its proof in the general case is simple. If X — S were not empty, then it
would contain a least element, say x. That implies that every element of the
initial segment $(x) must belong to S and, by the induction hypothesis, x £ S,
also. But this would be impossible, because x cannot belong both to S and to
X — S. Thus X — S must be empty and since S C X, S = X. This principle, as
applied to natural numbers, is called ‘strong induction,’ and its use will be
illustrated in Chapters 8, 9, and 10.
If two simply ordered sets are similar, they are said to be isomorphic (with
each other). Ordered sets that are isomorphic with each other are said to have
the same order type. The most important order types are those of well ordered
sets. Some writers define ‘ordinal numbers’ as the order types of well ordered
sets.47 An order type of a given nonempty well ordered set, however, cannot be
defined in ZF set theory as the set of all well ordered sets that are isomorphic
with the given well ordered set, because the union of such a set of all similar
well ordered sets would be the universal set—which we know does not exist.
This situation is the same as that which made the Frege-Russell definition of
cardinal number incompatible with ZF set theory. In the case of cardinal
numbers, discussed in Section 7.6, we let an unspecified representative set from

47Raymond L. Wilder, op. cit., p. 127.

Sec. 7.7] Ordinal Numbers and the Axioms of Replacement and Regularity 205

among a plurality of equivalent sets serve as the cardinal number of any of

the sets equivalent to it. Which set would serve as the cardinal number was
specified only for the finite case. In dealing with ordinal numbers, we can
specify which well ordered set will serve as the ordinal number of any well
ordered set isomorphic with it.
Every natural number n is a subset of the well ordered set co and is a well
ordered set that contains all of the natural numbers less than n. For any natural
numbers m and n, as they were defined in Section 7.5, m < n if and only if
m £ n. Each n = {in: m £ wro < n}. The ordinal number of any finite well
ordered set is defined to be the natural number that is the number of elements
in the well ordered set in question. Given any natural numbtfVs m and n with
rn £ n, the initial segment s(m) of n determined by m is {.v: x £ nmx < m).
That is, every natural number is a well ordered set such that the initial segment
determined by each element in it is the same as that element. This characteri¬
zation of natural numbers as ordinal numbers can be generalized to give a
definition of ordinal number that includes the infinite case.*48 That definition is

nr is an ordinal number = df a is a well ordered set such that .v(£) = £ for

every £ in nr.

The first and smallest of the infinite ordinal numbers covered by this definition
is co. (Infinite numbers are also called transfinitc numbers.) Using the same
definition of successor set that was given in Section 7.5, lU = u U {#/}, but
now allowing u to range over ordinal numbers, we moved beyond to to to *, then
to (co+) + , and so on. These are easily seen to be ordinal numbers also. For if nr
is an ordinal number, then so is n+, which is nr U {«}. For if £ £ or+, then either
£ £ a or £ £ {n}. In case £ £ nr, then since a is an ordinal number, .v(£) = £; and
in case £ £ {o }, £ = «, in which case s(£) = a, that is, s(<\) = or. So in either
case s(£) = £ for any £ £ n+, and n4 is an ordinal number.
Before stating the next ZF axiom, which will permit the construction of
transfinite ordinal numbers out to any distance that might be desired, it will be
convenient to define the various arithmetical operations that can be performed
with ordinal numbers.
Well ordered sets can be combined by putting one directly after the other.
For this to be done, of course, the sets must be disjoint. But if the set A and B
that we want to combine in this way are not disjoint, w'e can easily obtain other
sets A' and B' that are disjoint, and such that A s A' and B s B'. A' can be the
set of all ordered pairs <a, 0> with a in A and B' can be the set of all ordered
pairs <fo, 1> with b in B. The obvious one-to-one correspondences are
a -» <a, 0) and b <h, 1>, with the order within A' and B' simply borrowed
from A and B. Therefore, given that A and B are disjoint well ordered sets, we

48 See John von Neumann, ‘Uber die Definition durch transfinite Induktiori und verwandte
Fragen der allgemeinen Mengenlehre’, Mathematische Atmaltn. vol. 99 (1928), pp. 373-391. See
also R. M. Robinson, ‘The Theory of Classes. A Modification of von Neumann’s System’, rlhe
Journal of Symbolic ljogic, vol. 2 (1937), pp. 29-36.
206 Set Theory [Ch. 7

define the order in A U B so that pairs of elements, in A and pairs of elements in

B keep the order they had, and let each element of A precede each element of
B. The fact that A and B were well ordered entails that A U B is well ordered
also and it is said to be the ordinal sum of A and B.
Now, we can define addition for ordinal numbers. Given ordinal numbers a
and ft, let A and B be disjoint well ordered sets such that a = ord A (i.e., the
unique ordinal number similar to A) and /? = ord B, and let C be the ordinal
sum of A and B. The sum a 4- ft is now defined to be ord C; so ord A + ord
B = ord (A U B). It is clear that the sum a + 0 is independent of the particu¬
lar choice of sets A and B. Any disjoint pair of similar sets would give the same
result.
Some of the properties of addition for ordinal numbers are quite similar to
those for cardinal addition. Some examples are a + 0 = a = 0 -f a, a +
1 = a+, and the Association Law a 4- (P + y) = (a + fi) -I- y. Another
property is the fact that a < p if, and only if, there is an ordinal number y
(^0), such that ji = a + y. But the Commutative Law fails for the addition of
ordinals, where at least one is infinite. Thus 1 + <o = w but <o + 1 ^ io, for if
we put a new element in front of an infinite sequence (of type u>), the result is
similar to the original sequence. If we put a new element at the back of the
sequence, however, the similarity is gone because the new set has a last
element and the original set did not.
By analogy with multiplication for natural numbers, we conceive the
product of well ordered sets A and B to be the result of adding A to itself
B times. To do so, we must have B disjoint sets, each familiar to A. These can be
produced as well ordered sets Ab = A X {b} for each b in B. Then, the set of
all these disjoint Ab’s, {Ab: b £ B), has, as its sum, the union U {Ab: b £ B},
ordered in the following way: <a, B) < <«', B') if and only if either b < B' or
(B = b' and a < a1). We define the ordinal product of two well ordered sets A
and B as the Cartesian product A X B, ordered as explained above. To define
the product of ordinal numbers a and ft, we introduce well ordered sets A and
B such that ord A = a and ord B = /?, and let C be the ordinal product of A
and B. Then the product of the ordinal numbers a and /?, written aj8, is defined
to be ord C. The easiest well ordered sets to introduce and use here as A and B
are the ordinal numbers a and J8 themselves, since each is a well ordered set
whose ordinal number is itself.
Some of the properties of ordinal multiplication are quite similar to those for
cardinal multiplication: a0 = 0 = 0a, al = a = la, the associative law
a(j3y) = (aj8)y, the left distributive law a (ft -f y) = afi 4- ay, and the fact that
if the product of two ordinal numbers is zero, then one of the factors must be
zero. But the commutative law for multiplication fails where at least one of the
ordinal numbers is infinite. For example, 2to = w, for the product is an infinite
sequence of ordered pairs. However, <j2 ^ to because this product is an or¬
dered pair of infinite sequences. The right distributive law fails too: (a + /?)y
is, in general, different from ay -h /Vy. For example, (1 -t- l)a> = 2w = w, but
lw + lco = w -|- w = w2 96 w.
Sec. 7.7] Ordinal Numbers and the Axioms of Replacement and Regularity 207

Just as ordinal products were defined in terms of repeated addition, so

ordinal exponentiation can be defined in terms of repeated multiplication. The
result of careful definition along this line will have the following desired
properties: 0" = Ofora > 1, la = 1,«^Y = «^«Y, and«^Y = («t^)Y. But not
all the familiar laws of exponents hold: (afi)y is, in general, different from «Y/JY,
since (2*2)lJ = 440 = <o, but 2U32U5 = toco = to 2. A few more words on this topic
might be in order. If m is finite and greater than l, mu' is the limit of the
expanding product m-m-m . .. which is an infinite sequence of m-luples and
thus, co. Alternatively conceived, mu' is the limit of the set of all finite ordinals
m” for n < to, so again m* = to. Similarly, ^ is the limit of the set of all
powerscon for n <C w- Since 1 + w + • * • ■+■ = we may also write
(Ow==l+co-f-co2-h • • • cok +
We turn next to the problem of whether all of the ordinal numbers discussed
in the preceding paragraphs really exist in the ZF system. Since to is a set, we
can use the Axiom of Pairing and the Union Axiom to form successor sets on the
‘far’ side of to. The successor of to, to1, is to U {to}, and is written—using the
notation for ordinal addition already introduced—as to + l. Next, (W)4 is
co+ U {to+}, and is written to + 2, and so on. It is obvious that each of them
exists as a set and is an ordinal number. But is there a set containing them allP
Just as we needed an additional axiom, the Axiom of Infinity, to prove that
there was a set of all the natural numbers, so here we need still another axiom
(or ket of axioms presented as an axiom schema) to prove that sets of ordinal
numbers beyond co exist. The needed axioms here are

Z F-9 Axiom Schema o f R k p l a c k m e n t . // ? ( {/) is a

formula such that for each member x of the set X, q (x, ij) and q(\\ z) imply that
y = z, then there exists a set S such that ij € S if and only ij there is an x in
X such that <p(x, [/)•

Now, if we specify A to be « and cp(x, y), as y = a + v, we have an Axiom of

Replacement that asserts the existence of a set containing

CO, CJ -f l, to 4" 2, . . •

The union of this set and co we denote by co2. It is easily seen to be an ordinal
number. After co2 come further ordinal numbers: u:2 + 1. »2 + 2 and so on.
After all of them comes <A whose existence as a set is asserted by another
specification of Replacement. Continuing, we get «3 + 1, «3 + ^ and so on,
and then co4. In the same way, we obtain further ordinal numbers. Thus we get
successively, co, co2, co3.By Replacement, again we obtain an ordinal
number, co2, that is related to these other ordinal numliers in the series in the
way that co is to the natural numbers. It is convenient to introduce new symbols
for some of the transfinite ordinal numbers produced by he Axiom of Re-
placement. For example, the ‘next’ ordinal after the sequence «o-, co« >, co ,
208 Set Theory [Ch. 7

... is usually written as t0. Also, there is an ordinal number which will be
discussed later. . .
It might be helpful at this point to look at some particular transfinite well
ordered sets composed of familiar elements. Three examples follow.
2n — 1
1. The set of all fractions ———, together with 1, ordered by the familiar

‘<\ This set looks like

0, 1/2, 3/4, 7/8, ...; 1

in which the initial segment s( 1) is clearly similar to to. The whole well ordered
set must have an ordinal number greater than to. (In fact, its ordinal number is
to + 1.)
2. The set of all odd numbers in order of magnitude, followed by the set of
all even numbers in order of magnitude. This set looks like 1, 3, 5, ...; 2, 4,
6,... and must have an ordinal number twice as large as to. (In fact, its ordinal
number is to + to or to -2.)
3. The set of all positive integer powers of prime numbers pn, ordered
according to the relation < defined as follows: p^ < pjn if either pm < pn or
both pm = pn and i < /. This set looks like

2, 4, 8, 16, ... ; 3, 9, 27, 81, . . .; 5, 25, 125, 625, .

and must have an even larger ordinal number. (In fact, its ordinal number is
to-to or to2.)
Some remarks about the Axioms of Replacement are appropriate at this
point. First, the condition in the hypothesis that <p(x, y) and <p(x, z) implies that
y = z serves to prevent taking as a specification of <p(x, t/), the formula x C y.
If that were allowed, then—since 0 is a subset of every set—taking X as { 0 }
would make S contain all sets, which has been shown to be impossible. Perhaps
it might occur to the reader that the cdXdjtion of being single-valued could
more simply be expressed by specifying a function/such that/(x) = y, rather
than invoking an expression like <jp(x, y). A function is well defined, however,
only if its domain is a set and a set can be specified that contains its range. The
Replacement Axioms are needed precisely to produce a set that is the range of
/. The Replacement Axioms are enormously powerful. Indeed, the Separation
Axioms can be derived from Replacement. Given any set A and any condition
<p(x), if we specify the X in the Replacement Axiom to be A and we specify
<p(x, y) to be (x = y and qp(x)), then the ‘<p(x, y) and <p(x, z) imply y = z part of
the hypothesis of the Replacement Axiom is satisfied. Its conclusion then asserts
(leaving out the trivially true *x = x’)

(3S)(x)(x.e S = X€ A-<px)

which is the Axiom Scheme of Separation. Finally, it should be remarked that

the Axiom of Pairing is derivable from Replacement plus the Power Set Axiom.
Sec. 7.7] Ordinal Numbers and the Axioms of Replacement and Regularity 209

Here, we specify the set A in Replacement to be the power set (P(P( 0), which
is {0, {0 }}. If (i and b are the two objects whose pair set is wanted, we
specify q>(x, y) to be (x = 0 -y = a) v (x = { 0 } -i/ = b). Then, for each .r in
(P(P(0), there is exactly one y such that qr>(x, y): for x = 0, y = a, and for
x = {0},y = b. With these specifications, the hypothesis of the Replacement
Axiom is satisfied and its conclusion then states that

(3S)(x)(.r £ S = x = a v x = b)

which is the Axiom of Pairing.

The unending ascension of ordinal numbers can be indicated by the follow¬
ing:

0, 1, 2, ..., a?, (o + 1, to + 2, ..., to2, co2 -f- 1, co2 + 2, ..., io3, ...,
w4, . . ., w2, w2 4- 1, co2 + 2, . .., to-' + to, to2 4- to + I, to2 + to + 2.
w2 4- w2, to2 4* <o2 4* 1, . . ., to2 4- <o3, . . ., to2 4 <a:4, . . ., to22, . . .,
to23,. . ., to3,. . . , to4,...»to1*',..., to*0, to*** ,..., i 0 4- 1» *'|) 4* 2,...,
Eo + ^ • i P0 + w2, ..., i0 4 to2, . .., f0 4 tou\ .. . ,
f02, . . . , f:0(0, . . . , F0tOw, . . . , f0z.-

Up to any point in this ascension, the set of all ordinal numbers, to that point, is
a well ordered set, W, which has an ordinal number greater than any member
of W. If there were a set of all ordinal numbers, $2, it would have an ordinal
number greater than any ordinal number in ll, that is, an ordinal number
greater than any ordinal. This contradiction is called the Burali-Forti Paradox,
and shows that there can be no set containing all ordinal numbers.49
In Section 7.6 much was said about transfinite cardinal numbers, but their
definition was postponed to the present section. We did say that a cardinal
number was a representative set, serving as the cardinal number of any set
equivalent to it. We want now tcTspecify which sets are to serve as these
representatives. To do so we mak&Wise of an important theorem proved by
Zermelo, using the Axiom of Choice, that every set can be well ordered/*0 We
know that repeated uses of the Power Set Axiom establish the existence of an
endless series of sets with ever larger cardinality. By Zermelos theorem, every
set is equivalent to some ordinal number. This suggests that we look to ordinal
numbers for a specification of the representative sets to serve as cardinal
numbers.
An infinite set, of course, can be equivalent to a great many different ordinal
numbers. In fact, all of the transfinite ordinal numbers indicated on this page
have cardinality of only’ R0- But all of the ordinal numbers equivalent to a
given set S form a set. The argument here is quite simple. The power set of S,

49See C. Burali-Forti, ‘Una Questione Sui Numeri TransfinitiV fiendiconti del Circolo Male-
matico di Palermo, vol. 11 (1897), pp. 154-64, tr. in van Ileijenoort, op. cit., pp. 104-11. See also
I. Copi, ‘The Burali-Forti Paradox’, Philosophy of Science, vol. 25 (1958), pp. 281-86.
50 E. Zermelo, ‘Beweis, das jede Menge wohlgeordnet werden kann\ Mathematische Annalen,
vol. 59 (1904), pp. 514-516. Tr. in van Heijenoort, op. cit., pp. 139-141,
210 Set Theory [Ch. 7

(PS, has higher cardinality than S. Any ordinal number, say a, that is equivalent
to (PS, has cardinality greater than the cardinality, of S, and a is therefore larger
than all of the ordinal numbers equivalent to S. Since every ordinal number fi
less than o is a member of o, a is a set that contains every ordinal number
equivalent to S. Now the Separation Axiom gives us a subset a' of o containing
all and only ordinal numbers equivalent to S. Since o' is an ordered set, it
contains a smallest or least element a, which can then be specified as the
representative set equivalent to S which is selected to serve as the cardinal
number of S. We can define cardinal number now as an ordinal number a such
that if fi is another ordinal number equivalent to «, then a < fi. It is clear that
this definition accords with the essential characteristic of cardinal numbers,
that card A = card B if and only if A ~ B.
It should be realized that powers and products of ordinals must not be
interpreted cardinally. If A and B are well ordered sets, then, in general, card
((ord B)ordA) < (card B)cardA. For example, in the case of 2* we have card
((ord 2)*) = card to = K0, but (card 2)card* = 2*o = «, and K0 < N. And in
the case of tou, we have (card to)cardw = = K, but card ((ord to)(ordto)) = K0
because all positive integers can be arranged into a well ordered set whose
ordinal number is to*. And the set of all positive integers has cardinality K0.51
The nondenumerable set R of all real numbers with cardinal N has an ordinal
number located beyond all of the ordinal numbers indicated on page 209.
There are ordinal numbers with nondenumerably many members. The set of all
ordinal numbers, up to some point beyond the location of the ordinal number
of the well ordered set of real numbers is, itself, a well ordered set. Therefore,
it has a least member—which is called to1, the least ordinal number with
nondenumerable cardinality, and its cardinal number is Kr There is, in fact, a
whole sequence of alephs with increasing cardinality. Their cardinal num¬
bers are

^1* • * * > ^n> * * ■ » «co+l> * * ' » ^to2» • *

We know that 2K° = N and that K x is the least cardinal number greater than
K0. From this, we know that < K. The much discussed continuum hypoth¬
esis asserts that Kj = K or Kj = 2*<>. The generalized continuum hypothesis
states that for every ordinal number a, Ka+1 = 2K«. Efforts to prove the
continuum hypothesis were made by such distinguished mathematicians as
Cantor and Hilbert. In 1939, Godel proved that if the axioms of set theory are
consistent, the continuum hypothesis cannot be disprovedl, and in 1963 Paul
Cohen proved that the continuum hypothesis cannot be proved from the
axioms of set theory either.52
The nine axioms for ZF set theory stated thus far are believed to be sufficient
for the development of mathematics. It might seem desirable to add a final

51 See A. A. Fraenkel, Abstract Set Theory. 3rd rev. cd., op. cit., p. 205.
52See the references cited in footnotes 45 and 46 on page 200. Gddel’s and Cohen’s results also
apply to the generalized continuum hypothesis.
Sec. 7.7] Ordinal Numbers and the Axioms of Replacement and Regularity 211

restrictive axiom stating that there are no sets other than those whose existence
is required or entailed by the given axioms. A hint as to the direction that such
an axiom might take is given by another paradox, due to Mirimanoff.53 It
involves the concept of a grounded set, defined as any set, .v, for which there
is no sequence of (not necessarily distinct) sets tjv t/2, i/3, ... such that ...
€ !/3 € */2 £ !/i € *■ Let W be the set of all grounded sets. If W is grounded,
then W £ W> whence ... W £ W £ W, so W is not grounded. But if \V is not
grounded, then there is a sequence of sets yv y2, i/a, ..., such that ...
€ */3 € J/2 € !/i € W, whence yx is not grounded, and therefore cannot be a
member of W. In either case, W involves us in a contradiction. The limited
kind of comprehension stated in the Separation Axiom would seem to prevent
the formation of such a set as W. But the very idea of a nonempty set with no
ultimate members, but containing an infinite descending sequence of sets,
seems repugnant. In view of the need to distinguish between a singleton {.v}
and its only member x, the idea of a set’s being a member of itself also seems
suspect. But for the limitation in the Separation Axiom, the very condition
x £ x would lead to the Russell paradox. This paradox can be generalized into
infinitely many others involving what might be called ‘£ -cycles’: x £ yy £ x,
x £ ymy £ zmz £ x, and so on.54 Such ungrounded sets, self-membercd sets, or
£ -cycle sets, all seem both repugnant to our thinking and of no possible utility
in mathematics or logic. They are all ‘ruled out’ by the final ZF axiom:

ZF-10 Axiom of Regularity (Axiom der Fundierung). Every

nonempty set A contains an element b such that A D b — 0.

That this axiom does rule out the undesirable sets in question is easily shown.
1. If A is a set, then A ^ A. If A = 0 , A has no members whatever so A $ A.
If A ^ 0, by the Regularity Axiom, there is an x in A, such that {A } PI x = 0.
Since {A} is a singleton, only A £ {A} so that x = A and {A} HA = 0 . But
since A £ {A}, it follows that A $ A, which was to have been proved
2. There are no sets A and B such that A £ B-B £ A. Suppose that there
were such sets: then both

A£ {AyB} HB and B £ {A, B} D A (1)

By the Regularity Axiom there is an x in {A, B}, such that {A, B} flx = 0. But
{A, B} is the doubleton of A and B, so either x = A or x = B. Therefore, either
{A, B} fl A = 0 or (A,B) PI B = 0, which contradicts (1), proving the
supposition false.
3. There is no ungrounded set i/0, such that there is a sequence of sets y v t/2,

ft3D. Mirimanoff, ‘Les Antinomies de Russell et de Burali-Forti et le Probleme Fondamental.de

la Theorie des Ensembles’, L’Enseignement Mathematique, vol. 19 (1917), pp. 37-52. See also
S. Yuting, ‘Paradox of the Class of All Grounded Classes Journal of Symbolic Logic, vol. 18 (1953),
p. 114 and R. Montague, ‘On the Paradox of Grounded Classes’, ibid., vol. 20 (1955), p. 140.
54See W. V. O. Quine, Mathematical Logic, op. cit., pp. 128 ft,
212 Set Theory [Ch. 7

t/3,..., such that... € jfo € ifc € Vi € f/0. For if there were such a set y0 and
sequence of sets y,, t/2» 1/3, • - -, we could use the Axiom of Replacement with X
specified to be <0 and y(x,;/) specified to be yx+ £.yx. The condition that for
each x in <p(x, y) and qp(x, z) implies y = z must be satisfied because our
second result shows that there can be no £ -cycles. Therefore, all of the sets yv
y2, 1/3, ... must be distinct. With its hypothesis thus satisfied, the Axiom of
Replacement would guarantee the existence of a set R = (x: x = yn £ w}.
Hence there would be a function / with to = dom f and R = ran /, where
f(n+) £ f(n) for each n £ to. Then, any member r of R would have to equal f(n)
for some n, from which it would follow that f(n+) £ R H r. This implies that
for every r in R, R D r ^ 0. That, however, would contradict the Axiom of
Regularity.
This single chapter was intended to provide a bare introduction to the
fascinating and important branch of mathematics known as set theory. Only a
few of its topics and problems have been mentioned and, of course, not
everything said has been proved. It is hoped that the readers’ interest and
curiosity have been sufficiently aroused to make them want to learn more
about set theory.

SUGGESTED FURTHER READINGS

Useful books that can be consulted with pleasure and profit are listed here.
Cohen, Paul J., Set Theory and the Continuum Hypothesis, W. A. Benjamin,
Inc., Menlo Park, Calif., 1966.
Enderton, Herbert, Elements of Set Theory, Academic Press, Inc., New York,
1977.
Fraenkel, Abraham, Abstract Set Theory, 3rd rev. ed., North-Holland Publish¬
ing Company, Amsterdam, 1966.
Fraenkel, Abraham, and Yehoshua Bar-Hillel, Foundations of Set Theory,
North-Holland Publishing Company, Amsterdam, 1958.
Halmos, Paul, Naive Set Theory, D. Van Nostrand Company, New York, 1960.
Hausdorff, Felix, Set Theory, 2nd ed., Chelsea Publishing Company, New York,
1957.
Jech, Thomas J., Lectures in Set Theory, Springer-Verlag, Berlin-Heidelberg-
New York, 1971.
Quine, W. V. O., Set Theory and Its Logic, Harvard University Press, Cam¬
bridge, Mass., 1963.
Stoll, Robert R., Set Theory and Logic, W. II. Freeman and Company, Pub¬
lishers, San Francisco and London, 1963.
Suppes, Patrick, Axiomatic Set Theory, D. Van Nostrand Company, New York,
1960.
 8
A Propositional Calculus

8.1 Object Language and Metalanguage

The logistic system to be constructed in this chapter is intended to be

adequate for the formulation of arguments whose validity depends upon the
ways in which statements are truth functionally compounded. Our logistic
system will be a language, although for the sake of rigor in developing its
theorems, it will be regarded as uninterpreted. We shall talk about this
language. It will be the object of our discussion, and is, therefore, called the
object language*. Because it is uninterpreted, its symbols and formulas have no
meaning, and we cannot use it until it is given an interpretation—which will be
postponed until after its development. We must, therefore, use a different
language in order to talk about our object language. The language used in
talking about a language is called a ‘metalanguage*. In any investigation of
language, there is an object language that is the object of investigation, and
there is a metalanguage that is used by the investigators in talking about the
object language.
An object language may be discussed from alternative points of view. Its
relationship to its users may be investigated, as in a study of dialect changes in
English usage and pronunciation in various parts of the country. Or the
meaning or interpretation of a language may be investigated, as in compiling a
dictionary, and in this latter inquiry a semantical metalanguage must be used.
Finally, the formal structure of a language may be investigated, as in a
grammar textbook, or in describing the development of theorems in an
uninterpreted logistic system, for which a syntactical metalanguage or Syntax
Language is used. In discussing the object language to be constructed here, we
shall sometimes use a semantical metalanguage, and sometimes a syntactical
one. In discussing its consistency, completeness, and its adequacy for expressing
all truth-functional compound statements, we shall have to use a semantical
metalanguage, for these topics involve its intended interpretation. In describ¬
ing the purely formal criteria for its well formed formulas and the syntactical
validity* of its arguments, only a Syntax Language need be used, for no
references to meanings have to be made in these connections.
The language to be used as metalanguage in discussing our logistic system
213
214 A Propositional Calculus [Ch. 8

will be ordinary English, plus elementary arithmetic, with the addition of some
special symbolic devices that will be introduced and explained as they are
needed. It is assumed, of course, that the reader understands the metalanguage,
for the entire discussion occurs within it. Only this one metalanguage will be
used. It will function in some parts of our inquiry as a semantical meta¬
language, in other parts as a Syntax Language.
It should be emphasized that object language’ and ‘metalanguage* are
relative terms. Any language, no matter how simple or how complex, is an
object language when it is being talked about. And any language (which must
be an interpreted or meaningful one, of course) is a metalanguage when it is
used to discuss an object language. Since our object language is uninterpreted,
we cannot use it as a metalanguage. In another context, however, where the
object language is an interpreted or meaningful language, one and the same
language can function as both object language and metalanguage. Thus in our
first chapter the English language was being discussed (and hence was the
object language), and the discussion was carried on in English (which was,
therefore, the metalanguage, also). A sufficiently rich or complex language can
succeed in formulating the whole of its own syntax, and a good deal of its own
semantics. But no consistent language can express the whole of its own
semantics; certainly not the truth conditions for all of its own statements. That
limitation can easily be shown.
That no language can express exhaustively its own semantics is shown by the
following version of the ‘Paradox of the Liar’.1 Consider the following English
sentence:

The sentence printed on page 214, line 25 of this book is not true.

Let us abbreviate the preceding sentence by the letter ‘S’. Now just as it is
obvious that ‘Snow is white’ is true if, and only if, snow is white, so it is equally
obvious that

‘S’ is true if and only if the sentence printed on page 214, line 25 of this
book is not true.
jm

But counting the lines and looking at the page number verifies that ‘S’ is
identical with the sentence printed on page 214, line 25 of this book. Hence

‘S’ is true if and only if ‘S’ is not true.

which is an explicit contradiction. That such a contradictory result can appear

as a consequence of apparently innocent assumptions ought not to be regarded
as a joke or a sophistry. It is a serious matter which reveals that the assumptions

'Attributed to J.Lukasiewicz.
Sec. 8.2] Primitive Symbols and WeH Formed Formulas 215

were not as innocent as they appeared. The source of the trouble is generally
agreed to lie in the attempt to formulate the truth conditions for the sentences
of a language within that language itself. At least, if we distinguish sharply
between object language and metalanguage, and do not try to make an object
language serve as its own semantical metalanguage, the contradiction does not
arise. An alternative method of avoiding such contradictions is discussed in
Appendix C at the end of this book.

8.2 Primitive Symbols and Well Formed Formulas .

We now proceed to construct our logistic system. There are two kinds of
primitive symbols in our propositional calculus: ‘propositional’ symbols and
‘operator’ symbols. We use just four of the latter kind, these being

(,)•

We want infinitely many propositional symbols, for which we use the first four
capital letters of the alphabet (in boldface type), with and without subscripts:

A Ai A., A,
B Bt B2 B3
C C, C2 Cj
D Dj D2 Dj

These are the only symbols that our propositional calculus will contain,2 and in
proving theorems and deriving conclusions from premisses within the system
they are to be regarded as being completely uninterpreted. They may be
thought of, prior to the assignment of meanings, as being repea.able and
recognizable marks rather than ‘symbols’ at all—though it will be convenient
to refer to them as ‘symbols’.
Of course, we are guided in setting up our logistic system by the interpreta¬
tion we intend eventually to give it. This intended interpretation controls our
choice of primitive symbols and also governs our syntactical definition of ‘well
formed formula’. A formula of our system is defined to be any finite sequence of
primitive symbols of our system. The set of all formulas of our system includes
such sequences as the following:

2 We could use just A and' to provide a finite base for the infinite set of propositional symbols
written as

A, A', A", A"’, A"", ■ ■ ■ ■

As will be shown in Chapter 9, we could dispense with the symbols '(’ and *)’.
216 A Propositional Calculus [Ch. 8

Bi
(AHA)
~(D)~(~)
~((Aa)-(C 3))
B2B3A7~( )( )•( )
)))((

Only some of these are to count as well formed formulas, however. Our
definition of ‘well formed formula* will be stated, of course, in our Syntax
Language. It is convenient to introduce some special symbols into our Syntax
Language, as an aid in discussing the logistic system clearly and economically.
Although the first four capital letters of the alphabet, with and without
subscripts, printed in boldface type, are symbols in our object language, those
same letters printed in lightface italic type are symbols in our metalanguage.
Their meanings in the metalanguage follow the convention that a lightface
italic letter that is an element of our metalanguage denotes or means that same
letter printed in boldface, which is an element of our object language. In
addition, we introduce the capital letters ‘P*, *Q\ ‘R*, ‘S’,..., with and without
subscripts, and refer to them as ‘propositional variables*. Whereas the propo¬
sitional symbols in our propositional calculus are uninterpreted and have no
meaning, the propositional variables ‘P*, tQ\ ‘R’, ‘S*, ... in our Syntax Lan¬
guage are interpreted and do have meaning. Every propositional variable of our
Syntax Language, until further notice, denotes any formula of our object
language—subject to the following restriction. In any sentence or sequence of
sentences of our Syntax Language, two distinct propositional variables, say ‘P*
and tQ\ may denote either two distinct formulas of our object language, for
example, and ‘'-'((A1)*(C3))\ or one and the same formula of our object
language. Although a propositional variable may denote any formula of the
object language in any one context, it must continue to denote that same
formula wherever it occurs in that context. Thus the propositional variables ‘P*,
‘R*, ‘S*,... of our metalanguage may have substituted for them any name
in the metalanguage of any formula of the object language. At this point, we
introduce the symbols *•*, ‘~~*, and ‘)* into the Syntax Language, and explain
their meanings. Where any propositional variable, say ‘P\ of our Syntax
Language denotes in some context a particular formula of our object language,
say ‘A*, then the symbol ‘~~(P)* of our Syntax Language will, in that context,
denote the formula ‘~(A)* of our object language. Also, where any proposi¬
tional variables, say ‘P* and ‘Q\ denote in some context two formulas of our
object language, say ‘A* and respectively, then the symbol ‘(P) •(())* of
our Syntax Language will, in that context, denote the formula ‘(A) •(~(B2))’ of
our object language.
We cannot simply give a list .of all well formed formulas of our object
language, since there are infinitely .many of them. It is necessary to give an
inductive or recursive definition of ‘well formed formula*, which can be stated
Sec- 8-2l Primitive Symbols and Weil Formed Formulas 217

as follows, using the symbolic conventions explained in the preceding para¬

graph.

Recursive Definition for Well Formal Formula:i

(a) Any propositional symbol is a wff.

(b) If any formula P is a wff, then — (P) is a wff.
(c) If any formulas P and are both wffs, then (P)-(Q) is a wff.
(No formula of the object language shall be considered to be a wff unless its
being so follows from these rules.)
*
This definition permits infinitely many well formed formulas, but it provides
an effective criterion for recognizing them. No matter how long a (finite)
sequence of symbols of the object language may be, our recursive definition
permits us to decide in a finite number of steps whether or not it is a well
formed formula. Let us take a relatively simple example to illustrate this:

~((A)-(~(B»)

The question to be decided is whether the foregoing is a wff or not. By part (b)
of the recursive definition, it is a wff, provided that (A)•( — (R)) is a wff. The
latter is a wff by part (c) of the definition, provided that both A and -~(B) are
wffs. The first of these is a wff by part (a) of the definition, and by part (b), the
second is a wff provided that B is a wff, which it is by part (a) of the definition.
Thus the formula in question is a wff. From here on we shall use the proposi¬
tional variables T\ ‘R\ ‘S’, and so on, of our Syntax Language to denote
only well formed formulas of our object language.

EXERCISES

Using the Recursive Definition for Well Formed Formulas, show which of the
following are wffs of our object language:

1. (~(A1))-(A1)
2. MHW-HCY))))
3. HHW’W))
4. M(~(0.)M~(D4)))
*5. (H(~)-(C2)))-M(B3)-(B3)))
6. ~(((A)*(B)W~(£)))
1 Alternatives are available here, as elsewhere, in developing logistic systems. In place of the
recursive definition in the text, we might have used the following:

(a) If P is a propositional symbol, then (P) is a n:jf.

(b) If P is a wff, then (—P) is a wff.
(c) If I>and {) are wffs, then (P-Q) is a wff. (No formula is a wff unless its being so follows, from
these rules.)

This alternative definition of wff would make different'formulas wffs, but the resulting logistic-
system could be developed equally well.
218 A Propositional Calculus [Ch.8

7. H(A)-(B)WH(A)-(A))))
8. (H«(A))*(B)*(C))))*((HA))*(B))*HC)))
9. ^((H(A)-H(B))))*(H(H(B)-(C)))-(^(H(C)-(A))))))
MO. -(((-((A)‘(-(B))))‘(A))*H-((A)‘(HB)1

On the intended or ‘normal’ interpretation of our logistic system, its propo¬

sitional symbols are to be symbolic translations of English statements that
contain no other statements as (truth-functional) component parts. Where P is
the symbolic translation of any English statement whatever, ~~(P) is the
symbolic translation of its negation. Where P and Q are the symbolic transla¬
tions of any two English statements whatever, (?)•(()) is the symbolic transla¬
tion of their conjunction. The question now arises: Is our logistic system
adequate, when normally interpreted, to express all truth-functionally com¬
pound statements? A system that is adequate to express all truth functions will
be said to be 'functionally complete . This is a special case of the ‘expressive
completeness’ mentioned in Chapter 6.
We begin our discussion by showing that our logistic system (which we call
‘R.S.’, since it is Rosser’s System4 5) is adequate to the formulation of some
familiar truth functions. Negations and conjunctions have already been men¬
tioned, but there are also disjunctions, conditionals, and equivalences to be
considered. Where two English statements translate into P and Q, their weak
disjunction (which is true if either or both are true) is expressible in R.S. as
—((/^(P)) •(—(()))). Since it is a very common function, we introduce an
abbreviation for it into our Syntax Language, defining ‘P v Q* to denote
identically the same wffs of die object language R.S. as are denoted by
‘^((^(P)).(~w({))))\5 Where two English statements translate into P and Q,
their strong disjunction (which is true if either one is true but not both) is
expressible in R.S. as (~((i^) “(^>))) •(—-(("—^(^)) *(^—(^))))-
The symbolic translation of an English truth-functional conditional state¬
ment whose antecedent has the symbolic translation P and whose consequent
has the symbolic translation Q will be ~~((P)*(~~(()))). This concept, too, is
quite common, so we introduce the symbol ‘D’ into our Syntax Language,
using *P D Qy to denote identically the same wffs of R.S. as are denoted by

4 The author is indebted to Professor J. Barkley Rosser of Cornell University for permission to
include the following material. The calculi presented here and in Chapter 9 are early versions of
logistic systems that appear in revised form in Professor Rosser’s Logic for Mathematicians,
McGraw-Hill Book Company, New York, 1953.
5Because ‘P v Q’ is an abbreviation of ‘—((^—(^))*(~(^?)))'* it is proper to say that P v Q is, or
is identical with, ^((~-(P))*(MQ)))» but not that P v Q is an abbreviation of the latter, for there
are no abbreviations in the object language itself. (The symbol V is a significant symbol of the
Syntax Language, but does not even occur in the object language, although we could introduce it
there if we wished.) A parallel may help make this clear. Since ‘U.S.S.R.’ is an abbreviation of
‘Union of Soviet Socialist Republics’, the U.S.S.R. is identical with the Union of Soviet Socialist
Republics. The U.S.S.R. is not an abbreviation of anything because it is a nation covering one-sixth
of the globe. Since ‘U.S.S.R.’ denotes a.large nation, the U.S.S.R. is a large nation and has no
meaning in any literal or semantical sense. Similarly, since in any context ‘P v Qf denotes a wff of
R.S., in that context PvQisa wff of R.S. Because R.S. is uninterpreted, its wffs are meaningless
and P v Q, in particular, is meaningless.
Sec. 8.2] Primitive Symbols and Well Formed Formulas 219

((P)' (~~(Q))) • Where two English statements have the symbolic translations
Pand Q> the statement that they are (materially) equivalent, that is, that they
have the same truth values, or that each (materially) implies the other, has the
symbolic translation (P D Q)-(Q D P). This is identically the same wff of R.S.
as (^((P)*(^(Q))))-(^((P)*(~~(P)))). Since material equivalence too is a
frequently used notion, we introduce the symbol<==> into our Syntax Language,
defining ‘P= Q9 as an abbreviation for ‘(P D £))•(£) D P)\
At this point, two notational conventions will be adopted for our meta¬
language. The first is dispensing with the dot (for conjunction), so that (P)((J) is
identically the same formula of the object language as (P) •(£>). The other
notational convention is to replace parentheses by brackets or braces wherever
such replacement is conducive to easier reading.
Even using brackets and braces, too many punctuation marks make reading
difficult, so two further conventions that permit a minimum use of punctuation
will be adopted. The first convention is to assign the following order of
precedence to the symbols

= D v • —

of our Syntax Language. Of these, each has greater precedence or greater scope
than any listed to its right. What is intended here can be explained by the
following examples. The otherwise ambiguous expression ‘^PQP denotes
(—(P)) •(()) rather than •(())), because the connective 4 • ’ (which we have
agreed to represent by juxtaposition) has greater scope than the operator ■*.
The scope of the connective ‘extends over that of the symbol. The
otherwise ambiguous expression *Pv^)R’ denotes Pv(()*R) rather than
(P v ()) • R, because the connective V has, by our convention, precedence over
the connective *•’, and its scope extends over that of the latter. The otherwise
ambiguous expression ‘P D Qv RS' denotes P D [() v (R • S)], because ‘ D* has
precedence over both V and *•’, and V has precedence over And the
otherwise ambiguous expression 4P D Q = D ~~P denotes [P D ()] =
[(—Q) D (—P)], because *=’ has precedence over ‘D’ and and D* has
precedence over Our second convention is that of association to the left,
which means that where the convention of order of precedence does not suffice
to remove the ambiguity of an expression, its parts should be grouped by
parentheses to the left. That is, when an expression contains two (or more)
occurrences of the same connective, and their relative scopes within the
expression are not otherwise indicated, the occurrence to the right shall be
understood to have the wider (or widest) scope. This, too, is best explained by
means of an example or two. Since all of its connectives have equal order of
precedence, the ambiguity of the expression P D Q D P D P ca ~.ot be
resolved by our first convention. According to our second convention, however,
we interpret it as denoting [(P D Q) ^ P] ^ P- Part, but not all of the ambi¬
guity of the expression T = Q = PQ v ~P ^ Q' is resolved by the convention
regarding order of precedence. Once this convention has been appealed to, we
220 A Propositional Calculus [Ch. 8

know that it denotes either [P= Q \ =[(PmQ) v (^P^Q)] or

P= {Q=[(P'Q)v (~~Pm~~Q)]}- We decide that,the former is meant by
consulting the second convention, which instructs .us’ to associate to the left.

EXERCISES

Write the following expressions of the Syntax Language in unabbreviated form,

complete with parentheses:
1. Pv ~~ P 6. PDQD P
2. —P D P 7 .PD (PvQ)
3. P{) D P 8. {P D iV) 3 (P 3 Q)
4. P{) v R 9. (P D ())(() D R) D (P D R)
*5. n(pD P) *10. P D —P^ —P

The preceding remarks show how to express some truth functions in R.S. But
to prove that R.S. is functionally complete, that is, adequate to express all
possible truth functions of any number of statements, more is required. We
must have in our semantical metalanguage a method of expressing all possible
truth functions. Then we must prove that all of these, or all their substitution
instances, can be expressed in our object language R.S. also, in its normal or
standard interpretation. Such a method of expressing all possible truth func¬
tions is provided by truth tables, which were introduced in Chapter 2. The
method and notation of truth tables is, therefore, imported into our semantical
metalanguage, and will be used freely in discussing the various semantical
properties possessed by R.S. on its standard interpretation.
Truth functions may have one, two, or any number of arguments (in the
mathematical sense in which an ‘argument’ is an independent variable). Thus
/(P) is a truth function of P if and only if its truth or falsehood is completely
determined by the truth or falsehood of P. Similarly, f(P, Q) is a truth function
of P and Q if and only if its truth value is determined solely by the truth values
of P and Q.
There are exactly four different truth functions of a single argument, and
these may be expressed by the following truth tables:

P m P UP) P UP) r UP)

TF TT TF TT
FT FF FF FT

The functions/,(P),/2(P),/3(P), and /4(P) are completely defined by these four

truth tables. They are the only truth functions of a single argument, and are
called ‘unary’, ‘monadic’, or ‘singulary’ functions.6 That they can be expressed
in R.S. in its intended or normal interpretations is easily seen. First we note that

6 Here we are using ‘truth function’ in the strict and proper sense of a nonlinguistic correlation
between or among tmth values.
Sec. 8.2] Primitive Symbols end Well Formed Formulas 221

the intended or normal interpretations of ~~P and PQ are given by the truth
tables

l_and f P Kt
T F T T T
FT T F F
F T F
F F F

That R.S. is adequate to express ft(P\ f2(P), /3(f>), and f4(P) is proved by
actually formulating them in R.S. The function f2(P) is true when P is true and
false when P is false, and is, therefore, expressible in R.S. as P itself. The
function ft(P) is false when P is true and true when P is false, and is, therefore,
expressible in R.S. as — P. The function f3(P) is false no matter which truth
value Passumes, and is, therefore, expressible in R.S. as ~~PP. The function f4(P)
is true in every case and can therefore be expressed in R.S. as the negation of
/3(P), that is, as —PP). We have thereby shown that all singulary truth
functions are expressible in R.S.
There are, of course, more truth functions of two arguments than of one
argument. These are defined by the following truth tables:

p Q MQ) P Q m q) P MQ) P V m v)
T T F T~ T T T T T T T T
T F T T F F T F T T F T
F T T F T T F T F F T T
F F T F F T F F T F F F

P Q MQ) P Q m Q) P V JjMQ)
P Q m q)
T T F T T F T T F T T T
T F F T F T T F T T F F
F T T F T F F T T F T F
F F T F F T F F F F F T

P P P <? fniZQ) P .UK 9)

Q m q) Q
T T T f" T T T T F T~ T F
T F F T F T T F F T F F
F T T F T F F T F F T T
F F F F F F F F T F F F

P Q fl3(P.Q) P Q fu&Q) P Q V) P Q fuMQ)

T T F T~ T T t" T F T~ T T
T F T T F F T F F T F T
F F T F F T F F T T
T F
F F F F F F F F F F F T
 [Gh. 8

These are all the truth functions of two arguments, and are called ‘binary* or
‘dyadic* functions. That they are all expressible? in R.S. is easily shown by
actually expressing them by means of the — and * symbols: for example,
/14(P, Q) is expressible as PQ while /^P, Q) is expressible as ~~(PQ)-

EXERCISES

1. Express each of the binary truth functions fx(P, Q), *f2(P, Q), ■ • ■ > */s(P> Q)> • • ■ >
7n(P, <?)>••■> /16(^ Q) ^ wffs of R.S.
2. There are 256 ternary (or triadic) truth functions: ft{P, Qf R), f2(P.Q, fl),...,
/256(P, R), each of which is completely determined (or defined) by a different
eight-row truth table. Take any ten of them and express each as a wff of R.S.

To prove the functional completeness of R.S., it is necessary to show that any

truth function of any number of arguments is expressible by means of — and •.
Any truth function of n arguments is expressible by means of a truth table
having n initial columns and 2W rows. Thus any truth function
f(Pv P2,..., Pn_v Pn) is completely specified by writing a‘T* or an ‘F* in every
one of the 2n places in the last column of the following truth table:

••• *>,-1 Pn
Row 1: T T T T
Row 2: T T T F

Row 2" — 1: F F ... F T

Row 2": F F ... F F

The truth function/(P,, P2,..., .,, Pn) must have either an F in just one row
of its truth table, or F’s is more than one row of its truth table, or F’s in no rows
of its truth table. In any case the truth function can be represented by a wff
of R.S.

CASE 1. f(PvP2.Pn_vPn) has an F in just one row of its truth table. If

the F occurs in the first row, then tfTe function is represented in R.S. by the wff
^(P1-P2-.. ••Pn_1’Pn), which has anF in the first row of its truth table andT*s
in all other rows. If the F occurs in the second row, then the function is
represented in R.S. by the wff ~~(P1-P2*.. .•Pn„1*~~Pn). The 2” distinct truth
functions, the ith of which has an F in its ith row and T*s in all its others, are
represented in R.S. by the 2n wffs

Sr -(PrP2-...-P^-PJ
S2: ~(P1-P2' - 'P„-I—P„)

S2»_x: . .-~Pn_i'Pn)
S2»: ~(~PX-~P2—Pn)
Sec. 8.2] Primitive Symbols and Well Formed Formulas 223

case 2. /(Fj, P2,..., Pn_v Pn) has F’s in more than one row of its truth
table. If it has F s in the k( 1 < k < 2n) rows il# i2, • • •, ik, then it is represented
in R.S. by the wff . .-S, .

case 3. /(F1# P2» • • •»Pn-v Pn) has F’s in no row of its truth table.
Such a tautologous function is represented in R.S. bv the wff
-'Pn-i-Pn).
The preceding cases exhaust the possibilities, so we have shown how to
express any truth function of any number of arguments in R.S. This is not a
theorem of or in R.S., but a theorem about it, established in our semantical
metalanguage. We may, therefore, list it as

Metatheorem I. R.S. is functionally complete.

A rigorous proof of this metatheorem requires the use of mathematical

induction, which we discuss in the following paragraphs.
We could have chosen to present a logistic system other than R.S. In addition
to parentheses and the propositional symbols l\ (), R, S and so on, we could
have selected as primitive operators any of the following pairs of symbols:
and V’, and ‘D\ *•’ and V’, ‘ and *D\ or ‘v’ and *D\ instead of and
It is instructive to inquire whether or not any of those alternative choices of
primitives would have made the resulting system functionally complete (on the
normal interpretation of the symbols V, lD\ which we shall denote by
*•’, V, ‘ D’). Before addressing ourselves to these problems, it will be well
to explain briefly the two kinds of mathematical induction we shall use in
establishing results about (not in) R.S.
We shall use the term weak induction’ to refer to the more commonly used
type of mathematical induction. The schema for weak induction is

/(l)
for any arbitrary m, if f(m) then f(m + 1)
there, f(m) for every m

It is frequently used in proving theorems in elementary algebra. For example,

one proves that the sum of the first n odd integers is equal to n2 by first proving
the two premisses of the above schema and then drawing the general conclu¬
sion indicated. The first premiss, which we shall call the a-case , is here
established as the trivial equation 1 - l2. Then, the second premiss, which we
shall call the 7?-case\ is established by assuming f(m) to be true for an arbitrary
integer m and from this assumption, deriving the conclusion that/(w ■+■ 1) is
true. In this particular proof, the ft-case assumption is

1 + 3 + 5 + • • • + (2m -- 1) = m2

We derive the desired conclusion by adding (2m + 1) to both sides of the

equation and performing an elementary regrouping and factoring of terms:
224 A Propositional Calculus [Ch.8

1 + 3 -I- 5 4- • ■ * + (2m — 1) + (2m -f L) = m2 + (2m -I- 1)

1 + 3 + 5 + • • • + (2m — 1) -|- [2(m -4-.1) — 1] = (m + l)2

which shows that for an arbitrary m, if the sum of the first m odd integers is
m2, then the sum of the first (m + 1) odd integers is (m + l)2. We thus have
established the fi-ca.se; from it and the a-case, by weak induction, we draw the
desired conclusion that for every m, the sum of the first m odd integers is equal
to m2. Weak induction may be thought of as summarizing an unending
sequence of arguments of the form Modus Ponens:

/(1) m /(m)
if /(l) then /(2); if /(2) then /(3); ...; if f(m) then f(m
•‘•/(3) .\/(m -h 1)

A proof of Metatheorem I by weak induction can be stated as follows:

a-CASE. Any truth function f(P) of just one argument can be expressed in
R.S., as was shown on pages 220-221.

fi-case. Here we assume that any truth function of k arguments

f(P\v P2> • • •» 4) can be expressed in R.S. and prove that any truth function of
k -h 1 arguments g(Pv *2* • • • * PM) can also be expressed in R.S.
Let g(Pj, P2,..., Pk, Pk+,) be any truth function of the k + 1 arguments Pv
P»..'.Pk.Pk+t. It is defined by the truth table:

r_l P2 •• Pk PM g(Pv P2, . , Pk. Pk+J

T T T T Vx
T T T F V2

T F F T V2k_,
T F F F V2fc
F T T T V2fc+1
F T T F V2*+2

F F FT V2fc+i_l
F F F F V2Jfc+i

where each Vt (1 < i < 2fc+1) is a T or an F.

A truth function fi(P2, ^3* • • •» ?k+l) of the k arguments P2, P3, ..., Pk,
Pk+1 can be defined by the first 2k rd\ys‘of the preceding truth table if we delete
its first column:
Sec. 8.2] Primitive Symbols and Well Formed Formulas 22S

e. n-fi j i
T T T T V.
T T T F V2

F F F T ^2*-l
F F F F v2*
and by the /?-case assumption can be expressed in R.S. by the m/f S,. A truth
>
function f2(P2, P.„ • ■ ■ ,
Pfc+ of the arguments P.,, P„..., P, . P, , .■
can be
defined by the last 2 rows of the first truth table if we delete its first column:

13 /. .now.)
T T T T V +1
T T T F

F F FT V2fc+i
F F F F V2fc+i

and by the /i-case assumption can be expressed in R.S. by the wff S2.
Now the truth table for the wff l\mSA has the same entries as the original
truth table in its first 2k rows and F’s in its last 2k rows. The truth table tor the
wff has all F’s its first 2k rows and the same entries as the original truth
table in its last 2k rows. From this, it is clear that the truth table for the wff
(1\'S 1)v(^F1*S2) has exactly the same entries as those in the original truth
table, whence the arbitrary truth function #(/*,,P2,. .., Pk, l\ f,) of k -f 1
arguments can be expressed in R.S. by the wff (1\ • S v (—1\ *S2) or
H_(Pi.Si)._(_JVS2)). This establishes the (i-case.
Metatheorem I now follows by weak induction from <v and /f.
The term ‘strong induction’ will be used to reier to the somewhat less
frequently encountered type of mathematical induction whose schema is

m
for any arbitrary m, ii f(k) for every k < m then f(m)
therefore, f(m) for every iu

Strong induction may also be thought of as summarizing an unending sequence

of arguments of the form Modus Pone ns:

/(1) /(]) and /(2)

if /(1) then /(2); if /(l) and /(2) then f(3);.. .;

77m
/(1) and /(2) and ... and f(m - 1)
if/(l) and /(2) and ... and/(in - 1) then /(w);...
_____ . -
226 A Propositional Calculus [Ch.8

To illustrate the use of strong induction, we prove that a symbolic logic

based on the propositional symbols P, p, R, S, . v and the operators • and v
is not functionally complete. We do so by proving that no well formed
formula7 of the system based on •, v, and P, Q, R, S, ... can express a truth
function that has the value true when all of its arguments have the value false.
In our proof we use strong induction on the number of symbols in the well
formed formula g(P, p, R,...), ignoring parentheses, and counting each oc¬
currence of P, p, R, ..., •, and v as one symbol.

rt-cASE: If g(P, p, R,...) contains just one symbol, to be well formed it must
be either P alone, or p alone, or R alone, or .... Where the arguments P, p,
R, ... all have the value false, g(P, p, A,...) will also have the value false,
since it is one of them. Hence any well formed formula of the present system
that contains exactly one symbol cannot have the value true when its argu¬
ments all have the value false.

/?-case: Here we assume that any well formed formula g(P, p, R,...)
containing fewer than m symbols cannot have the value true when all of its
arguments have the value fake. We shall prove, under this assumption, that
any well formed formula containing exactly m symbols cannot have the value
true when all its arguments have the value fake. Consider any formula
g(P, p, R,...) that contains exactly m symbols (where m > 1). To be well
formed, g(P, p, R,...) must be either

gl(P,p, R,...)-g2(P,p, R,...)

or
gl(P,p,R,...)vg2(P,p,R,...)

where gj(P, p, R,...) and g2(P, p, R,...) are well formed formulas containing
fewer than m symbols. By the /3-case assumption, when P, p, R,... are aMfake,
both gy(P, p, R,...) and g2(P, p, R,. . .) will have the value fake also. Now,
given any two propositions that are fake, both their disjunction and their
conjunction are fake. Hence in this case g(P, p, R,.. .) has the value fake, also.
This establishes the /3-case. *
Having established both the a and /3 cases, by strong induction we infer that
no well formed formula of the system based on •, v, P, p, R, ... can express a
truth function that has the value true when all of its arguments have the value
fake. Hence the system is not functionally complete.

EXERCISES

Prove the functional completeness or incompleteness of the logistic systems based

on the propositional symbols P, p, R, S, , parentheses, and the operators:
*1. v and — 3. D and *.
*2. D and • 4. D and v
7 We assume a recursive definition that is analogous to the one stated on page 217.
Sec. 8.3] Axioms and Demonstrations 227

5. —and +, where 4- is the symbol for exclusive disjunction, defined by the tmth
table for f7 on page 221.
*6. 3 and 4- 8. 4* and == 10. = and
7. v and 4- 9. 4- and • 11. EE and D
12. 13 and where is the symbol for (material) noniinplication.8 defined by the
truth table for /13 on page 221.
13. D and , where (Z! is the symbol for converse nonimplication, defined by the
truth table for f12 on page 221.
14. and </L
15. Which, if any, of the sixteen binary truth functional operators fv f>,. . ., /l6
defined on page 221, can be added to the propositional symbols P, (). ft, S,. . . and
parentheses, to give a functionally complete logistic system with a single operator
symbol? 4

8.3 Axioms and Demonstrations

The rules for our system and the proofs of the theorems within it will he
much simplified if we assume infinitely many well formed formulas as axioms or
postulates. Of course, we cannot actually write out an infinite list of axioms
within our object language, but we can use our Syntax Language to specify
exactly which wffs of R.S. are axioms and which are not. Ihere can be no
objection to having an infinite number of axioms if there is an effective process
for determining whether or not any given tuff is an axiom. The infinite list of
axioms of R.S. may be written as

Axiom 1. P 3 (P"P)
Axiom 2. (P-Q) 3 P
Axiom 3. (P =3 Q) 3 IMP**) :j M«**)l

Each of these syntactical formulas (formulas within our Syntax Language)

represents or designates an infinite list of wffs of R.S. lhus Axiom 1 designates
all the following:

M(AM~((A)*(A))))
M(»)*M(®W®>)))
M(C)-M(C)-(C))))

M(Ai)'(M(A ,)•(.*,))))

~((~(A)HM(~tA))'(~(A)))))
^((^(B))-(M(MB))*(^(*)))))
M((A) -(D))- (—(((A) * (D)) • ((A) • (D)))))
~(((A3)-(B7))-M((As) ■•(*?))'*((Aa)'"(*7)))))

8 go named by Alonzo Church in Introduction to Mathematical Logic, vol. 1, Princeton

University Press, Princeton, N.J., 1956, p. 37.
228 A Propositional Calculus [Ch. 8

and infinitely many more wffs of R.S. It designates, in fact, every wff of R.S.
that is of the indicated pattern. It is effectively decidable whether any given
finite sequence of symbols of R.S. is of this pattern or not. Three patterns are
set forth, and every wff of R.S. that exemplifies any one of these patterns is
assumed as an axiom.
That these axioms are reasonable assumptions is evidenced by the fact that
on the normal interpretation of the symbols — and •, they are all tautologies.
Our logistic system is intended, upon interpretation, to be adequate for the
formulation of arguments. Arguments, as we know, consist of premisses and
conclusions, all of which are expressed in statements. Corresponding to such
arguments, we have in our logistic system sequences of well formed formulas of
which the last formula is the 'conclusion’. We shall set up a criterion that will
enable us to distinguish formally between two kinds of sequences of wffs within
R.S.: those which become valid arguments when interpreted normally and
those which do not.
In Chapter 3 a valid argument was characterized as one for which a formal
proof or demonstration could be given. A formal proof of an argument’s
validity was defined as a sequence of statements, each of which was either a
premiss or which followed from preceding statements by an elementary valid
argument, and whose last statement was the conclusion of the argument that
was being proved valid. Thus the question of the validity of any argument was
‘reduced’ to the question of the validity of certain elementary argument forms
or rules of inference. Something like that will be proposed for R.S., but it is not
desirable to make such a wholesale set of assumptions about valid elementary
inferences as was made in Chapter 3, where nineteen rules of inference were
postulated, plus rules of Conditional Proof and Indirect Proof. However, some
rule of inference must be assumed for our logistic system or there will be no
inference legitimized within it at all. One rule of inference will suffice for the
validation of all arguments and (hence) for the proof of all theorems within R.S.
We assume Modus Ponens, and state it (in our Syntax Language) as

rule 1. From P and P D Q to infer Q.

Examples of arguments in R.S. that afe legitimized as valid by the assumption

of this rule of inference are

~((A)-(~B)))
A and (AjWAg)
B ~(C)

The assumption of Modus Ponens as Rule 1 (abbreviated as R 1) of our

logistic system, by itself* permits only a special kind of argument having two
premisses to be regarded as valid. R.S., however, is adequate for the expression
of all formally valid arguments thatard certifiable by means of truth tables,
including extended ones that contain any number of premisses. So it is desirable
Sec. B.3] Axioms and Demonstrations 229

to introduce a method for validating extended arguments in R.S. by showing

how their conclusions follow from their premisses by repeated applications of
R 1. Here our treatment is roughly analogous to that for extended arguments
presented in Chapter 3, although with some important differences. VVe define
a demonstration of the validity of an argument that has as premisses the
formulas P4, P2, • • • > Pn and as conclusion the formula to be a sequence of
well formed formulas Sj, S2.Sk such that every S; is either one of the
premisses Pv P2,..., Pn, or is one of the axioms of R.S., or follows from two
preceding S s by R 1 and such that Sk is {). As before, an argument is to be
regarded as valid if and only if there exists a demonstration of its validity. A
special notation is introduced (into the Syntax Language) to represent this idea.
We introduce the special symbol V (which may be read ‘yields') so that

^ r*.P„ b V

assorts that there is a demonstration of the validity of the argument that has l\,
P2,..., P„ as premisses and as conclusion. For example, all of the infinitely
many arguments denoted by

P D()
-m)
~~(RP)

are valid because P 3 Q, ~(()fl) I-(BP). The demonstration consists of the

following sequence of S’s:

S,: (P 3 Q) 3 l3 '-(HP)}
• S2:PD()
S3: -(QR) 3 -(HP)
S4: ~(OT
S5: '-(HP)

in which St is Axiom 3, S2 is the first premiss of the argument, S3 follows from

Sj and S2 by R 1, S4 is the second premiss, S5 follows from S3 and S4 by R 1, and
S5 is the conclusion. Any such sequence of S’s will be called a demonstration
that Pv P2, ..., Pn I- Q. Each S, will be called a line of the demonstration.
Where a demonstration is given for the validity of all arguments in R.S. of a
particular form, as in the example cited, any particular argument of that form
may be regarded as having been validated by that demonstration and the
general form may be regarded as a derived rule of inference.
The definition of demonstration given in the preceding paragraphs should
have made it clear that the axioms of R.S. function as ‘understood’ premisses of
every argument formulated within the system. Where those axioms are the
only premisses of a valid argument, then the conclusion is a theorem of the
system. That a given formula Q is a theorem, is expressed by writing ‘1- Q\The
230 A Propositional Calculus [Ch. 8

notation T Q’ is defined more strictly as the assertion that there exists a

demonstration of Q, which is a sequence of well formed formulas Sv S2,..., Sk
such that every S4 is either an axiom of R.S. or follows from two preceding S’s
by R 1 and Sk is Q.
It should be noted that demonstration, either of a theorem or of the validity
of an argument, is an effective concept. Given any sequence of S’s, however
long, it can be decided quite mechanically, in a finite number of operations,
whether or not it is a demonstration. It can be decided effectively of any S
whether or not it is an axiom (and whether or not it is one of the premisses, if
an argument is involved). And if some S, say S;, is neither an axiom nor a
premiss, then since only a finite number of S’s precede in the sequence, a
finite number of inspections will reveal whether or not two of the preceding
lines are S( and D Sj. Only if these occur earlier in the sequence of S’s will
follow from two preceding lines by R 1.
With this definition of theorem, the three axioms and the one rule of R.S. can
be regarded as a sort of symbolic machine for generating well formed formulas.
Each axiom itself generates an infinite number of wffs, and by repeated
applications of R 1 to them infinitely many more wffs are produced as
theorems. Two questions quite naturally arise at this point. First, is the system
consistent? And second, are all the theorems (on the normal interpretation of
the primitive symbols) tautologies?
A logistic system that is a propositional calculus (such as R.S.) will be called
analytic if and only if all of its theorems become tautologies on their normal
interpretations. R.S. is analytic, by this definition, provided that bP implies that
P is tautologous. We shall now prove that R.S. is analytic.

Metatheorem II. R.S. is analytic (that is, if kP then Pisa tautology).

Proof: We use strong induction on the number of uses of R 1 in the

demoastration that kP.

(\-case: Suppose P results from a single use of R 1. Then R 1 must be

applied to the axioms. The axioms are all tautologies, as is easily verified by
actually constructing truth tables forthem. Hence P results from applying R 1
to S and S D P, where S and S DP are both tautologies. Clearly, P must be a
tautology in this case, for if it were not, there would be an F in at least one row
of its truth table, say row j. But since S is a tautology, it has all T’s in its truth
table, including of course a T in row /. Hence m the jth row, S would have a
T and P an F, so that S DP would have an F—contrary to the fact that it is
tautologous and has only T*s in its truth table. Hence the Metatheorem is true
for m uses of R 1, where m = 1.

/3-case: Here we assume the Metatheorem to be true for any number

fc < m uses of R 1. That is, if S is a theorem whose demonstration involves
fewer than m uses of R 1, then S is a tautology. Consider the status of any P
Sec. 8.4] of the Axioms 231

that has a demonstration which involves m uses of R 1. The formula P is either

an axiom (in which case it is clearly a tautology) or must result from earlier
lines S and S D P by the mth use of R 1. The earlier line S is either an axiom (in
which case it is clearly a tautology) or is obtained by k < rn uses of R 1, and
hence is a tautology by the assumption of the ft-case. Similarly, the earlier line
S D P must also be a tautology. Now, because S and S D P are tautologies, P
must be one also, by the argument of the a-case.
From a and /?, by strong induction, we conclude that if bP (by any number of
applications of R 1), then P is a tautology, which means that R.S. is analytic.
The analyticity of R.S. having been established, its consistency follows
immediately, and may be regarded as a corollary of Metatlveorem II. In our
proof we use the Post criterion for consistency, which states that a deductive
system is consistent if it contains a formula that is not provable within it as a
theorem.

Corollary. R.S. is consistent.

Proof: The formula P* ~~Pis a wffof R.S. but is not a tautology; therefore, by
Metatheorem II, it is not a theorem of R.S. Hence R.S. contains a formula that
is not provable as a theorem, so R.S. is consistent by the Post criterion.

We define the ‘deductive completeness’ of a logistic! system for the proposi¬

tional calculus as the converse of analyticity. An analytic system is one, all of
whose theorems are tautologies. A propositional calculus will be called de¬
ductively complete when all tautologies are provable as theorems within it. The
deductive completeness of R.S. is more difficult to establish than the first two
Metatheorems and requires that we first develop some theorems of the system.
The proof, then, will be postponed to Section 8.0, following the development
of R.S. itself in Section 8.5, after we prove the independence of the axioms of
R.S. in the next section.

8.4 Independence of the Axioms

A set of axioms is said to be independent if each of them is independent of

the others, that is, if none of them can be derived as theorems from the others.
To prove the independence of each axiom, it suffices to find a characteristic of
wffs of the system such that:

1. The axiom to be proved independent lacks that characteristic.

2. All other axioms have that characteristic.
3. The characteristic in question is hereditary with respect to the rules of
inference of the system.

Roughly the same sense of ‘hereditary’ is used here that is explained in

Appendix A: a characteristic is hereditary with respect to a set of rules of
 [Ch. 8

inference if and only if whenever it belongs to one or more formulas, it also

belongs to every formula deduced from these formulas by means of those rules
of inference. The characteristic of being a tautology, pn the normal or intended
interpretation, is hereditary in the sense defined. But it cannot be used to prove
the independence of any axiom, because all axioms of R.S. have this charac¬
teristic.
To prove the independence of the axioms of R.S., we must use a model
containing more than two elements, a model similar to that used in Appendix A
to prove the incompleteness of the original nineteen Rules of Inference. For
R.S., a three-element model will suffice, but for other truth-functional axiom
systems, models containing more than three elements may be required.
To prove the independence of Ax. 1 of R.S., we introduce a three-element
model in terms of which the symbols and wffs of R.S. can be interpreted. The
three elements are the ‘values' 0, 1, and 2, which play roles analogous to those
of the truth values T and F. Every propositional symbol has one of these three
values assigned to it. Every u;jff that is not a propositional symbol will have one
of the three values 0, 1, 2 in accordance with the following tables:

_Q_ P-Q
0 2 0 0 IT
110 1 1
2 0 0 2 2
1 0 1
1 1 2
1 2 2
2 0 2
2 1 2
2 2 2

The defined symbol ‘D’, whose definition is given by

PDQ = df ~(P-~p)

has, pursuant to that definition, the* table

P P=>P
0 0 0
0 1 1
0 2 2
1 0 0
1 1 0
1 2 . 1
2 0 0
2 1 • 0
2 2 ' 0
Sec. 8.4] Independence ol the Axioms 233

The characteristic of wffs that we utilize here is that of having the value 0,
regardless of the values assigned to its component propositional symbols. This
characteristic is easily seen to belong to Ax. 2 and Ax. 3 of R.S. by the following
tables (analogous to the table explained on page 30).

^_V) d r
0 0 0 0 0
0 1 1 0 0
0 2 2 0 o
1 1 0 0 1
1 2 1 0 l
1 2 2 0 1
2 2 0 0 2
2 2 1 0 2
2 2 2 0 2

3 - («_• _P)J
(p Q) 2 _Ll (V_• JO
0 0 0 0 2 0 0 0 0 2 0 0 0
0 0 0 0 1 0 1 1 0 1 110
ooooo 0 2 2 0 0 2 2 0
1 1 0 1 2 0 0 0
0 110 1
0 110 0 J 2 1 11110
0 110 0 1 2 2 0 0 2 2 0
0 2 2 0 0 2 2 0 2 2 0 0 0
0 2 2 0 0 2 2 1 11110
0 2 2 0 0 2 2 2 0 0 2 2 0
1 0 0 0 2 0 0 0 0 10 11
1 0 0 0 1 0 1 1 0 0 12 1
1 0 0 0 0 0 2 2 0 0 2 2 1
1 1 0 0 10 11
10 10 1
1 2 1 0 0 12 1
10 10 0
1 2 2 0 0 2 2 1
10 10 0
2 2 0 110 11
112 0 0
2 2 1 0 0 12 1
112 0 0
2 2 2 0 0 2 2 1
112 0 0
ooo 0 0 0 2 2
2 0 0 0 2
0 0 12 2
2 0 0 0 1 0 1 1
0 2 2 0 0 2 2 2
2 0 0 0 0
1 1 0
0 0 0 2 2
2 0 10 1
1 2 1 0 0 12 2
2 0 10 0
12 2 0 0 2 2 2
2 0 10 0
2 2 0 0 0 0 2 2
2 0 2 0 0
2 2 1 0 0 12 2
2 0 2 0 0
2 2 2 0 0 2 2 2
2 0 2 0 0
 [Ch.8

The characteristic is easily seen to be hereditary with respect to R 1 of R.S.

by consulting the table given for 4 D\ In the only row in which both P and
P D Q have the value 0, Q also has the value 0. Hence if the characteristic
belongs to one or more wffs, it also belongs to every wff deduced from them by
R 1.
Finally, it is readily seen that the characteristic in question does not belong
to Ax. 1. When P is assigned the value 1, P D (P*P) has the value 1 rather than
0, for 1 D (1*1) is 1 3 2, which is 1. Hence Ax. 1 is independent.
Three questions naturally arise in attempting to prove the independence of
an axiom in a propositional calculus. First, how does one decide how big a
model to use? Second, how does one decide what values (elements of the
model) to assign in making a table for the primitive symbols of the system?
Third, how does one decide what element (or elements) of the model to
designate as the (hereditary) characteristic? There is no effective or mechanical
answer to these questions. Methodology here is still a matter of trial and error.
For simplicity and economy, one should try using a three-element model. If it
works, good. If not, one should try a four-element model, and so on. The
answers to the second and third questions are interrelated. To prove the
independence of Axiom 1, we wanted it to take on an undesignated value in at
least one case, where the other axioms take on only the designated value (or
values) in every case, and the designated value is hereditary with respect to
R 1. This need led us to assign, for at least one case, a different value for P-P
than for P (by assigning a different value (2) to P*Q for at least one case in
which both P and Q take on the same value (1)). Then to make Axiom 1,
P 3 (P-P), take on an undesignated value in at least one case, we specified that
the conditional with antecedent (value) 1 and consequent (value) 2 take on an
undesignated value, in this case 1. To show the independence of Axiom 2,
(P*@) 3 P, that is, to make it take on an undesignated value in at least one
case, we make its antecedent P*Q take on a designated value in at least one
case in which its consequent P takes on an undesignated value. Several such
assignments are possible, of course, but any chosen must be consistent with
other constraints, such as making the other axioms take on only designated
values, and making the designated value hereditary with respect to R 1.
To prove the independence of A*r 2 of R.S., we use the same three-element
model and the same table for 4~^P\ The difference lies in the table for *PmQ\
which follows, along with the derivative table for 4P 3 Q\

P_Q P-Q PDQ

0 0 0 0
0 1 0 2
0 2 . 2 2
1 0 0 0
1 1 • 0 2
1 2 2 2
Sec. M] Independence of the Axiom* 238

p Q np
2 0 2 0
2 12 0
2 2 2 0

The characteristic of wffs that we utilize here is (again) that of having the
value 0, regardless of the values assigned to its component propositional
symbols. This characteristic is easily seen to belong to Ax. 1 and Ax. 3 of R.S. by
the following tables.

P D (P • P)
0 0 0 0~ 0
10 10 1
2 0 2 2 2

(P D Q) D (g • R) 3 - (« • P)]
0000 20000 2 0 0 0
0000 20010 2 10 0
000000220 0 2 2 0
02 1 0 2 1000 2 0 0 0
0 2 1 0 2 1 0 1 0 2 10 0
021001220 0 2 2 0
022002202 2 0 0 0
022002212 2 10 0
02 2 0 0 2220 0 2 2 0
10 0 0 2 0000 2 0 0 1
1000 2 0010 2 10 1
100000220 0 2 2 1
12 1 0 2 1000 2 0 0 1
12 102 101° 2 10 1
12100 1220 0 2 2 1
122002202 2 0 0 1
122002212 2 10 1
1220 0 2220 0 2 2 1
20 0 0 2 0000 0 0 2 2
2000 2 0010 0 12 2
2000 0 0220 0 2 2 2
20 10 2 1000 0 0 2 2
2 0 1 0 2 1 0 1 0 0 12 2
2 0 1 0 0 122 0 0 2 2 2
2020 0 2200 0 0 2 2
2020 022 1 0 0 12 2
202002220 0 2 2 2
236 A Propositional Calculus [Ch.8

The characteristic is easily seen to be hereditary with respect to R 1 of R.S.

by consulting the table for 4 D\ In the only row in which both P and P D Q
have the value 0, Q also has the value 0. Hence if thh characteristic belongs to
one or more tvffs it also belongs to every toff deduced from them by R I.
Finally, it is readily seen that the characteristic in question does not belong
to Ax. 2. When Pand Q are both assigned the value 1, (P'Q) D Phas the value
2 rather than 0 for (1 • 1) D I is 0 D 1 which is 2. Hence Ax. 2 is independent.
To prove the independence of Ax. 3 of R.S., we use the same three-element
model and the same table for 4~~P\ The difference lies in the table for *P*Q\
which follows, along with the derivative table for T D ^)\

P Q P-Q PDQ
0 0 0 0~
011 1
0 2 2 2
10 2 0
112 0
12 2 0
2 0 2 0
2 12 0
2 2 2 0

The characteristic of wffs that we utilize here is (yet again) that of having
the value 0, regardless of the values assigned to its component propositional
symbols. This characteristic is easily seen to belong to Ax. 1 and Ax. 2 of R.S. by
the following tables.

p D (P • p)
0 0 0 0 0
1 0 1 2 1
2 0 2 2 2

[p • Q) D p
0 ~ <r~ 0 0 0
0 i 1 0 0
0 2 2 0 0
1 2 0 0 1
1 2 1 0 1
1 2 2 0 1
2 2 . 0 0 2
2 2 . i 0 2
2 2 2 0 2
Sec. 8.5]
Development of the Calculus 237
The characteristic is easily seen to be hereditary with respect to R 1 of R.S.
by consulting the table for ‘ D'.In the only row in which P and P D Q have the
value 0, Q also has the value 0. Hence if the characteristic belongs to one or
more wffs, it also belongs to every wff deduced from them by R 1.
Finally, it is readily seen that the characteristic in question does not belong
to Ax. 3. When P and Q are both assigned the value 1 and R is assigned the
value 0, (P D Q) D [~~(QaR) D ~~(ft*P)] has the value 1 rather than 0 for
[—(1*®) ^ ^(0-1)] reduces successively to 0 D [—2 D ^1], to
0 D [0 D 1], to 0 D 1, and finally to 1. Hence Ax. 3 is independent.

EXERCISES

For each of the following sets of axioms for propositional calculi, prove the
independence of each axiom:

*1. Church’s system PN has as primitive operators — and v, with P D defined as

~~Pv Its Rule is: From P and P D to infer (). Its four axioms are

Axiom l.(PvP) DP
Axiom 2. P D (P v Q)
Axiom 3. [Pv(pvR)] D ft) v(PvR)]
Axiom 4. (QDR)D [(P\Q) D (Pvfl)]

2. The Gotlind-Rasiowa system P(J has the same primitive operators, definition of D,
and Rule as PN above. Its three axioms are

Axiom 1. (P v P) D P
Axiom 2. P D (PvQ)
Axiom 3. (<? D R) 3 [(Pv Q) D (R v P)\

3. Frege’s system F.S. has as primitive operators — and D, with P v {) defined as

~~P D Q. Its Rule is: From P and P TJ {) to infer <;). Its five axioms (after deleting
Frege’s original Axiom 3, [P D (Q D R)] D [p D (P D fl)], which was not inde¬
pendent) are

Axiom 1. P D (Q D P)
Axiom 2. [P D (Q D R)] D [(P D Q) D (P DR) J
Axiom 3. (P D p) D D —P)
Axiom 4. — —P D P
Axiom 5. P D ~~~P

4. Lukasiewicz’s system L.S. has the same primitive operators, definition of v, and
Rule as F.S. above. Its three axioms are

Axiom 1. P D (Q 3 P)
Axiom 2. [F D (Q 3 H)] -* [(^ D Q) ^ (P ^ H)1
Axiom 3. (~P 3 ~Q) D (Q ^ P)

8.5 Development of the Calculus

In the actual development of derived rules and theorems of our object

language, we regard all of its symbols as completely uninterpreted, since we
238 A Propositional Calculus [Ch.8

are motivated by the desire for rigor in its development. When interpreted
normally, the formulas ~~RP and P~~R are logically equivalent. But they
cannot be so regarded as wffs of R.S., in its development, and the wff
—RP = P~~R cannot be accepted as a theorem until it has been formally
derived from the axioms of the system.
It is convenient to begin with the demonstration of a derived rule of
inference for R.S., which will validate infinitely many arguments in it. We state
it as

DR 1. P D Q, <? D R h ~~(~RP)

Its demonstration requires a sequence of just five well formed formulas, the
third of which we shall write twice, once in unabbreviated and once in
abbreviated form:

Sr (P 3 Q) 3 [~[Q~R) 3 ^RP)]
■V p => Q
S3; D
S3: (Q 3 R) 3 ~(~RP)
S4: Q D R
S5: ~(~RP)

That this sequence of S’s is a deinonstraction is easily verified. The first line, S1?
is Axiom 3 of R.S. It is true that there is an apparent difference between and
our syntactical formulation of Axiom 3:

St: (P D (?) => ^ ~~(~RP)]

Ax. 3: (P D (?) D [~(QR) D ~(KP)]

because Sx contains ~~R wherever Axiom 3 contains R. The point is that we are
here talking about well formed formulas of R.S. Both Sj and Axiom 3 denote
infinitely many wffs of R.S., and every wff of R.S. which is denoted by S1 is also
denoted by Axiom 3 (though not conversely). We can put the matter another
way. Our first derived rule (DR 1) validates infinitely many arguments formu-
lable within R.S., for example:

~((A )•(-(«))) ~((B)-(~(C)))

~((B)-(~(C))) and -((C) ■•(-(»)))
~((~(C))-(A)) ~((~(D))-(B))

as well as

~((A) -(~(A))) ~(((A) -(B)) -(~(~(C))))

~((A)«(~(A))) and ~((~(C))-(~((D1)-(D2))))
~((~(A))-(A)) .~((~((D1)-(D2)))-((A)-(B)))

The sequence of S’s in the given demonstration denotes infinitely many

sequences of wffs of R.S., in fact, one for each of the different arguments whose
Sec. 8.5]

validity is being demonstrated. The first line in the demonstration in R.S. for
the first of die four examples given is the wffoi R.S. denoted by Sx when T\
Q , and R are taken to denote ‘A\ ‘B’, and ‘C\ respectively. But this is
identically the same wff of R.S. denoted by our syntactical formulation of
Axiom 3 when P , and ‘R’ in it are taken to denote ‘A\ 4B\ and ‘~(C)\
respectively. Hence the first wff of the demonstration sequence is one of the
infinitely many axioms of R.S. supplied by Axiom 3. The situation is the same
with respect to the demonstration of the validity of every other argument in
R.S. that is validated by DR 1.
It is readily seen that the other lines of the sequence conform to the
requirements laid down in our definition of demonstration. The two premisses
of the argument occur as S2 and S4, while S3 follows from S, and S2 by R 1; and
S5 follows from S3(S3) and S4 by R 1. It is helpful to write in the ‘justification'
for each line in a demonstration: ‘Ax. 3* to the right of S,, ‘premiss' to the right
of S2, and so on. These labels are not part of the demonstration, but are helpful
to both the writer and the reader.
At this point the question naturally arises as to whether derived rules of
inference can be appealed to in deriving theorems from the axioms of the
system. It is most conveniently discussed in connection with an actual example.
Let us take as our first theorem of R.S. the formula

Th. 1. h~~(~~PP)

This formula follows directly from Axioms 1 and 2 by means of our first derived
rule, DR 1. A sequence of formulas which has already been demonstrated to be
a valid argument (by our demonstration oi DR 1) is

SA: P D PP Ax. 1
S2: PP 13 P Ax. 2
S3: DR 1

That S2 is Ax. 2 should be clear from the previous discussion. Every wff of R.S.
denoted by our syntactical formulation of Ax. 2, P(J D P, is an axiom, and
every wff of R.S. denoted by the syntactical expression ‘PP D P' is (also)
denoted by 'PQ 3 P\ so S2 denotes infinitely many of the axioms of R.S. that
are denoted by our syntactical formulation of Ax. 2. And that S3 really follows
by DR 1 from Sx and S2 is seen by observing that DR 1 validates any argument
of the form

PDQ
QD R

no matter what wffs of R.S. the formulas P, Qf and R may be. Thus DR 1
includes the case in which T and ‘R’ denote identically the same formulas,
while 'Q9 and ‘PP* also denote the same wffs.
 [Ch. 8

Although the sequence S1# S2, S3 may be regarded as a 'proof' of Theorem 1,

it does not constitute a demonstration, for by definition, a demonstration
involves the use of R 1 exclusively. But if we have a * proof ’ in which a derived
rule is used, as DR 1 is used in the given sequence, the demonstration of the
derived rule can be inserted into the sequence to produce a demonstration
proper. Thus in place of S3 in the given sequence, the demonstration of DR 1
can be substituted to produce

Sx: P D PP Ax. 1
S2: PP DP Ax. 2
S3: (P D PP) D [~~(PP~P) D —(—^FP)] Ax. 3
S4: —(PP~-P) D ~~(~~PP) R 1
SJ: (PP DP)D ~~(~~PP)
S5: -(-PP) R 1

This sequence is a demonstration because Sv S2, and S3 are axioms; S4 follows

by R 1 from and S3; and S5 follows by R 1 from S2 and S4. Moreover, the
present demonstration results from the earlier ‘proof’ by certain changes that
are indicated in the proof itself. The original sequence, which we call a ‘proof’,
is not a demonstration, but a description of a demonstration that can be given.
A proof may be regarded as a prescription or recipe for the construction of a
demonstration.
The situation is analogous to that which occurs when later theorems of a
deductive system are derived, not directly from the axioms, but from earlier,
already established theorems. Here, again, an example will aid the discussion.
We shall take as our second theorem the formula

Th. 2. h-PDP

that follows directly from Theorem 1 by definition. Its proof may be written as:

Sx: ~~(^^P~~P) Th. 1

Si:-PD P df.

This sequence is clearly not a demonstration, but it is a proof, for it tells us

exactly how to construct a demonstration of Th. 2. In place of Sx we need only
write our demonstration of Th. 1—or rather the demonstration of that version
of it which is relevant to the desired conclusion, Th. 2. The general statement
of Th. 1 is

|~ ~~(-~PP)

which denotes every wff that is of this form, no matter which wff Hie syntacti¬
cal variable ‘P’ denotes. Every wffdenoted by *— (- P~~P)’ is of that form,
and is therefore included among the infinitely many provable formulas of R.S.
SBC. 8.5] Development of the Calculus 241

that are labeled Theorem 1. The demonstration described by the indicated

proof of Th. 2 can be written as follows:

S\: ^P D -P-P Ax. 1

S2: -P-P D ^P Ax. 2
S3: (-P D -P-P) D t~(~P~i3-P) D -P-P)] Ax. 3
S4: -(-P-P--P) D -(-P~~P) R 1
Si: (-P-P ^ -P) D -(-P^P) df.
S5: P-P) R 1
S^:-P D P df.

Compared with demonstrations, proofs are shorter and are therefore easier
to write out. Since any proof can be made into a demonstration by replacing
any line that is a previously established theorem bv the demonstration of that
theorem, and by replacing any line that results from the use of a derived rule
by the demonstration of that rule, proofs can he regarded as shorthand
notations for demonstrations. However, proofs are different, and should not be
confused with demonstrations.
Before proceeding with the development of additional theorems and derived
rules for R.S., it should be remarked that Th. 2 can be equally well expressed as
b P v P, which is a version of the principle of the Excluded Middle. For, by
our definition of the symbol V, Pv P* is an abbreviation of
which has (Th. 2) —P D P' as an alternative abbreviation. In the latter form
it constitutes part of the principle of Double Negation.
Some additional theorems of R.S., together with their proofs (not demon¬
strations) are these:

H. 3. h ~~(QR) D (R D

Proof: b D Q Th. 2
b (—() D <;)) D [~(W 3 ~(R~~C>] Ax. 3
R 1
b ~(C«) D
1 —(QR) 3 (R D -{)) df.

ii. 4. b R D-R

Proof: 1-(~RR) D(RD Th. 3

b ~-(~RR) Th. 1
bR D ~~R R 1

H. 5. b (p d P) d (~p =) ~V)

Proof: b ~(p~R) =) (—F D ~~Q) Th. 3

b (e d p) d (~r 3 —e) df.

It will be observed that Th. 5 is part of the principle of Transposition, and

that Theorems 2 and 4 are parts of the principle of Double Negation. But
 [Ch.8

although both P D ~~^P and ~~~~P D P have been proved to be theorems,

the principle of Double Negation, P= which abbreviates
(P D ~~'^P)'(’--P D P), has not (yet) been proved to be a theorem. It would
follow from Theorems 2 and 4 by the principle of Conjunction, P, Q b P*Q, but
the latter has not (yet) been established as a valid principle of inference or
derived rule for R.S. These remarks are intended to throw additional light on
the meaning of the V symbol. Writing ‘b P’ asserts that there is a sequence of
wffs ending with P, which is a demonstration. Writing ‘b P and b Q' asserts that
there are two sequences of wffs, both demonstrations, one ending with P, the
other ending with Q. But writing ‘b P* Q’ asserts that there is a single sequence
of wffs, which is a demonstration, and ends with PmQ. This assertion, although
different, follows from the preceding by the principle of Conjunction, which
will be established as DR 14.
The next derived rule is proved as follows:

DR 2. —P D b() D P

(~P D ~p) D D ~(<?~P)] Ax. 3

~P D premiss
~(~W) b ~(^~P) R 1
~(~W) Th. 1
R 1
QDP df.

Although Theorem 5, b (£) DP)D (—P D —(}), is part of the principle of

Transposition, DR 2, —P D b Q D P, is not. That principle asserts that
(Q D P) = (—P D ~~()), which is our abbreviation for [(() D P) D
(—P D -—C?)] * [('~P 3 ^Q) D (Q 3 P)]. The left-hand conjunct is Theorem 5,
but the right-hand conjunct is not DR 2. There is an important difference
between

-PD-pbpDP and b (P D ~~ Q) D (Q D P)

The first of these asserts that there is a sequence of wffs, each of which is either
~~P D or an axiom or follows from two preceding wffs by R 1 and whose
last wffis Q D P. The second, however, asserts that there is a sequence of wffs,
each of which is an axiom or follows from two preceding wffs by R 1 and whose
last wffis (~~P D ~~Q) D (Q D P). (The second has not yet been established.)
Of course there is some connection between them as there is between any two
statements such as P b Q and b P D Q. Given the latter, the former is easily
established. To the sequence of wffs S^, S2,..., Sk (where Sk is P D Q), which
constitutes a demonstration for b P D Q, we need only add P as S^+1 and derive
Q as Sfc+2* f°r if follows from Sk and Sk+1 by R 1. But although b P D Q follows
from P b Q, the proof that is does is less simple. It will be established as
 Development of the Calculus 243
Metatheorem III (the ‘Deduction Theorem ); but until it has been proved, it
cannot be assumed to hold for R.S.
Some additional derived rules, together with their proofs, are these:

DR 3. P D pbflP D QR

Proof: (PDQ)D [—(pfl) D ~(RP)] Ax. 3

PDQ premiss
~(QR) D —(RP) R 1
RP D QR DR 2

DR 4. P D Q, R D S h ~[~(^S)(7^)]

Proof: PDQ premiss

SP D QS DR 3
RD S premiss
PR D SP DR 3
~~[~~(QS)(PR)] DR 1

DR 5. P D Q, Q D R, R D S h P D S

Proof: RDS premiss

(R D S) D (~~S D ~R) Th. 5
—S D ~~R R 1
(~S D ~R) D [—(~RP) D —(P~S)] Ax. 3
(—RP) D ~(P~S) R 1
PDQ premiss
QDR premiss
~(~RP) . DR 1
'(■P S) R 1
PDS df.

The last derived rule mentioned may be thought of as a ‘generalized' Hypo¬

thetical Syllogism. In developing R.S., it is convenient to establish DR 5 before
proving the familiar Hypothetical Syllogism P D Q D H 1- P D R. The
latter will be established as DR 6. For ease in proving the next derived rule, it
is useful to prove three additional theorems first. Their proofs are left as
exercises for the reader:

*T h . 6. h(R-~P) D (PR)

Th. 7. h PDF

Th. 8. h RPDPR

It is convenient to state and prove a corollary of Theorem 7:

Th. 7, Cor. hPv-F

244 A Propositional Calculus [Ck. 8

Proof: ~~PD ~~P Th. 7

df.
Pv~P df.

It will be instructive for the reader to construct a demonstration (not merely a

proof) for Theorem 7. It is so obvious a tautology as to seem trivial, and yet its
demonstration in R.S. is not short.
Together, DR 5 and Theorem 7 provide an easy proof of the validity of the
Hypothetical Syllogism

*DR 6. PDQ,QDR\-PDR

Some additional theorems and derived rules are useful in proving the next
Metatheorem.

Th. 9. b ~~(PR) D ~(RP)

DR 7. P D Q, R D S\-PR D QS

It is convenient to record two corollaries of DR 7:

DR 7, Cor. 1. P D Q h PR D QR
DR 7, Coh. 2. R D S h PR D PS
DR 8. P D Q, P D R h P D QR

*Th. 10. \-(PQ)R D P(QR)

A useful corollary of Th. 10 is the other half of the principle of Association

for

Th. 10, Cor. I- P{QR) D (PQ)R

DR 9. P D R, Q D S\-(PvQlD (RvS)
*DR 10. P D R, Q D R\-(PvQ) D R

Th. 11. h (PvQ)D(QvP)

Th. 12. h (PvQ)vRD Pv(QvR)

A useful corollary of Th. 12 is the other half of the principle of Association

for V:

Th. 12, Cor. Pv (Qv R) D (Pv Q)v R

Th. 13. \-[PD(QDR]]D-[PQDR]

*Th. 14. \-[PQDR]D[PD(QDR)]

Sec. 8.5] Development of the Cdcuhis 249

These last two theorems are the two halves of the principle of Exportation, but
before the principle itself can be derived from them, the principle of Con¬
junction must be established (as DR 14).

DR 11. P ID Q, P D (Q D R) b P D R
TH . 15. I-PD(()D PQ)

Th. 16. b P D (<? D P)

Having established these derived rules and theorems, we are now able to
prove the Deduction Theorem for R.S. as

Metatheorem 111. If Pu P2, ..., Pnh{) then 1\. P...

Pn-

Proof: We assume that Pv P2, ..., ,, Pn b {), that is, that there is a
demonstration or sequence of toffs Sl5 s2,... , SN, such that each S- is either an
axiom, or a P}(i = 1, 2,.. ., n), or follows from two previous S\s by H 1, and SH
is Q. Now consider the sequence of ioffs I’n => S„ Pn DS2.Pn D S„. If we
can ‘fill in’ wffs before each Pn D S4 in such a way that the resulting total
sequence is a demonstration from pvp2,...,pn v so each line of the resulting
total sequence is either an axiom or a Pt (i = 1, 2, ..., n — 1), or follows
from two previous lines by R 1, then since Pn D Ss is Pn D {)% we shall have a
demonstration that Pv P2, .. ., Pn { h P n D {l That we can ‘fill in’ to get the
desired demonstration, is proved by weak induction on the number of formulas
Pn D S( involved.

(a) In case i = 1, we have only the formula Pn D S, to consider. By as¬

sumption, S-j is either an axiom or a Pi (i = 1, 2, ..., n).

case I, Sj is an axiom. Here we fill in with the demonstration of Theorem

16, h S, D (Pn D St), and S, itself. From the last two formulas we derive
Pn D S1 by R 1. So the total sequence of wffs up to and including Pn D S, is
a demonstration that h Pn D Sl and hence that PJt P2, ..., Pn. x HPn D Sr

case 2. is a Pi (i = 1,2,..., n — 1). Here we fill in with the demonstration

of Theorem 16, h Sa D (.Pn D SJ, and Sv itself. From the last two formulas we
derive Pn D Sx by R 1, so the total sequence of wffs up to and including Pn D S\
is a demonstration that S1\- Pn D Sr Since Sx is a Pt (i = 1, 2, ..., n — 1), we
have a demonstration that Px, P2, .... Pn_i f" Pn ^ ‘V

case 3. Sj is Pn. Here we fill in with the demonstration of Theorem 7,

h Pn D Pn, that is, h Pn D Sv Thus the total sequence of wffs us to and including
Pn D Sj is a demonstration that h Pw D Sv and hence that Pv P2, ...,

pn-i
 [Ch.8

(/?) Now suppose that we have properly filled in all the lines up to and
including Pn D Sk_v so we have a sequence of wffs> .which is a demonstration
that Pv P2, • • • > Pn~ 1 ^ S/c-1- Under this assumptioh we show how to fill in
to include Pn D Sk in the sequence, which will then be a demonstration that Pv
P2, ...» Pn_j I- Pn D Sk. By assumption, Sk is either an axiom, or a
Pi (i = 1, 2,..., n), or resulted in the original demonstration from the appli¬
cation of R 1 to two previous S’s, say and (i, \ < k).

case 1. is an axiom. Insert the demonstration of Theorem 16,

b Sk D (Pn D Sk), and Sk itself, and derive Pn D Sk by R 1. The entire sequence
will then be a demonstration that P1 P2, ..., Pn_t b Pn D Sk.

Sk is Pi (i = 1, 2,..., n — 1). Insert the demonstration of Theorem

case 2.
16, b D (Pn D Sk), and Sk itself, and derive Pn D Sk by R 1. The entire
sequence will then be a demonstration that Pv P2, ..., Pn_1 b Pn D Sk.

case 3. Sk is Pn. Insert the demonstration of Theorem 7, b Pn DPn, that is,

b PM D Sk, and the entire sequence will be a demonstration that Pj, P2, ...,

case 4. Sk resulted (in the original demonstration that Pv ^2* • • •» Pn Q)>

from the application of R 1 to two earlier S’s, say St and Sy , where », / < and
Sj is of the form S; D Sfc. By the assumption of the fi-case, we have already
filled in up to and including both Pn D Sj and Pn D (S;. D SJ. By DR 11,
stated as

PDQiPD(QDR)\-PDR

we have

PwDS,.,PnD(S;.DSJhPnDSk

Insert the demonstration of this derived rule, whose last line is Pn D Sk, and the
entire sequence will then be a demonstration that Pv P2, ..., Pn_1 b PM D Sk.

Now by weak induction, we conclude that we can fill in for any number of
lines Pn D in such a way that the resulting sequence will be a demonstration
that Pv P2,..., Pn_j b Pn DSit We can, therefore, do it for the demonstration
of Pv P2,..., Pn_v Pn b Q no matter how many lines Sv S2,..., Ss it contains.
And since Ss is (), we can construct a demonstration that Pv P2, ...,
Pn_x b Pn D Q. This concludes our proof of the Deduction Theorem.
An immediate consequence is

MT III, Corollary: If.PV Q then b P D Q,

Sec. 8.5]

An equally obvious corollary is the more general conclusion that: The

Deduction Theorem holds for any propositional calculus that has only the rule
Modus Ponens and that contains demonstrations for P D P, P D (O D P), and
(P D Q) D {[P D (Q D R)] D (P D R)).
The way in which Metatheorem III (abbreviated ‘D.T.’) may be used in
proofs is indicated in the following proof of DR 6. First, we demonstrate the
relatively trivial DR 6P D Q, Q D R, Ph R by the following steps.

PDQ premiss
P premiss
Q R 1
QDR premiss
R R 1

Then we can prove DR 6 by simply applying the D.T. once, the proof reading-

P D Q, Q D R, Pi- R DR 6'
P D Q, Q D R I- P D R D.T.

Since the D.T. gives us an effective method of constructing a new demonstra¬

tion for DR 6 on the basis of the old one for DR 6', the preceding two step
proof is a perfectly adequate prescription for the desired demonstration. It
should not be thought that effort has been ‘wasted’ in constructing more
difficult proofs of our earlier theorems, for they had to be established first in
order to prove the Deduction theorem itself.
Some additional theorems and derived rules of R.S. are

Th. 17. (-PD(pvP)

*Th. 17, Cor. hPD(Pv<;))

Th. 18. f- (P v Q)R D (PR v QR)

This theorem constitutes a part of the principle of Distribution—the distribu¬

tion of with respect to V.

DR 12. PD-phPD ~(QR)

DR 13. PD~RhPD -(QR)
*DR 14. P,Q\-PQ

Here, finally, we have the principle of Conjunction. It permits us to establish

the principle of Double Negation as our next theorem, which follows directly
from Th- 2 and Th. 4 by DR 14.

Th. 19. hP= ~~P

248 A Propositional Calculus [Ch.8

Theorem 19, by itself, however, does not permit us to replace by P

wherever it may occur in the interior of a larger itfj[f.*That is, where we are able
to demoastrate a wff b (-—P...), the mere equivalence of ~~^P and P
does not permit us simply to infer that h (—P...). The inference would be
valid, but we must prove that it is valid within R.S. The legitimacy of any such
replacement is asserted by our next Metatheorem. Before stating and proving
it, we shall find it convenient to establish the following:

DR 15. P=gb-P=-t>
DR 16. P=Q, R = S\-PR = QS
DR 16, Cor. P= Q9 R = S b Pv R = Q v S

Proofs will be left as exercises for the reader.

Now we are ready to prove the Rule of Replacement for R.S.

Metatheorem IV (Rule of Replacement). Let be any

toffs, let be any wff that does not occur in any Pi9 and let S be any wff that
contains no components other than Q and P{ (1 < i < n). Where S* is a wff that
results from replacing any number of occurrences of Q in S by R, then
Q = R bS = S*.

Proof: Strong induction is used on the number of symbols in S, counting

each occurrence of or any Pi as a single symbol.

(a) n = 1. In this case S is either Q alone or a P{ alone.

case 1. S is {) and S* is R. It is obvious that Q = R b Q = R, which can be

written £) EE fl b S = S*.

case 2. S is Q and S* is Q also. Since b Q = Q by Th. 7 and DR 14,

b S EE S*, whence {) = R b S EE S*.

case 3. S is a Pt. In this case S* also is the same P{. Since bPi = P{ by
Th. 7 and DR 14, b S = S* wherice Q = R b S = S*.

(fi) Here the Metatheorem is assumed to be true for any S containing < n
symbols. Consider any S containing exactly n symbols (n > 1). It is clear that S
must be either or S1,S2.

case 1. S is ~-'S1. Because S contains n symbols, Sj contains < n symbols, so

by the /I-case assumption, Q = R b Sx = S *, where S * is a wff that results
from replacing any number of (Occurrences of {) in SL by R. But
= S* b ^(Sx) = ^(5!*) by DR 15, and since any — (S/) is obviously the
same as a (^5!)* or S*, we have Q = R b S EE S*.
Sec. 8.5]

case 2. S is S1*S2. Here Sj and S2 each contains < n symbols, so by the

P-case assumption,^ = R\-SX=S* andpEERh S2 = S2*. But by DR 16,
Sj == Sx , S2= S2* h Sa • S2 = Sj*«S2* and since anv S* is an S/vS.,\
Q = R hS = S*. '

Hence, by strong induction, we infer that regardless of the number of symbols

in S, Q = R\~S = S*.

M T IV, Corollary: If Q, ft, S, and S* are as in Metatheorem /V, then

Q = R, S h S*.

The proof of the corollary is obvious.

Of the list of Rules of Inference used in proving the validity of arguments
back in Chapter 3, the first nine were elementary valid argument forms proper,
and the last ten were equivalences whose mutual replaceability was assumed.
Of the first nine, the first, Modus Ponens, is the primitive rule R 1 of R.S. The
third, the Hypothetical Syllogism, has already been established as DR 6; and
the eighth, the principle of Conjunction, has been proved as DR 14. The
remaining six rules can easily be proved as derived rules of R.S. They may be
listed as

DR 17. FDp, ~~Q\-~~P (Modus Pollens)

DR 18. PvQ, ^P\~Q (Disjunctive Syllogism)
DR 19. PQ h P (Simplification)
DR 19, Cor. PQ b Q
*DR 20. (P D Q)(R D S), PvRtQvS (Constructive Dilemma)
DR 21. (P D Q)(R D S), v -S h ~~Pv ~~R (Destructive Dilemma)
DR 22. P\~Pv Q (Addition)

The various equivalences that made up the last ten of the nineteen Rules of
Inference can easily be established. The principle of Double Negation has
already been proved as Theorem 19. The various principles of Commutation
and Association are obtained by simply applying DR 14 to already established
theorems and corollaries.

Th. 20. h PvQ=QvP (Commutation of V)

TH 21. h PQ = QP * (Commutation of *•’)

By Theorem 21, we have h (P=Q) = (Q = P), which can be used, without

any special mention, in proving the following theorems.

Th. 22. h[Pv(QvR)]=[(PyQ)vR\ (Association of 4v’)

Th. 23. \-P(QR) = (PQ)R (Association of

250 A Propositional Calculus [Ch. 8

The principle of Transposition is obtainable by DR 14 from Theorem 5 and the

result of applying the Deduction Theorem to DR 2.

Th. 24. b (P D Q) = D ~~P) (Transposition)

The proof of the principle of Exportation is even more obvious.

T h . 25. b [(PQ) D R] = [P D (Q D R)] (Exportation)

The proofs of the final group of theorems of the R.S. will be left as exercises for
the reader.

Th. 26. h P=E PP (Tautology)

Th. 26, Cor. b P= Pv P (Tautology)

T h . 2 7. b ^(PQ) = (—Pv ^^)) (De Morgans Theorem)

Th. 2 8. b ^(P v Q) = (De Morgan’s Theorem)

Th. 29. b(PD^)E (^Pv @) (Material Implication)

*Th. 3 0. b P(Q v R) = PQ\ PR (Distribution of ‘ • ’ over V

Th. 30. Cor. b (Pv 0)R = PR v OR

Th. 31. b (P=Q)= [P£> v ^P—Q] (Material Equivalence)

T h . 3 2. b P v QR = (Pv())(Pvfl) (Distribution of ‘v’ over ‘ • ’)

With the establishment of this final group of theorems, R.S. has been shown
to contain all of the logical principles appealed to in validating extended
arguments in Chapter 3. Containing also, as it does, the Deduction Theorem
and the principle of Double Negation, it is adequate also for the methods of
Conditional Proof and Indirect Proof that were discussed in Chapter 3. It still
remains to be proved that the system is deductively complete, which will be
established in the next section.

8.6 Deductive Completeness

In Theorems 22, 23, and 30 we already have the Association of V and *•’,
and the Distribution of *■’ with respect to V. But as stated, these properties
have been established only for cases involving exactly three wffs. In proving
the deductive completeness of R.S., it is convenient to make use of more
general Association and Distribution principles. These will be established as
our next three Meta theorems. The-first of them will establish the general
Association and Commutation of the conjunction symbol, stating that no
Sec. 8.6] 251
matter in what order or grouping any number of wffs are conjoined, the
resulting wff will be equivalent to the result of conjoining them in any other
order or grouping. We may state this formally as follows:

M etathe orEM V . Let Pv P2, Pn he any wffs and let Q and R be

any wffs constructed out of them by means of'-'. If each Pi {1 < i < n) occurs
exactly once in each of the wffs Q and R, then h () = R.

Proof: We use strong induction on the number of 'factors* (i.e., conjuncts) Pi

in Q and R.
*

(«) Where n = 1, Q and R are identically the same wff Pv so b (J == R by

Th. 7 and DR 14.

(/?) Here we assume the Metatheorem to be true for every k < n factors 1\,
P2,. •., Pfc. Now consider and R, each constructed out of n > 1 factors 1\, P2,
..., Pn. Q is S-T and R is X• Y.
Each of the wffs S and T contains at least one of the factors Pv P2, . . ., Pn.
We can assume that Px is a factor of S, because if not we can apply Th. 21 and
relabel to obtain h Q = S • T where S now does contain l\ as a factor.
Because T contains at least one of P2, ..., Pn as a factor, S contains < n
of the factors Pv P2, ..., Pn. Hence either S is I\ and b = I\ • 1\ or by the
j8-case assumption b S == where S' is a wff that contains all the factors of
S except Pv In the latter case, by MT IV, Cor. we have

b Q = (PrS')-T

and by Th. 23 and MT IV, Cor.

b Q=I\-(S'-T)

In either case, there is a wff, call it T\ such that

b Q = i\-r

By the same reasoning, we can show that there is a wff, call it Y\ such that

b R = Px-Yf

Each of the wffs T and T contains the n - 1 factors P2, I\,..., Pn, so by the
/?-case assumption

bT'= Y'

By Th. 7 and DR 14 we have b Px = Pv so by DR 16 we have

252 A Propositional Calculus [Ch.8

b Pl-Tf = P1-Y'

and by MT IV, Cor.

\-Q = R

Metatheorem V now follows by strong induction.

The next principle concerns the general Association and Commutation of the
disjunction symbol V*. No matter in what order or grouping any tvffs are
connected by V’, the resulting disjunction or ‘logical sum* will be equivalent to
the result of connecting them by V in any other order or grouping. We may
state this formally as

Meta theorem VI. Let Pv P2, ... 9 Pn be any tvffs and let Q and R be
any two wffs constructed out of them by means of V. If each P- (1 < i < n)
occurs exactly once in each of the wffs Q and R, then b Q = R.

The proof will be left as an exercise for the reader.

MT VI, Cor. If Q and R are as in Metatheorem VI, then Q b R. The

proof of the corollary is obvious.

Finally, we wish to establish a generalized statement of the Distribution of

conjunction with respect to disjunction, that is, of *•’ over ‘v’. This is ex¬
pressed as

Metatheorem VII. If Q is the logical sum of Pv p*. .., Pn, that is

(P{ v P2 v ... v Pn) with association to the left given as a convention, and S is the
logical sum of PtR, P2R, ..., PnR, then b QR = S.

Proof: We use weak induction on the number of ‘summands’ (i.e. disjuncts)

^2* • • •» Pn’

(a) Where n = 1, Q is Pv QR is f\R, and S is PXR also. By Th. 7 and DR 14,

b PXR = PXR which is h QR = S.

(fi) Assume the Meta theorem true for k summands Pv p»... , Pk. Now let Q
be the logical sum or disjunction of Pv P2y..., Pk> Pk+1, and let S be the logical
sum of P^R, P2R9 .... PkR, Pk+iR- Now we argue that

b (Pn v P2 v ... v Pk)R = P.R v P2R v ... v PkR by the j3-case assumption
b Pk+1R = Pk+1R Th. 7 and DR 14
b (Pj v P2 v ... v Pk)R v Pk+1R
(PjR v P2R v ... v P^R) v Pk+1R DR 16, Cor.
Sec. S.6] Deductive Completeness 253
h[(P1vP2v...vPJvPt+1]R =
(PlVP2v...vPfc)flvPk+1R Th. 30, Cor.
h[(P1vP2v...vPk)vPm]R =
(Pjfi v P2R v ... v PkR) v Pk+lR MT IV, Cor.

By our convention of association to the left, the preceding line can also be
written as

h(P1vP2v...vPtv Pk+l)R EP1RvP2Rv...v PkR v Pk+,R

which is h QR = S where Q contains k -f 1 summands. Metatheorem VII now

follows by weak induction.
To prove that R.S. is deductively complete, we show that all tautologies are
demonstrable as theorems in the system. Since all tautologies are expressible as
wffs of R.S. (by MT I), the deductive completeness of H.S. is expressed as: If S
is a wff that is a tautology, then b S. A criterion for deciding whether or not any
wff is a tautology is supplied by the method of truth tables. Any wff S has a
truth table with as many initial columns in it as there are distinct propositional
symbols in S. Where there are n of them, say Pv P2, . . . , Pw, the tautology S
will have this truth table:

Any such truth table has 2n rows, each of which represents a different assign¬
ment of T’s and F’s to the Pt*s, one row for every possible assignment. That
only T’s appear in the column under S indicates that every possible assignment
of T’s and F’s to the P/s must assign a T to S.
To show that b S for every such S, we establish the following:

1. Each row of its truth table, that is, each assignment of truth values to the
P.’s, can be represented by a wff of R.S., the first row by Qv the second
by Q2, ...» and the last or 2nth by Q2n‘>
2. If Qv Q2i..., Q2n are the 2n wffs representing all possible assignments
of T « and F’s to the P/s, then b (Q1 v Q2 v ... v ()r); and
3. If a particular assignment of T’s and F’s to the P/s assigns a T to S, then
where Q represents that particular assignment, we have b D S.
254 A ProposNIonal Calculus [Ch.8

That these will suffice to prove that b S is easily seen. Where the truth table
for S has all T’s in the column under S, then .1- PjDS, b Q2 3 S, ...,
b Q2n D S. From these, by 2n — 1 uses of DR 10> we have b v Q2 v ...
v Q2n) D S. And once we have established that b (Qj v()2v...v @2")> we
obtain b S by R 1.
Now we proceed to attack the problem in detail. First, we must show that
each possible assignment of T’s and F\s to the P^s of a set (Px, P2,..., Pn) can be
represented by a wff. It is convenient to lay down the following:

Definition. A wff is said to represent a particular assignment of truth values

to propositional symbols Pv P2,..., Pn if and only if (on the normal interpre¬
tation of our operator symbols) that truth value assignment is the only one
which makes that wff true.

Where T’s are assigned to every P{, this assignment is represented by the
conjunction Px *P2 • ... *Pn, which we denote by *QX\ Where T’s are assigned to
every Pi except Pn, which is assigned an F, the assignment is represented by the
conjunction PxmP2m • • • * Pn-1' ^Pn, which we denote by ‘Q2 , Similarly, we use
this procedure for every other possible assignment corresponding to every
row of the truth table, ending with the conjunction —Pt • ~~P2 • ... -^Pn,
which we denote by *Q2n’. In this way any row of any truth table can be
represented by a wff of R.S., which establishes the first result mentioned in the
preceding paragraph.
Next, we turn our attention to the second result, expressed as

Metatheorem VIII. IfQv Q2, • • • > CV represent all possible distinct

assignments of truth values to the n distinct propositional symbols Pv P2, ...,
Pn, then b ((?! v g2 v ... v Qr).

Proof: We use weak induction on the number of Pfs.

(a) Where n = 1, 2W = 2, and is Px and Q2 is ~~PV Here we have

b Px v — Pj by Th. 7, Cor. which is b (Qx v £)2).

(/?) Assume that the Metatheorem is true for Px, P2, ..., Pk. Now consider
the set PXi ^2» * * * Jk’^k + V Where ^)1, Q2,..., Q# represent all possible distinct
assignments of truth values to P\> P%> • • • > we have b (Qx vp2v...v Q2k) by
the /?-case assumption.
Then we continue the argument as follows:

l-^+iV~Pk+1 Th. 7, Cor.

1- (Qi V g2 V ... V Q2*)(Pk+! V ~Pk+,), DR 14

•- (Pi V Q2 V • • • v @2‘ )(Pk+l V ~?k+1) =

((?i v Qt v • • •v y «?i V Q2 V • • • V Q2*)~Pk+1 Th. 30
'-(Qiv<?2v -v <?2*)Pk+1 V ((?!V g2v ... v C*2»)~Pk+i MT IV, Cor.
Sac. 8.6]
Deductive Completeness 255
>-(CivC>2 v---v^2‘)4+i=(g1i,t+1v^+1v, MT VII
•v(VUi)
V • • • V ft.) Pk+l —
(Cl~Pfc+lv (V~J*+1V... vp2*-^+|) MT VII
MPA+1'' (?2^c+l v • •VP2*^ + I)V(()I~?1. + ,V
Q*~~Pk+ lV--‘ V C2‘~flt+|) MT IV, Cor.
i-PA+iv Cl~^fc+lV QzPk + l V Q>~~Pk + 1 V ... V
<?2‘A + 1 v V2* ^t+1 MT VI, Cor.

The preceding expression contains 2k + 2* = 2-2fc = 2*+1 distinct summands,

each of which represents a different assignment of truth values to Pv P2,
Pk+v Each QA+1 and each is a different where the 2*^ ^;’s
represent all possible distinct assignments of truth values to l\, P2.PfcM.
Hence b (@i v Q'2 v ... v i). Metatheorem VIII now follows by weak
induction.

We next prove that if the truth value assignment represented by assigns a

T to S, then b Qj D S. We shall prove this by establishing a slightly more
general result, which includes also the case in which the truth value assignment
assigns an F to S instead. This is stated and established as our next meta¬
theorem.

Metatheorem IX. Ixt Qj represent any possible assignment of truth

values to the n propositional symbols Pl? P2, . .., Pn; and let S be any wjf which
has no propositional components other than I] (I < i < n). If the truth value
assignment represented by Qj assigns a T to S, then h D S; and if the truth
value assignment represented by ^ assigns an F to S, then b (^)- D ~~S.

Proof: We use strong induction on the number of symbols in S, counting

each occurrence of •, of —, and of any Pt as a single symbol.

(a) Where n = 1, S must be a single symbol, and since it is a wjf, it must be

a Pi (I < i < n).

case 1. Qi assigns a T to S, that is, to 1\. Hence Pi rather than ^l] must be
a factor of Qjm By MT V, b Qj = P, where K is a conjunction of all the factors
of Qj except Pt. Now we argue

b P{R D Pi Ax. 2
b Qj D ^ MT IV, Cor.

which is b Qj D S.

case2. Qj assigns an F to S, that is, to P-. Hence —Pi must be a factor of

q By MT V, b Qi = ~~PjR, where R is a conjunction of all the factors of Qj
256 A Propositional Calculus [Ch. 8

except —Pj. Now we argue

I-D -P{ Ax. 2

b Qi D -Pi MT IV, Cor.

which is b Qj D —S.

(ft) Assume that the Metatheorem is true for any S containing any number
k < n symbols. Now consider any S containing n ( >1) symbols. The wff S is
either S1*S2 or —S3.

case 1. S is Sj-Sjj.
Subcase A: Qj assigns a T to S. Here Qj must assign a T to St and a T to
S2. Since Sv S2 contain fewer than n symbols each, we argue as follows:

b Qi D by the ft-case assumption

b Qi D S2 by the ft-case assumption
b Qi D S1*S2 DR 8

which is b Qi D S.
Subcase B: Qf assigns an F to S. Here Qj must assign an F to S1 or an F to
S2. If to Sx, then b Qj D -Sx by the ft-case assumption, and hence by DR 12
b Qj D —(Si -S2), which is b D — S. If to S2 then b Qi D —S2 by the /1-case
assumption, and hence by DR 13 b Q} D —-(S1 *S2), which is b Qj D —S.

case 2. S is — S-j.
Subcase A: Qj assigns a T to S. Here Qj must assign an F to S3. Hence by
the /1-case assumption, b D ~~S3 which is b Qj D S.
Subcase B: Qi assigns an F to S. Here Qj must assign a T to S3. Hence by
the /1-case assumption, b Qj D S3. But b S3 D-S3 by Th. 4, so by DR 6 we
have b ^ D --S3, which is b Qj D —S.

Metatheorem IX now follows by strong induction.

JB

The deductive completeness of the system now follows easily, and may be
proved as

Metatheorem X. R.S. is deductively complete (that is, if S is a

tautology, then b S).

Proof: If S is a tautology, then every possible assignment of truth values to

its components Pv P2, ..., Pn must assign a T to S. Hence by MT IX:

b JQX ? S
b Q-2 D S

b Q2n DS
Sec- 8a6l Deductive Completeness 257

where Q2,..., Q2„ represent all possible assignments of truth values to Px%
P2, . .., Pn. Now by 2n - 1 uses of DR 10, we have

h v <?2 v .. v C>2») D S

and by MT VIII,

h (Qi v v • • • v y2n)

From these, we derive h S by R 1, which completes the proof of Meta¬

theorem X.

The decision problem for any deductive system is the problem of stating an
effective criterion for deciding whether or not any statement or well formed
formula is a theorem of the system. In view of the analvticity and deductive
completeness of R.S. (Metatheorems II and X), the method of truth tables
constitutes a solution to the decision problem. Truth tables enable us to decide
effectively whether or not any wff is a tautology. By MT II, only tautologies
are theorems, and by MT X, all tautologies are theorems. Hence truth tables
enable us to decide effectively whether or not any wff is a theorem. Moreover,
the proofs up to and including that for MT X do not merely assure us that for
any tautologous wff there exists a demonstration—they prescribe effectively a
method of actually constructing its demonstration. The demonstration con¬
structed by following the directions contained in the proof of deductive
completeness will usually be longer than one discovered through the exercise of
ingenuity and inventiveness. That is to be admitted. But it is significant and
important that through the use of the directions contained in the proofs up to
and including that of MT X, a demonstration within the logistic system can be
written out for any tautology—without any need for ingenuity or inventive¬
ness. Our effective solution of the decision problem for the system guarantees
that this can be done.
It is clear from the foregoing that any argument whose validity can be
established by the use of truth tables can be proved valid in R.S. In Chapter 3,
the claim was made that any such argument could be proved valid by using the
list of nineteen Rules of Inference augmented by the principles of Conditional
Proof and Indirect Proof. We are now in a position to substantiate that claim,
which is equivalent to the assertion that the method of deduction set forth in
Chapter 3 is deductively complete. We can do so by showing that every
argument which can be proved valid in R.S. can also be proved valid by the
methods of Chapter 3.
Consider any argument Pv ..., Pn Q that can be proved valid in R.S. To
say that it can be proved valid in R.S. is to say that there is a demonstration in
R.S. for the derived rule of inference Pv .. ., Pn b Q. By n uses of the Deduction
Theorem, Exportation, and the Rule of Replacement, we have h P D Q where
P is a conjunction of Pv ..., Pn. By the analyticity of R.S., P D Q is a truth
table tautology, so P- is a contradiction. Now there is a formal proof of
 [Ch. 8
validity for the argument

Pl.Pn> ~Q ' P-Q.. (1)

using the methods of Chapter 3 (by iterated uses of the principle of Conjunc¬
tion). Hence there is a formal proof using the methods of Chapter 3 for the
argument.

.\N (2)

where N is a disjunctive normal form9 of the formula Pm ~~Q. This is so because

the equivalences included among the Rules of Inference in Chapter 3 are
sufficient to permit deriving the disjunctive normal form of any line in a formal
proof.
Because P* is a contradiction, N is a disjunction in which every disjunct
contains a contradiction as a conjunct. Hence by repeated uses of variants of
the formal proof of validity for

qv[(p-~~p)-r] q

the formal proof of validity for (2) can be extended to a formal proof of validity
for

Pv-*rn>~Q •••*! (3)

where Nx is a single disjunct of N. If Nt is not itself an explicit contradiction, it

is a conjunction containing a contradiction as a conjunct. Hence by Com. and
Simp, (and possibly Assoc.) the formal proof of validity for (3) can be extended
to provide a derivation of an explicit contradiction from the set of premisses Pv
..., Pn, And this derivation constitutes an Indirect Proof of validity for
the original argument Pv... 9Pn . Q. Hence any argument that can be proved
valid in R.S. can be proved valid by the methods of Chapter 3. This proof
suffices to show that the method of deduction set forth in Chapter 3 is also a
deductively complete system of logic.10

9 See Appendix B, pages 337-343.

10The preceding proof is adapted from John Thomas Canty’s ‘Completeness of Copi’s Method
of Deduction,’ Notre Dame Journal of Formal Logic„ vol. 4 (1963), pp. 142-144. See also M. C.
Bradley, ‘Copi’s Method of Deduction Again,’ ibid., vol. 12 (1971), pp. 454-458.
Alternative Systems
and
9.1 Alternative Systems of Logic

There are three different senses in which the phrase ‘alternative systems of
logic* can he understood. These parallel the three senses of the phrase ‘alter¬
native systems of geometry*, and can most easily be explained by analogy with
the latter. We may speak of Euclidean plane geometry and Euclidean solid
geometry as ‘alternative systems* in the sense that the first can be studied
independently of the second, and they are certainly different in that the second
is more inclusive than the first. Analogously, we may speak of a Propositional
Calculus and a Function Calculus as ‘alternative systems of logic* in that the
first can be studied independently of the second, and that the second is more
inclusive than the first—where a Function Calculus contains all the tautologies
and rules of the Propositional Calculus plus Quantification Axioms, Rules, and
Theorems. It is not this sense of an alternative system with which we shall be
concerned in the present chapter.
A second sense is that in which we can speak of Euclidean geometry and
Riemannian (or Lobachevskian) geometry as alternative systems. These are
alternative in the sense that, although they may possess some theorems in
common, each contains some theorems not included in the other. Parallel to
this situation in geometry, there are alternative systems of logic exhibiting the
same sort of differences. An ordinary ‘two-valued* system of logic, whose
formulas—on interpretation—-are either true or false, can be contrasted with
‘three-valued* or ‘many-valued* systems of logic whose formulas are supposed
to take—on interpretation—either three or n > 3 different ‘truth values’.
Alternative systems of logic in this sense have been extensively developed, first
by Jan Lukasiewicz in Poland and independently by E. L. Post in this country,
more recently by J. B. Rosser and A. R. Turquette.1 A study of these is beyond

iSee ]. Lukasiewicz, O logice trojwartosciowej/, Ruch Filnzoficzny (Lwow), vol. 5 (1920), pp.
169-171.'
E. U-Post, ‘Introduction to a General Theory of Elementary Propositions’, American Journal of
Mathematics, vol. 43 (1921), pp. 163-185.
J. B. Rosser, ‘On the Many-Valued Logics’, American Journal of Physics, vol. 9 (1941), pp.
207-212.
]. B. Rosser and A. R. Turquette, ‘Axiom Schemes for M-Valued Propositional Calculi', Journal

259
260 Alternative Systems and Notations [Ch. 9

the scope of this book. It is not this sense of alternative systems with which we
shall be concerned in the present chapter.
The third sense in which one can speak of alternative systems of geometry is
that in which different axiomatic bases are assumed, but identically the same
theorems are derivable from each of them. Thus many different axiom sets for
Euclidean geometry have been devised, all of which yield the same theorems.
In alternative systems of this sort, different terms may be taken as primitive or
undefined, and different formulas are assumed as axioms or postulates. What
may be a primitive term in one system may be defined by means of other
primitives in the other, and what is assumed as an axiom in one may be derived
as a theorem from the axioms of the other—where those axioms correspond to
theorems of the first. It is this sense of alternative systems that will be discussed
in the present chapter.
The logical truths whose systematization is under consideration are truth
functional tautologies. Any system wholly adequate to their expression and
development must be functionally complete, analytic, and deductively com¬
plete, in the senses in which R.S. was proved to possess these properties in
Metatheorems I, II, and X of the preceding chapter. Any such system will be
called a Model System of Logic, and any other axiom system for logic will be a
genuine or acceptable alternative to R.S. if and only if it is a Model System.
There are many different Model Systems, either different in that they start with
different primitive or undefined terms, or different in that they assume different
formulas as axioms. They are all equivalent, however, first in being able to
express^—on their normal interpretations—all truth functions, second in in¬
cluding all tautologies as theorems, and third in having only tautologies as
theorems. One such alternative system will be proved to be a Model System in
the following section.

9.2 The Hilbert-Ackermann System

The Hilbert-Ackermann system for the propositional calculus was ac¬
knowledged by D. Hilbert and W. Ackermann to be “due in essence to
Whitehead and Russell (Principia Mathematical 1st ed.)”. It has as its primitive
symbols infinitely many single letters'With and without subscripts:

A At Ao Aq ...
B Bx B2 B3 ...
C Cx C2 C2 ...
D Dx D2 D3 ...

of Symbolic Logic, vol. 10 (1945), pp. 61-82, and Many-Valued Logics, North-Holland Publishing
Company, Amsterdam, 1952.
See also: Alan Ross Anderson et al., ‘Proceedings of a Colloquium on Modal and Many-Valued
Logics’, Acta Philosophica Fennica, Fasc. 16 (1963); and A. A. Zinov’ev, Philosophical Problems of
Many-Valued Logic, G. Kung and D. D. Corney, eds. and trs., D. Reidel Publishing Company,
Dordrecht-Holland, 1963.
Sec. 9.2] The Hilbert-Ackennenii System 261

which on their normal interpretations express noncompound propositions, and

has in addition to parentheses, the two operator symbols and V (desig¬
nated in our metalanguage by W and V), which have as their normal
interpretations the operations of negation and weak (or inclusive) disjunction.
We continue to use the symbols T\ *Q\ *R\ \S\ .... with and without
subscripts, in our metalanguage, to denote well formed formulas of H.A. (the
Hilbert-Ackermann system of logic). The notion of a wff of H.A. is defined
recursively as follows:

RECURSIVE RULE FOR wffs IN H.A.

1. Any single letter of H.A. is a wff.

2. If P is a wff‘ then ~~(P) is a wff.
3. If P and p are wffs, then (P) v (p) is a wff.
(No formula of H.A. will be regarded as being a wff unless its being so
follows from this definition.)

The symbols ‘ D \ ‘=\ are defined syntactically for our metalanguage by

the following:

P D () = df Vp
P-p = df H^Pv
Pp = dfP-p
P=P = df(PDp)(pDP)

We shall continue to use the same conventions regarding parentheses that were
adopted in the preceding chapter.
Four (patterns of) axioms or postulates are assumed in H.A.

P 1. (PvP) D P
V 2 .PD (PvQ)
V 3. (PvQ) D (QvP)
p 4. (PDQ) 3 [(flvP) D(RvQ))

Each of these syntactical expressions denotes infinitely many u# of our object

language H.A.. just as in the metalogical development of R.S
Finally, a single rule of inference is assumed, winch we may state as

R' 1. From P and P D Q to infer Q.

It should be realized that R' 1 is different from R 1. because R' 1 legitimizes

arguments within H.A. of the form
 [Ch. 9

~PvQ
Q
whereas R 1 legitimizes arguments within R.S. of the form

P
~(P~Q)
Q
and these are clearly different. We may make the contrast more vivid by
writing them as

R 1. From P and ' ~(P~~ ()) to infer Q.

R' 1. From P and ^Pv Q to infer Q.

By MTIV, Cor., and Th. 29 of R.S., any wffin R.S. that follows by R 1 from two
other wffs must also follow by R' 1 and conversely. But this statement cannot
be assumed true of H.A. until it is proved.
A ‘demonstration in H.A.’ of the validity of an argument having premisses
Pv P2.Pn and conclusion () is defined to be a sequence of wffs Sv S2, ...,
St (of H.A.), each of which is either a postulate P 1, P 2, P 3, or P 4 or a
Pi (1 < i < n) or follows from two preceding S’s by R' 1, and such that St is
Q. That there is such a demonstration in H.A. is written

pvp2,...,pn ke

Similarly, that the formula P is a theorem of H.A. is written

|fr:P

which asserts that there is a sequence of wffs Sv S2,..., St (of H.A.) each of
which is either a postulate P 1, P 2, P 3, or P 4 or follows from two preceding
S’s by R' 1, and such that St is P.
The functional completeness of H.A. is easily established. (It was Exercise 1
on page 226 in the preceding chapter.) A proof of the analyticity of H.A. is
easily given by using truth tables to show that any postulate P 1, P 2, P 3, P 4
is a tautology, and then proving that any wff which follows from tautologies by
repeated applications of R' 1 must be tautologous also.
The independence of the H.A. postulates is established by the following
models.
To prove that Postulate 1 is independent, we use the three-element model
Sec. 9.2] The Hilbert-Ackermann System 263
{0,1,2}, of which 0 is the designated element. Wffs are assigned values in
accordance with the tables:

P ~P P _Q_ PvQ PDQ

0 2 0 0 0 0
1 1 0 1 0 1
2 0 0 2 0 2
1 0 0 0
1 1 0 0
1 2 1 1
2 0 0 0'
2 1 1 0
2 2 2 0

To prove Postulate 2 independent, we use the three-element model (0, 1,2}

of which 0 is designated, with the tables:

p ~~P P 9 PvQ PDQ

0 1 ~0~ 0 0 0
1 0 0 1 0 1
2 2 0 2 0 1
1 0 0 0
1 1 1 0
1 2 1 0
2 0 0 0
2 1 1 1
2 2 1 1

To prove Postulate 3 independent, we use {0,1, 2} with 0 designated and

tables:

P _Q_ PvQ PDQ

2 0 0 0 0
0
1 0 0 1 0 2
2 1 0 2 0 2
1 0 0 0
1 1 1 0
1 2 0 0
2 0 0 0
2 1 2 1
2 2 2 0
264 Alternative Systems and Notations [Ch. 9

To prove Postulate 4 independent, we use the four-element model

{0,1, 2,3} with 0 designated and tables: . .

P ~~P P Q PvQ P D Q
o r~ 0 0 0 0~
1 0 0 10 1
2 3 0 2 0 2
3 0 0 3 0 3
10 0 0
111 0
12 2 0
13 3 0
2 0 0 0
2 12 3
2 2 2 0
2 3 0 3
3 0 0 0
3 13 0
3 2 0 0
3 3 3 0

The deductive completeness of R.S. was proved in Chapter 7 by showing

that all tautologies are derivable by its rule R 1 from its three axioms. One
might think that, to prove the deductive completeness of II.A., it would suffice
to derive the three axioms of R.S. as theorems of H. A., and the rule R 1 of R.S.
as a derived rule of H.A. For would that not show all tautologies to be
derivable, via those three theorems and one derived rule, from the four
postulates of H.A. by its rule R' 1? This question is not just rhetorical. In fact,
the answer to it is negative. The trouble is that the systems H.A. and R.S. have
different primitive symbols. Despite the fact that both R.S. and H.A. are
deductively complete, not all tautologies can be derived by the R.S. rule R 1
from the three R.S. axioms when these are regarded as constructed or formu¬
lated in the primitive base of H.A. rather than that ofR.S., that is, with — and
v as undefined symbols rather than — afid •. This can be shown by proving the
tautology Pv~~P to be independent of the H.A. formulations of the
three axioms of R.S.: ~~[Pm ~~(PmP)]9 ~~[(P*Q)m~~P], and
^Q) • •?)]}, in the sense of not being derivable
from them by the H.A. formulation of the R.S. Rule: P, ~~(P*^Q) |nx Q.
To prove that a specified tautology cannot be proved in a given axiom
system, one proceeds in the same way as in proving an axiom of a system to be
independent of the other axioms. One regards the specified tautology as if it
were an axiom to be proved independent, and attempts to show that it does not
follow by the given system’s rule from the (other) axioms of the system, just as
described on pages 231 ff.
To prove independence (nonderivability) of the tautology P v we use the
Sec. 9.2] The Hilbert-Ackermenn System 265
six element model {0,1,2,3,4,5}, with the tables for ~ and v (and the
derivative table for •), given below.

P-Q
P_ <? Pvp ~P ~Pv ~(~P v ~~Q)
V 0 0 ~5~ 5 5 0
0 1 0 5 5 5 0
0 2 3 5 4 5 0
0 3 3 5 1 0 5
0 4 0 5 0 0 5
0 5 0 5 0 0 5
1 0 0 5 5 5 0
1 1 0 5 5 5 0
1 2 3 5 4 5 0
1 3 3 5 1 0 5
1 4 0 5 0 0 5
1 5 0 5 0 0 5
2 0 3 4 5 5 0
2 1 3 4 5 5 0
2 2 3 4 4 5 0
2 3 3 4 1 0 5
2 4 3 4 0 0 5
2 5 3 4 0 0 5
3 0 3 1 5 0 5
3 1 3 1 5 0 5
3 2 3 1 4 0 5
3 3 3. 1 1 0 5
3 4 3 1 0 0 5
3 5 3 1 0 0 5
4 0 0 0 5 0 5
4 1 0 0 5 0 5
4 2 3 0 4 0 5
4 3 3 0 1 0 5
4 4 5 0 0 0 5
4 5 5 0 0 0 5
5 0 0 0 5 0 5
5 1 0 0 5 0 5
5 2 3 0 4 0 5
5 3 3 0 1 0 5
5 4 5 0 0 0 5
5 5 5 0 0 0 5

In this model the three elements 0, 1, 2 are designated. The characteristic of

fairing only designated values is hereditary with respect to the rule: From P and
266 Alternative Systems and Notations [Ch. 9

~~(Pm ~~Q) to infer Q: and the three H.A. formulations of the R.S. axioms take
only designated values. But for the value 2 for. P, we have P v ~~P =
2 v ^2 = 2 v 4 = 3, which is not a designated value.2'
To prove the deductive completeness of H.A., we first establish some
theorems, derived rules, and metatheorems for it.

Theorem 1. \sa(Q D R) D [(P D Q) D (P D R)]

Demonstration: 1. (Q D R) D [(~Pv Q) D (~PvR)] P 4
2. (Q D R) D [(P DQ)D(PD R)] df.

DR 1. PDQ,QDr\skPDR

Proof: 1. (QDR)D [(P D Q) D (P D R)] Th. 1

2. Q D R premiss
3. (P D Q) D (P D R) R' 1
4. P D Q premiss
5. PDR R' 1

Theorem 2. (ha P D (Qv P)

Proof: 1 ,PD(P\Q) P 2
2. (PvQ) D (QvP) P 3
3. PD (Qv P) DR 1

DR 2. Q D r\ha(PvQ) D (PvR)

Demonstration: 1. (Q D R) D [(Pv Q) D (PvR)] P 4

2. Q D R premiss
3. (PvQ) D (PvR) R' 1

Theorem 3. |nx:P D P
Proof: l.PD(PvP) P 2
2. (PvP) D P P 1
3. PDP DR 1

DR 3. PvQ IhxQvP

Demonstration: 1. (Pv Q) D (QvP) P 3

2. PvQ premiss
3. QvP R' 1

2See Henry Hiz, ‘A Warning About Translating Axioms’, American Mathematical Monthly, vol.
65 (1958), pp. 613 ff.; Thomas W. Scharle, 'Are Definitions Eliminable in Formal Systems?*
(Abstract), The Journal of Symbolic Logic, vol. 35 (1970), pp. 182 ff.; and Alonzo Church,
Introduction to Mathematical Logic, Princetop University Press, Princeton, N.J., 1956, pp.
125-128. On this matter I have benefited from correspondence with Professor ]ean Porte and from
discussion with Professor Anjan Shukla.
Sec. 9.2] The Hllbert-Ackermenn System 267
Theorem 4. IhaPv^P

Proof: 1. P D P Th. 3
2. ~PvP df.
3. Pv~P DR 3

Theorem 5. [haP D —P

Proof: 1. ~Pv-P Th. 4

2. P D --P df.

Theorem 6. |ha ^~~P d P

Proo/; 1. —P D ~~~P Th. 5

2. (Pv ~P) D (Pv --P) DR 2
3. Pv~P Th. 4
4. pv —-P R' 1 (2,3)
5. -—--PvP DR 3
6.-PDP df.

Theorem 7. |ha [P v (£) v R)] D [^) v (P v R)]

Proo/: 1. flD(PvR) Th. 2

2. (pvR) D [pv(PvR)] DR 2
3. [Pv(pvR)] D (Pv[pv(PvR)]} DR 2
4. {Pv [p v (PvR)]} D {[pv(PvR)]vP} P 3
5. [Pv(pvR)] 3 {[Qv(PvR)\vP) DR 1 (3,4)
6. P D (PvR) P 2
7. (PvR) D[Qv(PvR)] Th. 2
8. P D [p v (Pv R)] DR 1 (6, 7)
9. {[O v (P v R)] v P} =>
{[O v (Pv R)] v [p v (Pv R)]} DR 2 (8)
10. {[Ov(PvR)]v[^v(PvR)]} D
[p v (Pv R)] P 1
11. {[p v (Pv R)] v P) D [p v (P v R)] DR 1 (9,10)
12. [Pv(pvfl)] D [pv(PvR)] DR 1 (5, 11)

Theorem 8. [ha [Pv (() v R)] D [(P v p) v fi]

Proof: 1. (pvR) D (Rvp) P 3

2. [Pv(pvR)] D [Pv (R v P)] DR 2
3. [Pv (R v p)] D [flv(Pvp)] Th. 7
4. [Pv (pv R)] D [R v (Pv P)] DR 1 (2,3)
5. [Rv (Pv P)] D [(Pv p) v R] P 3
6. [Pv(pvR)] D [(PvP)vR] DR 1 (4,5)
268 Alternative Systems and Notations [Ch.9

Theorem 9. |hs:[(Fv())vR] D [Pv(^)vfi)]

Proof: 1. [(Pv0vfl] D [flv(Pv())] P 3

2. [Rv(Pv0] D [Pv(Rv£))] Th. 7
3. [(Pvg)vR] D [Pv(Rvg)] DR 1
4. [Pv(Rvg)] D [(Pvfl)vgj Th. 8
5. [(Pvg)vfl] D [(PvR)vgj DR 1 (3,4)
6. [(PvR)vg] D [gv(PvR)] P 3
7. [(Pvg)vR] D [gv(Pvfl)] DR 1 (5,6)
8. [gv(Pvfl)] D [Pv(gvR)] Th. 7
9. [(Pvg)vR] D [Pv(gvfl)] DR 1 (7,8)

Theorem 10. |haP D [Q D (Pg)]

Proof: 1. (~Pv ~g) v ~(~Pv ~g) Th. 4
2. [(~Pv~g)v~(~Pv~g)] d
{—Pv [~gv ~(~Pv ~g)]} Th. 9
3. ~Pv[~gv~(~pv~g)] R' 1
4. p D [g D (pg)] df.

DR 4. p, g liiAPg
Proof: l. p d [g d (pg)] Th. 10
2. P premiss
3. g d (pg) R' 1
4. g premiss
5. pg R' 1

The next three theorems follow by DR 4 from Theorems 5 and 6, Theorems

8 and 9, and Theorem 3, respectively:

Theorem 11. |haP = —

Theorem 12. |ha [P v (£) v itf] = [(P v Q) v R]

Theorem 13. |haP=P

DR 5. PDQ |HA~g D ~P

Proof: 1. P D Q premiss
2. Q D Th. 5
3 .PD ~~P DR 1
4. ~Pv -—~Q df.
5. -Q v ~P DR 3
6. —Q D ~~P df.
Sec. 9.2] The Hilbert-Ackermann System 269
Theorem 14. (ha(PQ) D P
Proof: 1. —P D (~Pv ~~Q) P2
2. ~(~Pv 3 ~_^p DR 5
3.-P D P Th. 6
4. ~(~Pv ~<p) D P DR 1
5. (PQ) D P df.

Theorem 15. |iIa(P()) D

Proof: 1. (~^)v ~P) D (~Pv ~<p) P 3

2. ~(~P v ~p) D ~(~<pv ~P) DR 5
3. (P<?) D (()P) df.
4. (<?P) D Th. 14
5. (PQ) D Q DR 1

The next two derived rules follow from the two preceding theorems by R' 1.

DR 6. PQ |haP

DR 7. Pq\haQ
DR 8. P=Q biA ~P= ~<p

Proof: 1. P=Q premiss

2. (P D (>)(p D P) df.
3. PDQ DR 6
4. D ~P DR 5
5. Q D P DR 7 (2)
6. ~P D ~() DR 5
7. (~P D D ~P) DR 4
8. -Pee ~<p df.

DR9. Pee(), pHEft 1itaP = R

Proof: 1. P=Q premiss

2. (P D (?)((? D P) df.
3. P D Q DR 6
4. Q D P DR 7
5. Q=R premiss
6. (p D fl)(R 3 (?) df.
7. QDR DR 6
8. RD Q DR 7
9. P D R DR 1 (3, 7)
10. R D P DR 1 (8, 4)
11. (P D R)(R D P) DR 4
12. P=R df.
270 Alternative Systems and Notations [Ch. •

Theorem 16. |haPe=:(PvP)

Proof: 1. PD(PvP) P 2
2. (PvP) D P p 1
3. [P D (Pv P)][(Pv P) DP] DR 4
4. P=(PvP) df.

Theorem 17. bx'P^PP)

Proof: 1. ~P= (~Pv ~P) Th. 16

2. ~~P = ~(~Pv ~P) DR 8
3. P= ~~P Th. 11
4. P=~(~Pv~P) DR 9
5. P=(PP) df.

DR 10. g DR bv(Pvg) D (RvP)

Proof: 1. g D fl premiss
2. (Pvg) D (PvR) DR 2
3. (PvR) D (RvP) P 3
4. (Pvg) D (R vP) DR 1

DR 11. P D g, R D S bA(Pvfl) D (gvS)

Proof: 1. fl D S premiss
2. (PvR) D (SvP) DR 10
3. pdg premiss
4. (SvP) D (gvS) DR 10
5. (PvR) D (gvS) DR 1 (2,4)

DR 12. P=g,fl = S IffA(PvR) = (gvS)

Proof- 1. P=g premiss
2. (P D gxg D P) df.
3. p d g DR 6
4. g d p DR 7
5. R = S premiss
6. (R D S)(S D R) df.
7. R D S DR 6
8. S DR DR 7
9. (PvR) D (g.v-S) DR 11 (3, 7)
10. (gvS) D (PvR) . DR 11 (4,8)
11. [(PvR) D (gvS)][(g‘vS) D (PvR)] DR 4
12. (Pvfl) = (gvS) df.
Sec. 9.2] The Hllbert-Ackermann System 271
At this point it will be useful to prove the Rule of Replacement as our first
metatheorem.

Metatheorem I. (Rule of Replacement). Let Pv P2,..., Pn be any wffs,

let Q be any wff that does not occur in any Pi, and let S be any wff which
contains no components other than Q and P{ (l <i < n). If S* is any wff which
results from replacing any number of occurrences of Q in S by R, then
Q = R |haS = S*.

Proof: Strong induction on the number of symbols in S, counting each

occurrence of v, —, Q, or any Pi as a single symbol.

(a) n = 1. Here S is either Q alone or a Pi alone.

case 1. S is Q and S* is R. It is obvious that Q=R [ha = R, which can

be written Q = R [ha S = S*.

case 2. S is Q and S* is also. Here [ha () ==£ {) (Th. 13), that is,
[ha S = S*, whence Q = R IiIa S = S*.

case 3. S is a Pr Here S* is Pr also. Here [ha l\ === Pi (Th. 13), that is,
[ha S = S*, whence () = K [ha S = S*.

(/?) Here the Metatheorem is assumed to be true for any S containing fewer
than n symbols. Now consider any S containing exactly n ( >1) symbols. S is
either —Sx or S1 v S2.

case 1. S is — Sv Here S\ contains fewer than n symbols, so by the ft-case

assumption, =R [haS\ = S*. Rut S, = S* [ha —by DR 8. It is
obvious that = (~~St)*, so is S*, whence {) = R [ha S e= S*.

case 2. S is Sj v S2. Here Sf and S2 each contains fewer than n symbols, so

by the /8-case assumption, Q=lR [ha Sj = S,* and Q=R ItIa S2 = S2*. Now
by DR 12: S^S^, S2=S2* [ha (S, v S2) = (S* v S2*), so <2=eR Iha^
v S2) = (S* v S./). Since any S* v S2* is an S*, we have Q = R (ha S = S*.

Hence by strong induction we infer that regardless of the number of symbols

in S, Q = R |ha S= S*.

M T I, Corollary: If Q, R, S, and S* are as in Metatheorem I, then

Q?=R,S IrrxS*.
The proof of this corollary is obvious.

A few more theorems and derived rules will move us closer to a proof of
deductive completeness for H.A.
272 Alternative Systems and Notations [Ch.9

Theorem 18. |ha (P v Q) = (Q v P)

Proof: 1. (PvQ) D (QvP) P 3

2. (QvP) 3 (PvQ) P 3
3. ((PvQ)3(QvP)][(QvP)D(PvQ)] DR 4
4. (PvQ) = (QvP) df.

DR 13. P D Q, P D R (haP D (QR)

Proof: 1. PDQ premiss

2. D ~P DR 5
3. PDR premiss
4. ~fl D ~P DR 5
5. (~£v~R) D (~Pv —P) DR 11
6. ~(~Pv ~P) D ~(~p v —fl) DR 5
7. (PP) D (pfl) df.
8. P D ((5R) MT I, Cor., Th. 17

Theohem 19. li!AtPv(gR)] 3 [(Pv0)(Pvfi)]

Proof: 1. p) D () Th. 14
2. [Pv(pR)] D (Pv()) DR 2
3. (^R) D fl Th. 15
4. [Pv(QR)] D (PvR) DR 2
5. [Pv«?R)J D[(Pv<?)(PvR)] DR 13

DR 14. ? D (() D fl) (haQ D (P D R)

Proof: 1. PD(QDR) premiss

2. ~Pv(~£>vR) df.
3. [~Pv(~9vfl)] D [~9v(~PvR)] Th. 7
4. ~£> v (—PvR) R' 1
5. pD(PDfl) df.

DR 15. P D (Q D R) (ha(PQ) DR

Proof: 1. PD (Q DR) premiss

2. ~Pv(~pvfl) df.
3. (—Pv —Q)v R MT I, Cor., Th. 12
4. ~~(~P v ~Q) v R MT I, Cor., Th. 11
5. (PQ)DR df.
Sec. 9.2] The Hllbert-Ackermann System 273
Theorem 20. \ha[(PvQ)(PvR)] D [Pv(QR)]

Proof: 1. Q D [R D (QR)] Th. 10

2. [R D (gfl)] D {(PvR) D [Pv(QR)]} P 4
3. Q D {(PvR) D [Pv(QR)]} DR 1
4. (PvR) D {g D [P\(QR)]} DR 14
5. (g D [Pv(gR)]} D
{(Pvg) D {Pv[Pv(gfl)])} P 4
6. (PvR) D {(Pvg) D (Pv[Pv(gR)]}} DR 1 (4,5)
7. (PvR) D {(Pvg) D [(PvP)v(gfl)]} MT I, Cor., Th. 12
8. (PvR)D {(Pvg) D [Pv(gR)]} MT J, Cor.. Th. 16
9. (Pvg) D {(PvR) D [Pv(gR)]} DR 14
10. [(PvgXPvR)] D [Pv(gR)] DR 15

Theorem 21. (ha [Pv (gR)] = [(Pv g)(Pv R)]

Proof: Th. 21 follows from Th. 19 and Th. 20 by DR 4.

Theorem 22. ha = (~~Pv

Proof: 1. ~(Pg)=~(^P) Th. 13

2. ~(pg)= -—-(~pv ~g) df.
3. ~(pg) = (~Pv ~g) MT I, Cor., Th. 11

Theorem 23. [ha *~(P v Q) =

Proof: 1. ~(Pv g) = ~(Pv Q) Th. 13

2. ~(pvg) = ~(—~Pv~~g) MT I, Cor., Th. 11
3. ~(-Pvg) = (~p~g) df.

DR 16. P (ha Pvg

Demonstration: 1. P D (Pv Q) P 2
2. P premiss
3. Pvg R' 1

At this point it is useful to state and prove

Metatheorem II (Generalized Association and Commutation of v).

Let Pv P2,..., Pn be any wjfs and let Q and R be any two wffs constructed out
of them by means ofv. If each Pi (1 < i < n) occurs exactly once in each of the
wffs Q and R then |ha Q=R.

Proof: Strong induction on the number of disjuncts Px in O and in R.

(a) n = 1. Here Q and R are identically the same wff Pv so wkQ = R by
Th. 13.
 [Ch. 9
(fi) Here the Metatheorem is assumed to be true for any k < n disjuncts Px, P2,
..., Pk. Now consider Q and R each constructed .out of exactly n (>1)
disjuncts Pv P2, Pn. Q is S v T and R is X v Y. • # '
Each of the tvffs S and T contains at least one of the wffs P{ (1 < i < n). We
can assume that Px is a disjunct of S, because if not, we can use Th. 18 and MT
I, Cor. to obtain |HE Q = (S v T), where S now does contain Px as a disjunct.
Because T contains at least one of P2,P3,...,Pn as a disjunct, S contains
fewer than n of the disjuncts Pi. Hence either S is Px and (ha Q = (Px v T), or
by the /?-case assumption |ha S = (Px v S'), where S' is a u>j(f that contains all
the disjuncts of S except Pv In the latter case, by MT I, Cor. we have

\iiZQ = [(PivS')vT]

and by Th. 12 and MT I, Cor.

k? = |PlV(S'vr)]

In either case there is a tvff, call it T', such that

taf Q = (Pj V T')

By the same reasoning we can show that there is a wff\ call it Y', such that

lifAK = (P1vy')

Each of the wffs T' and Y' contains the n — 1 disjuncts P2, ^3> • • •»Pn , so by the
j8-case assumption

By Th. 13 we have |ha Px = Pv so by DR 12 we have

(ha (P1 v T') = (P, v Y')

jm

which by MT I, Cor. gives

Metatheorem II now follows by strong induction.

To prove the deductive completeness of H.A., we use a somewhat different

method than that used in Chapter 7 to prove the deductive completeness of
R.S. Here we make use of the notion of a Conjunctive Normal Form, as
discussed in Appendix B. Our representation of a wff is in Conjunctive Normal
Form (abbreviated as C.N.F.) if and only if: (1) it contains only propositional
symbols, parentheses, and the symbols v, and *; (2) the negation signs apply
Sec. 9.2] The Nllbert-Ackermann System 275

only to propositional symbols; and (3) no disjunct is a conjunction, that is, the
symbol v is nowhere adjacent to a conjunction. In symbols, the ioff S is in
C.N.F. if and only if S is

(associated in any way at all), where each Si is

(associated in any way at all), where each T{ is either a propositional symbol or

the negation of a propositional symbol.
Next we state and prove

Metatheorem III. Given any representation of a wff S there is a

Conjunctive Normal Form formula SCNF such that filA S EE SCNF.
Proof: If S is itself in C.N.F., then [iTa S = SCNF by Th. 13. If S fails to be
in C.N.F., it can only be through failing to satisfy conditions 1, 2, or 3 in the
preceding paragraph. If S fails to satisfy condition l, it contains defined
symbols E or D. Here we replace every well formed part of S of the form
= S2 by (—SL v S2)(-S2 v Sj) and every well formed part of S of the form
SA D S2 by 'S1 v S2. The result of these replacements is S', where (ha S = S'
(by df.) and S' satisfies condition 1.
If S (or S') fails to satisfy condition 2, we replace every w.f. part of the form
—(S1S2) by —St v —S2 and every w.f. part of the form ^(SI v S2) by —S^S2.
After all such replacements have been made, we replace every w.f. part of the
form ^-"-*'Sl by Sv The result of these replacements is S", where flTA S' = S"
and hence \nX S = S", by MT I and Theorems 22, 23, and 11, and S" satisfies
conditions 1 and 2.
If S (or S' or S") fails to satisfy condition 3, it can only be because it contains
w.f. parts of either the form Sx v (S2S3) or the form (S2S3) v Sr Now we replace
each w.f. part. Sa v (S2S3) by (S1 v S2)(Sl v S3), and each w.f. part (S2S3) v S1 first
by S1 v (S2S3) and then by (S1 v SjXS, v S3). The result of these replacements
satisfies all the conditions 1, 2, and 3 and is therefore an SCNF. Moreover,
|hx S = SCNF by MT I and Theorems 18 and 19.
Hence for any wff S there is a wfj SCNF in C.N.F. such that (lTA‘ S == SCNF.
All H.A. equivalences are truth preserving, so if S is a tautology, then any
SCNF such that |ha S = SCNF is a tautology also.
It is clear that if a wff S is a disjunction of wffs Tv 7 2,..., Tn where each 1)
is either a propositional symbol or the negation of a propositional symbol, then
if S is a tautology there must be a propositional symbol P such that both P and
are disjuncts in S. For if not, then suitable truth value assignments (false to
each T) which is a propositional symbol and true to each propositional symbol
whose negation is a Tfj will make every disjunct of S false and hence S itself
276 Alternative Systems and Notations [Ch.9

false. This is contrary to the assumption that S is a tautology and therefore true
for any assignment of truth values to its component propositional symbols.
With these remarks in mind, we can state and prove

Metatheorem IV. If a wff in C.N.F. SCNF is a tautology, then

fiTA SCNF.

Proof: Let SCNF be a wff in C.N.F., which is a tautology. Then SCNF must be
a conjunction Sj S2... Ss each of whose conjunets Si (1 < i < s) is a disjunction
of wffs Tv T2, ..., Tn. each of which is either a propositional symbol or the
negation of a propositional symbol. Since SCNF is a tautology, each of its
conjunets Sj is tautologous, because a conjunction is true if and only if all of its
conjunets are true. We have already remarked that a disjunction of proposi¬
tional symbols and negations of propositional symbols can be a tautology only
if there is a propositional symbol P such that both P and —P are disjuncts of Si.
But (ns: P v ~~P by Theorem 4, from which we derive |fia (P v — P) v () by DR
16, where in this case () is a disjunction of all of the disjuncts of S,j other than P
and — P. But by MT II, fnA Si = \(P v —P) v (_)], whence [iTA Si by MT I, Cor.
Hence every conjunct Si of SCNF is a theorem, that is: [ha S\, |itaS2, ...,
[iTA S8. Now by s — 1 uses of DR 4 we obtain [ha Sl S2 . .. Ss which is [tta Scnf.
This completes our proof of Metatheorem IV.

Now the deductive completeness of H.A. follows.

M e tat ii e o re m V. The system Il.A. is deductively complete.

Proof: By the functional completeness of II.A., any tautology is expressible
in H.A., say by the wff S. By MT III there is a C.N.F. formula SCNF such that
[da S EE SCNF. The wff SCNF is also tautologous, whence by MT IV, [ha Scnf,
and by MT I, Cor., fnA S. So H.A. is deductively complete.

EXERCISES

Prove that each of the propositional calculi described in the exercises on page 237 is
a model system of logic.

9.3 The Use of Dots as Brackets

It has been remarked that the language of symlxilic logic requires punctua¬
tion if ambiguity is to be avoided. It shares this characteristic with natural
languages as well as with other artificial languages like (ordinary) algebra. We
have been using three kinds of punctuation marks in our logical language:
parentheses, brackets, and braces. In the following discussion, for convenience,
we shall use the word ‘brackets’ Jto refer to all of these.
Even moderately complicated formulas require many pairs of brackets,
which make them difficult to read. Any use of paired punctuation marks
Sec. 9.3] The Use of Dots as Brackets 277
involves redundancy. In the formula

(PDQ)v(PD

some punctuation is necessary to avoid ambiguity. Brackets are essential, but

not paired brackets, for the outermost parentheses can be dropped without
incurring any ambiguity, leaving

PDQ)v(PD ^Q

The same punctuation effect can be obtained by replacing the remaining

parentheses by dots, which would give

P D Q. v .P D

There is no danger of confusing the punctuation dot with the conjunction dot
because the punctuation dot can occur only adjacently to a connective symbol
such as V’, ‘D', or *=', whereas the conjunction dot can never do so.
Dots are symmetrical in shape, in contrast to the asymmetry of brackets.
Thus ‘(’, ‘[’, ‘{' are all concave on the right, which indicates that they group or
‘operate toward' the right, while ‘)\ ‘]\ ‘}’ are concave on and ‘operate toward*
the left. The symmetry of the dot notation is compensated for by introducing
the convention that punctuation dots always operate away from the connec¬
tive symbol to which they are adjacent.
There is another analogy between bracketing logical formulas and punctu¬
ating sentences of natural languages. The latter have punctuation marks of
different degrees of strength, the stronger of which ‘take precedence over* or
‘extend over* the weaker. For example, a period is stronger than a semicolon,
and a semicolon is stronger than a comma. Of the three kinds of brackets used
in logical formulas, the unstated convention has been to regard braces as
stronger than brackets, and brackets as stronger than parentheses. We have
used parentheses to group symbols within brackets, but not conversely, and we
have used brackets to group symbols within braces, but not conversely.
When this convention is adhered to, it permits additional redundant brackets
to be dropped. Thus the formula

(P D Q) D [(R v P) D {R v Q)]

remains unambiguous when rewritten as

PDQ)D[RvP)D (R v Q

If we use a single dot in place of a parenthesis, and a double dot in place of a

bracket, the preceding formula can be rewritten as

P D Q. D :RvP. D .RvQ
278 Alternative Systems and Notations ‘ [Ch. 9

where the convention is that two dots bind more strongly or have greater scope
than one dot. Using three dots as a punctuation mark, that has greater scope
than either one or two dots, the formula

[Pv(QvR)] D {Pv [Qv (Pv R)]}

can be rewritten as

Pv ..Qv R : D Pv : Qv . Pv R

In writing some formulas, we were forced to include one pair of braces

within another, because we had only three kinds of brackets available. Now the
use of dots permits the generation of as many different punctuation marks of
different degrees of strength as may be desired, by the simple expedient of
adding dots one at a time. The general convention here is that the scope of a
group of n dots extends over that of any number of groups of less than n dots,
and that the scope of a group of n dots extends to but not beyond the nearest
group of n or more dots. The formula

{(Qv R) D [Q v (PvR)]} D {[Pv(QvR)] D {Pv [Qv (Pv R)]}}

can be written as

QvR. D :Qv .PvR:. D ::Pv .QvR: D :.Pv:Qv .PvR

For the sake of symmetry and greater ease of reading, one frequently adds dots
which are not strictly necessary for the avoidance of ambiguity. Thus the
formula

P D .QDP

is frequently written

P.D.QDP

The preceding lengthy formula may be more easily read when rewritten
according to this convention as

QvR: D :Q . v .Pv R :: D :: P .v .Qv R:. D P :v : Q .v . Pv R

Now what should be done in case two bracketed expressions are connected
by the conjunction symbol, as in the formula

(P D QHQ 3 P)
Sec. 9.4] A Parenthesis-Free Notation 27S
Applying the technique already described, we should emerge with the awk¬
ward expression

PDQ.-.QDP

To avoid this awkwardness, it is customary to let the conjunction dot itself do

the bracketing, writing the formula simply as

P D p.p D P

Here the single dot is to be thought of as operating in both directions, left and
right. This convention is satisfactory, as can be seen by observing that the
different formulas

1. PD[(Q.Q)DP]
2. [P D (p.p)] D P
3. [(PDQ).(J]DP
4. PD[p.(pDF)]

can be written in distinguishable and unambiguous fashion as

1'. P D -., p.p : D P

2'. PD .Q.Q-.. DP
3'. P D p.p : 3 P
4'. PD-.Q.QDP

The use of dots as brackets has the advantage of economy, and the further
advantage of providing an infinite number of different punctuation marks of
different degrees of strength, where the strength or scope of each may be
determined by the simple expedient of counting its constituent dots.3

9.4 A Parenthesis-Free Notation

A logical notation that dispenses with brackets entirely has been devised by
the Polish logician, J. Lukasiewicz, and has been extensively used by logicians
of the Polish School. Corresponding to the four most commonly used operator
symbols
~ D • v

3 A more extended discussion of this topic will be found in (,. I. Lewis and (,. II. Langford,
Symbolic Logic The Century Co., Appendix 1, New York, 1932, pp. 486-489. For more technical
discussions' the reader is referred to H. B. Curry, ‘On the Use of Dots as Brackets in Logical
Expressions’, The Journal of Symbolic lagic, vol. 2 (1937), pp. 26-28 and A. M. Turing, ‘The Use
of Dots as Brackets in Church’s System’, ibid., vol. 7 (1942), pp. J46-156.
280 Alternative Systems and Notations [Ch.9

are the four symbols

NC KA

In writing their formulas, they use lower-case letters ‘p\ *q\ V, V,... instead
of the capital letters ‘JP\ *Q\ TV, ‘S’, ... Instead of the connective symbols *C\
*K\ ‘A’ being written between the formulas they connect, they are placed
directly to the left of the two formulas to be joined. Thus

—P is written Np
P D Q is written Cpq
P'Q is written Kpq
Pv Q is written Apq

That this notation is unambiguous can be seen by comparing the following

formulas:

P D (Q D R) is written CpCqr
(P D Q) D fl is written CCpqr
P D (Q'R) is written CpKqr
(P D @)*R is written KCpqr
(Pv Q) D (RmS) is written CApqKrs

The axioms of R.S. translate into the Polish notation as

Ax. 1. CpKpp
Ax. 2. CKpqp
Ax. 3. CCpqCNKqrNKrp

And the three postulates of Lukasiewicz’s own system are written in his
notation as

P 1. CpCqp
P 2. CCpCqrCCpqCpr
P 3. CCNqNpCpq

The Polish notation has the obvious advantage of dispensing with all special
punctuation marks, for the order in which its symbols are written suffices to
make any formula unambiguous.

EXERCISES

*1. Translate the postulates of H.A. into the Polish notation.

2. Translate the postulates of F.S. iitfo the Polish notation.
3. Translate the postulates of PN into the Polish notation.
Sec. 9.5]

9.5 The Stroke and Dagger Operators

Any of the following pairs of operators provide a functionally complete

logic: and •, ~~ and v, ~ and D, or D and +. We can construct a
functionally complete system of logic containing just a single operator, and we
can do it in either of two ways.
The first way is to adopt as the single primitive operator the so-called ‘stroke
function . This operator symbol, called ‘alternative denial' by Quine,* operates
on or connects two formulas and is written ‘P | Q\ Its standard interpretation is
to deny that both of the formulas P and Q are true, which'is the same as
affirming that at least one is false. It is defined by the truth table

_P\V
T t F
TFT
F T T
F F T

The other operators, ~, •, v, and D can all be defined in terms of the stroke
function. That the following definitions preserve the standard interpretations of
the symbols being defined is easily verified by the method of truth tables:

-P = df P|P
P.Q = df j
P Q. | .P | Q
PvQ= df P|P.|.Q|Q
P D Q = df P.|.Q|Q

The other operator that suffices for a functionally complete logic is that of
‘joint denial’, symbolized by a dagger with its point down. Written ‘PJQ’, its
standard interpretation is to deny that either of the formulas P or Q is true,
which is the same as affirming that they are both false. It is defined by the truth
table

p_q_nv
T T F
T F F
F T F
F F T
tr

4 See W. V. O. Quine, Mathematical Logic, Harvard University Press, Cambridge, Mass., 1947,
pp. 48-49. The stroke function is frequently referred to as the ‘Shcffer stroke function’ after
Professor H. M. Shefler, although it was first used by C. S. Peirce.
282 Alternative Systems and Notations [Ch. 9

The other operators can all be defined in terms of, the dagger function alone.
That the following definitions preserve the standard interpretations of the
symbols that are being defined is easily verified by the. method of truth tables:

= df PIP
PQ = df PlP.l.QlQ
PvQ = di PlQ.l.PlQ
PDQ = di PlP.l.Q.l-.PlP.l.Q

Thus we see that functionally complete systems of logic based on a single

operator can be constructed in terms either of the stroke or of the dagger. It
is interesting to note the parallel between the definition of the stroke function
in terms of the dagger and the definition of the dagger function in terms of the
stroke. These are

PIQ = df P\P.\.Q\Q-.\iP\P.\.Q\Q
P K? = df PiP^Qm-nP'l QlQ

as may easily be verified.

EXERCISES

1. Express Ax. 1 of R.S. in terms of the stroke function.

*2. Express Ax. 1 of R.S. in terms of the dagger function.
3. Express P 1 of H.A. in terms of the dagger function.
4. Express Ax. 3 of R.S. in terms of the stroke function.
5. Express Ax. 1 of L.S. in terms of the dagger function.
*6. Express Ax. 2 of R.S. in terms of the stroke function.
7. Express Ax. 5 of F.S. in terms of the stroke function.
*8. Express Ax. 5 of F.S. in terms of the dagger function.
9. Express Ax. 1 of F.S. in terms of the stroke function.
10. Express P 2 of H.A. in terms of the dagger function.

9.6 The Nicod System

Thus far in the text and exercises, a number of alternative Model Systems of
Logic have been set forth. Each of them is based on two primitive operator
symbols, and the number of their axioms or postulates ranges from three (for
R.S.) to five (for F.S.). A system that is more economical both in primitive
operators and postulates is attributable to J. G. P. Nicod.5 The Nicod System
can be set forth as follows.

5 J. G. P. Nicod,4A Reduction in the Number of the Primitive Propositions of Logic', Proceed¬

ings of the Cambridge Philosophical Society, vol. 19 (1916), pp. 32-40.
See also J. Jorgensen, A Treatise of Formal Logic, vol. 2, Russell and Russell Publishers, New
York, 1962, pp. 149-172.
Sec. 9.6] The Nlcod System 283
The primitive symbols are infinitely many propositional symbols P,Q, R, S,
T, with and without subscripts, parentheses (or dots), and the single operator
symbol | . The recursive rule for well formed formulas in N may be stated as:

1. Any single letter of N is a wff.

2. If P and Q are wffs, then (P)\(Q) is a wff.
(No formula of N will be regarded as being a wff unless its being so follows
from this definition.)

It should be noted that even the recursive rule for well formed formulas is
simpler in the Nicod System, since it needs only two rather than three clauses.
We have here the first fruit of using only one operator symbol.
The single axiom required (or rather, the single pattern for infinitely many
axioms of the object logic) is stated in our metalanguage, in which we use dots
for brackets, as

Ax. P.\.Q\R:\::T.\/r\T:.\:.S\Q:\:P\S.\.P\S

The single rule of inference that is required may be stated as

Rule. From P and P. | .R |() to infer Q.

The definitions of ‘valid argument in N’ and of ‘theorem of N\ that is, of

Pv F2, ..., Pn ky and (n P

are strictly analogous to those given for ‘ |rs’ and ‘ (ha\

Although more economical in the respects indicated, the Nicod System can
scarcely be said to be simpler than the systems already mentioned. There is only
one axiom for N, but it is more complicated than any axiom or postulate of any
of the other systems. Not only is it longer, but it involves five distinct proposi¬
tional symbols, ‘F\ <Q\ *R\ ‘S’,and T\ whereas the entire set of axioms for any
of the other systems can be stated in terms of just three distinct propositional
symbols. The situation is similar with respect to the rules of inference of the
several systems. The Nicod Rule is not Modus Ponens, but a more powerful

W. V. Quine, ‘A Note on Nicod’s Postulate , Mind, n.s. vol. 41 (1932), pp. 345-350.
B. A. Bernstein, ‘Remark on Nicod’s Reduction of Principia Mathematical, The Journal of
Symbolic Logic, vol. 2 (1937), pp. 165-166.
Thomas W. Scharle, ‘Axiomatization of Propositional Calculus with Sheffer Functors’, Notre
Dame Journal of Formal Logic, vol. 6 (1965), pp. 209-217.
Several alternative, one-axiom systems are listed in A. N. Prior, Formal Logic, Oxford University
Press, England, 1962, p. 306. Two in particular are noteworthy because they use only four
propositional variables, in contrast to Nicod’s five. They are also different from Nicod’s axiom in
that they contain no proper part that is itself a theorem. These systems are attributable to
Wajsberg (1931) and to Lukasiewicz (1931). I am indebted to Professor P. T. Geach of the
University of Leeds for calling them to my attention.
284 Alternative Systems and Notations [Ch. 9
instrument for deduction. Modus Ponens, which may be stated as

From P and P. | .£) | Q to infer jp*

is simply a special case of the Nicod Rule, in which ‘R’ and ‘Q* denote
identically the same well formed formula of the object language. Nicod’s Rule
as well as his axiom is more complicated than those required in less economical
systems of logic.
That the Nicod System is functionally complete has already been indicated
in our discussion of the stroke function in Section 9.5. The analyticity of N is
simply enough shown by establishing via truth tables that the Axiom is
tautologous, and that the Rule can lead only to tautologies from tautologies.
The deductive completeness of N remains to be proved to complete the proof
that the Nicod System is a Model Logic. When this is attempted, we find that
for all the economy of its single axiom and the greater deductive strength of
its Rule, it is very difficult indeed to derive theorems in the Nicod System.
Nevertheless, it is interesting from the point of view of seeing how far one can
go in the direction of reducing the number of postulates and still have a
deductively complete system of logic.
We shall develop seventeen theorems of the Nicod System, and one derived
rule. The last four theorems will be the four axioms of the Ililbert-Ackermann
System, and the derived rule will be the Ililbert-Ackermann System’s rule R' 1.
Of course, these must be stated in their unabbreviated form, and then expressed
in the Nicod System’s notation. The Hilbert-Ackermann axioms, expressed in
terms of that system’s primitive symbols — and v are

1. ^(PvP)vP
2. -Pv(Pv())
3. -(Pv())v(()vP)
4. ~~(^PvQ) v [-~(R v P) v (R v ())]

The Hilbert-Ackermann rule, similarly unabbreviated, is

From P and — P v £) to infer ()

These, expressed in the Nicod System’s notation, in which is written

\P |P\ and Tv Q’ is written T \ P. | .Q | Q\ are stated as:

1. P\P.\.P\P:\:P\P.\.P\P:.\ :.P \P. \ .P\P:\:P \P.\.P \P:: \ ::P\P.

2. P\P.\.P\P:.\:.P\P.\.Q\Q:\:P\P.\.Q\Q.

P:|:0|£.|.P|P.
4. P\P.\.P\P:\:Q\Q:.\:.P\r.\.P\P:\:Q\Q-.:\::P\P.\.r\P:\:Q\Q:.\:.P\

•PI P I :-R \R.\.Q\Q:\:R\ R-[ | ::.R | R. | .P| P:| :R| R.| .P| P:. |
:.R|R.|.P|P:|:fl|R.|.P|P::|::R|fl.|.gig:|:R|R.|.p|^.
R' 1. From P and P | P. | .P | P: | :Q \ Q to infer Q.
Sec. 9.8]
The Nlcod System 285

Our development of the Nicod System will be entirely in terms of the stroke
function. Since the formulas of the proofs are of almost intolerable length, they
will be described rather than written out. Our descriptions, however, will be
sufficiently complete to permit the reader to write them out for himself.

Theorem 1. k p:. |:.T.\.T\T: |:S::. |::.S: |:T. I .T12’:. |:.0;; |;:S: |:T. I .TI
T:.\:.Q.
Proof: Line 1 is the Nicod Axiom with T in place of P, of (), and of R. Line
2 is the Nicod Axiom with T.\.T\T in place of P and of Q, and with
S | T: | :T | S. | .T | S in place of R. Line 3 is the result of applying the Nicod Rule
to lines 1 and 2. Line 4 is the Nicod Axiom with S: | :T. | .T | T in place of P, with
T. | .T | T: | :S in place of £) and of R, and with Q in place of S. Line 5 is the
result of applying the Nicod Rule to lines 3 and 4.

Theorem 2. (nT.|.T|T

Proof: Line 1 is Theorem 1 with T. |.T|T in place of £), and with

SI71: \:T |S. \.T |S in place of S. Line 2 is the Nicod Axiom with 7\|.T|
TV. | ::T. | .T | T:. | :.S | T: | :T |S. | .T |S in place of P, with S |T: |:T |S. | .T |S:. |
:.T.\.T\T::\ ::T. \ .T | T in place of Q and of R, and with S | T::: | :::S | T: | :T | S. |
T | S:. | :.T. | .T \ T:: | ::S::. | ::.S | T: | :T | S. | .T | S:. | :.P. | .T | T:s |
| \ .T\T in place of S. Line 3 is the result of applying the Nicod Rule
to lines 1 and 2. Line 4 is the Nicod Axiom with S \T: \:T |S. | .7' |S:. | :.7\ |
.T\ T in place of P, with T in place of Q and with T | T in place of R. Line 5
is Theorem 1 with S\T:\:T\S.\.T\S:.\:.T.\.T \ 7:: | ::7\ | .T | Pin place of Q and
with S | T::: | :::S | T: | :T | S. | .T | S:. | :.T. | .T | T:: | | ::.S | T: | :T | S. |.7*| S:. |
:.T. | .T | T:: | ::S in place of S. Line 6 is the result of applying the Nicod Rule to
lines 4 and 5. Line 7 is the result of applying the Nicod Rule to lines 3 and 6.
Line 8 is the Nicod Axiom with T in place of P, of Q, and of R. Line 9 is the
result of applying the Nicod Rule to lines 7 and 8.

Theorem 3. |nS|P:|:P|S. |.P|S

Proof: Line 1 is the Nicod Axiom with P in place of Q and of R. Line 2 is

Theorem 2 with P in place of T. Line 3 is the result of applying the Nicod Rule
to lines 1 and 2.

Theorem 4. |nP|P. |.P

Proof: Line 1 is Theorem 3 with P |P in place of P and with P in place of

S. Line 2 is Theorem 2 with P in place of T. Line 3 is the result of applying the
Nicod Rule to lines 1 and 2.

Theorem 5. |nP|P:|:S|P. |.S|P

Proof: Line 1 is Theorem 3 with P | S in place of S and with S |P. | .S |P in

place of P. Line 2 is Theorem 3 with P in place of S and S in place of P. Line 3
is the result of applying the Nicod Rule to lines 1 and 2. Line 4 is the Nicod
286 Alternative Systems and Notations [Ch.9

Axiom with S |P. | .S |P in place of P, with P in place of Q with S in place of

R, and with P | P in place of S. Line 5 is the result of applying the Nicod Rule to
lines 3 and 4. Line 6 is Theorem 4. Line 7 is the result of applying the Nicod
Rule to lines 5 and 6. Line 8 is Theorem 3 with S |P. | .S |P in place of S, and
with P | P in place of P. Line 9 is the result of applying the Nicod Rule to lines 7
and 8.

Theorem 6. \sP. | .p|R:: |::S\Q: |:P|S. |.P|S:. |:.S|p: |:P|S. | .P\S

Proof: Line 1 is the Nicod Axiom with S|£): | :P |S. | .P |S:. |:.S |(): |:
P |S. | .P |S in place of P, with T. | .T |T:. |:.S \Q: |:P|S. | .P |S in place of Q and
of R, and with P. |.Q|R in place of S. Line 2 is Theorem 5 with S|£):|:
P | S. | .P | S in place of P, and with T. | .T | T in place of S. Line 3 is the result of
applying the Nicod Rule to lines 1 and 2. Line 4 is the Nicod Axiom. Line
5 is the result of applying the Nicod Rule to lines 3 and 4. Line 6 is Theorem 3
withS1^: |:P|S.|.P|Ss.|:.S|p:|:P|S.|.P|SinplaceofS,andwithP. \.Q\Rin
place of P. Line 7 is the result of applying the Nicod Rule to lines 5 and 6.

Theorem 7. k(>|S.\.U: |:g|S.\.U:. |:.S|^>.\.U

Proof: Line 1 is Theorem 6 with Q | S in place of P, with S | p in place of Q

and of R, and with U in place of S. Line 2 is Theorem 3 with Q in place of S
and with S in place of P. Line 3 is the result of applying the Nicod Rule to lines
1 and 2. Line 4 is Theorem 3 with (7. |.S|£> in place of S and with
Q | S. |. (7: | | S. |. U in place of P. Line 5 is the result of applying the Nicod
Rule to lines 3 and 4. Line 6 is Theorem 6 with S | Q. |. U in place of P, with
U. | .S | Q in place of Q and of R, and with Q | S. |. U: | | S. |. U in place of
S. Line 7 is Theorem 3 with S | Q in place of S and with U in place of P. Line 8
is the result of applying the Nicod Rule to lines 6 and 7. Line 9 is the result of
applying the Nicod Rule to lines 5 and 8. Line 10 is Theorem 3 with S | £). |. U
in place of S, and with Q | S. |.U: | :Q | S. |. U in place of P Line 11 is the result
of applying the Nicod Rule to lines 9 and 10.

Theorem 8. |NP.|.^|fi::|::9|S:|:P|S.|.P|S:.|:.g|S:|:f>|S.\.P\S

Proof: Line 1 is Theorem 6 with P. \-Q\R in place of P, with S|Q:|

:P|S. |.P|S in place of Q and of R, and with Q\S: |:P|S. |.P|S:.\:.Q\S: |:
P | S. | .P | S in place of S. Line 2 is Theorem 6. Line 3 is the result of applying
the Nicod Rule to lines 1 and 2. Line 4 is Theorem 7 with P | S. | .P | S in place of
U. Line 5 is the result of applying the Nicod Rule to lines 3 and 4.

Theorem 9. (n S:. |:.P. |.S|S: |:P. |.S|S

Proof: Line 1 is Theorem 3 with S | S. | .P in place of S, and with P. | .S | S:

| :P. | .S | S in place of P. Line 2 is Theorem 1 widi S | S in place of S. Line 3 is
the result of applying the Nicod Rulje ‘to lines 1 and 2. Line 4 is the Nicod
Axiom with P. | .S | S: | :P. |.S | S in place of P, with S | S in place of Qt and with P
Sec. 9.6] TI10 Nlcod System 287

in place of R. Line 5 is the result of applying the Nicod Rule to lines 3 and 4.
Line 6 is Theorem 2 with S in place of T. Line 7 is the result of applying the
Nicod Rule to lines 5 and 6. Line 8 is Theorem 3 with P. | .S | S: | :P. | .S | S in
place of S, and with S in place of P. Line 9 is the result of applying the Nicod
Rule to lines 7 and 8.

Theorem 10. fetC>|():|:p|S.|.p|S

Proof: Line 1 is Theorem 8 with Q | Q in place of P, with S | in place of

Q and of R, and with Q | S. | ,Q | S in place of S. Line 2 is Theorem 5 with Q
in place of P. Line 3 is the result of applying the Nicod RuIq to lines 1 and 2.
Line 4 is Theorem 3 with Q in place of P. Line 5 is the result of applying the
Nicod Rule to lines 3 and 4.

Theorem 11. |n Q:. |:.0|P. | .T:\:Q\P.\.P

Proof: Line 1 is Theorem 8 with Q in place of R and with P in place of 8.

Line 2 is Theorem 9 with Q in place of S. Line 3 is Theorem 8 with () in
place of P, with P. in place of £) and of R, and with Q\P:\:
P\P.\,P\P:.\:.Q\P:\:P\P.\.P\Pin place of S. Line 4 is the result of applying
the Nicod Rule to lines 2 and 3. Line 5 is the result of applying the Nicod Rule
to lines 1 and 4. Line 6 is Theorem 9 with P |P. | .P in place of S and with
in place of P. Line 7 is Theorem 4. Line 8 is the result of applying the Nicod
Rule to lines 6 and 7. Line 9 is Theorem 8 with Q jPin place of P, with P |P in
place of Q and of R, and with P in place of S. Line 10 is Theorem 8 with Q
in place of P, with Q\P:\:P\P.\.P\P in place of Q and of R, and with
P|P. |.P:.|:.g|P.|.P:|:^)|P.|.P::|::P|P.|.P:.|:.piP.|.P:|:()|P.|.Pinplaceof
S. Line 11 is the result of applying the Nicod Rule to lines 5 and 10. Line 12
is the result of applying the Nicod Rule to lines 9 and 11. Line 13 is Theorem 8
with <? in place of P, with P |P. |.P in place of <p and of R, and with Q\P.\.
P: | :Q |P. | .P in place of S. Line 14 is the result of applying the Nicod Rule to
lines 8 and 13. Line 15 is Theorem 8 with Q in place of P, with P | P. | .P:. |:.
Q\P.\.P:\:Q\P.\.P in place of Q and of R, and with Q:. | :.Q |P. | .P: j
|P. | .P::| |:.@|P. \.P:\’.Q\P-\-P in place of S. Line 16 is the result of
applying the Nicod Rule to lines 12 and 15. Line 17 is the result of applying the
Nicod Rule to lines 14 and 16. Line 18 is Theorem 10 with Q | P. | .P: | :Q | P. | .P
in place of S. Line 19 is Theorem 3 with £)|() in place of S, and with
Q:. |:.@|P. | .P:\ :Q\P. | .P:: ||:.() |P. | .P:\:Q\P. | .P in place of P. Line 20
is the result of applying the Nicod Rule to lines 18 and 19. Line 21 is Theorem
8 with Q:. \ :.Q |P. | .P: | :Q |P. | -P” I ::£>:• I: Q |P-1 P: I '-Q\P-1 .Pinplace of Pand
of S, and with Q in place of R. Line 22 is the result of applying the Nicod Rule
to lines 20 and 21. Line 23 is the result of applying the Nicod Rule to lines 17
and 22. Line 24 is Theorem 4 with Q:.\:.Q\P.\.P:\iQ\P.\.P::\::Q:.\
:.Q\P.\.P:\:Q\P.\.Pin place of P. Line 25 is the result of applying the Nicod
Rule to lines 23 and 24.
288 Alternative Systems and Notations [Ch. 9

Theorem 12. |n P: | :Q| R. | .0| R:: | ::Q: \ :P\R. \ .P\R:. \ :.Q: \ :P\R. \
■P\R

Proof: Line 1 is Theorem 8 with p in place of P, with Q |R. | .R in place of

Q and of R, and with P |R. | .P |R in place of S. Line 2 is Theorem 11 with R
in place of P. Line 3 is the result of applying the Nicod Rule to lines 1 and 2.
Line 4 is Theorem 8 with P:|:p|R. |.p|R in place of P, with p|R.|.R:|
:P|R.| -P|R in place of p and of R, and with p: |:P|R. | .P|R:.|:.p: |:P|R.
| .P | R in place of S. Line 5 is Theorem 8 with p \ R in place of p and of R, and
with R in place of S. Line 6 is the result of applying the Nicod Rule to lines 4
and 5. Line 7 is the result of applying the Nicod Rule to lines 3 and 6.

Theorem 13. |wP.|.p|g::.|:s.H|H.|.P|F:.|:.H|H.|.g|p:|:fl|H.|.C|

Proof: Line 1 is Theorem 12 with R | R. | .P | P in place of P, with P. |. p | p in

place of p, and with R |R. | .p|p*. |:R |R. | .p|p in place of R. Line 2 is
Theorem 8 with R | R in place of P, with P in place of p and of R, and with
P | p in place of S. Line 3 is the result of applying the Nicod Rule to lines 1
and 2.

Theorem 14. \n P\P.\.P\P:\:P\P.\.P\P:.\:.P\P.\.P\P:\:P\P.\.P\Pr.

\::P\P(Ax. 1 of II.A.)

Proof: Line 1 is Theorem 4 with P\P.\.P\P:\:P\P.\.P\Pin place of P. Line

2 is Theorem 4 with P | P in place of P. Line 3 is Theorem 8 with P |P. | .P |
P:|:P|P.|.P|P:.|:.P|P.|.P|P:|:P|P.|.P|P in place of P, with P|P.|.P|P in
place of Q and of R, and with P | P in place of S. Line 4 is the result of applying
the Nicod Rule to lines 1 and 3. Line 5 is the result of applying the Nicod Rule
to lines 2 and 4.

Theorem 15. \n P|P. |.P|P:. |:.P|P. | .@|p: |:P|P. | .Q |^(Ax.2ofH.A.)

Proof: Line 1 is Theorem 10 with P'j P in place of Q, and with Q in place

of S.

Theorem 16. kP|P.|.()|p:|:P|P.|.£|p:. |:.P|P. | .Q|£:|:P|P. | .Q\

Q::\::Q\Q.\.P\P:\:Q\Q.\.P\P(Ax.3oiU.A.)

Proof: Line 1 is Theorem 3 with P | P in place of S, and with p | p in place of

P. Line 2 is Theorem 4 with P |P. | .p | p: | :P |P. | .p | p in place of P. Line 3
is Theorem 8 with P|P.|.p|p:|:P|P.|.p|p:.|:.P|P.|.p|p:|:P|P.|.p|p
in place of P, with P|P. |.p|p in place of p and of R, and with p|p. |.
P | P: | :P | p. | .P | P in place of S. Line 4.is'the result of applying the Nicod Rule
to lines 2 and 3. Line 5 is the result of applying the Nicod Rule to lines 1 and 4.
 9-61 The Nlcod System 289

Theorem 17. In P\P. \ ,P\l>: |:p|p:.\:.P\P. I. P\P: I :p in:: i [:p\p, | .pi

P: t :Q 11>:.| :.P| P. | .P| P: | :p| p::; | :::R | R. | .p| J>; \:R\R.\ .?| f>;. | ;.R \ R. |
|P:|:fl|fl.|.P|P::|::fl|fi.|.p|p:|:fl|ji.|.p|p!:.|;:.R\R,| .P|F;|;«|R.| ,R|
P:. | :.R |R. | .P |P: | :R |R. | .P|p:; | ::fl |R. | .p \ Q; \ :R \ R. \ .p \ p (Ax. 4 of Il.A.)

Proof: Line 1 is Theorem 4 with P \P.\.P\P-.\:p | p:. | :.p |}'. \ .p | p; | :p | q iu

place of P. Line 2 is Theorem 9 with P for S and P \ P for P. Line 3 is Theorem 8
with P\P.\.P\P in place of p and of R, and with p|p in place of S. Line 4
is the resrdt of applying the Nicod Rule to lines 2 and 3. Line 5 is Theorem 8
withP |P. | -P |P: |:() K>:. | :.P |P. | .P |P: |:{) \Q-.: \ ::P \P. \ .P\P: \:{) \ {):. |:./’ | P. |
.P\P-.\:Q\Q in place of P, with P\P.\.P\P:\-.Q\Q in place of () and of R,
and with P. \ .Q | Q: \ :P. \.() | ^ in place of S. Line 6 is the resrdt of applying the
Nicod Rule to lines 1 and 5.1 ane 7 is the resrdt of applying the Nicod Rrtle to
lines 4 and 6. Lirre 8 is Theorem 13. Line 9 is Theorem 4 with R\R.\ ,P\P: \
:R|R.|.P|P in place of P. Line 10 is Theorem 8 with R|H.|.P|P:
\:R\R.\.P\P:.\:.R\R.\.P\P:\:R\R.\.P\P in place of P, with R|R.|.P|P
in place of (f and of R, and with R |R. |.{) |{)■. | :R |R. |.{) \in place of S.
Line J1 is the result of applying the Nicod Rrtle to lines 9 and 10. Line 12 is
Theorem 8 with /’. | .01(1 in place of /' with R|R.| .P\P:.\ :.R |R. |.()|(): | :R|
R.|.pKl in place of (1 and of R, and with R|R.|./’| P: | :R | fl. |./’ | P:. \ :.R |
R. | .P|P: | :R |fl. | .P\ | ::R | R. | .<;) | Q: \ :R | R. | .((l | | ::.R | R. | .P\P:
|:fl|R.|.P|P:.|:.R| R. | .P\P: | :R | R. | .P| Pi: | ::R | R. |.{) K): | :R | R. |.() | pin
place of S. Line 13 is the result of applying the Nicod Rule to lines 8 and 12.
Line 14 is the result of applying the Nicod Rrtle to lines 11 and 13. Line 15 is
Theorem 8 with P|P. |.P| P: \:() | p:. |:./’ | /’. |./’ | P: | =p | p:: |::/' | P. | .P \ P: \ :p |
p:.|:.P|P. |.P|P:|:PIP in place of with P. |.p|P in place of p and of
R,smdwHhR\R.\.P\P:\-.R\R.\.P\P:.\-..R\R.\.P\P:\:R\R.\.P\P:.\::R\R.\.
PIP:|:R|R.|.p|p::,|::.R|R.|.P|P:!:R!R.|.P|P:.|:.R|R.|.P|P:|:R|R.|.
P\P:: |::R |R. |.p |p: |:R |R. |.p |pitt place of S. Line Hi is the resrdt of apply¬
ing the Nicod Rule to lines 7 and 15. Line 17 is the result of applying the
Nicod Rrtle to lines 14 and Hi.

DR i . P, P | P. | .P |P: |:p |p h P (R' 1 of Il.A.)

Proof: Line 1 is Theorem 9 with P in place of S, and with P \ P in place of
P. Line 2 is the premiss P. Line 3 is the result of applying the Nicod Rrtle to
lines 1 and 2. Line 4 is the premiss P\P. | .P\P: |:p |p. Line 5 is the resrdt of
applying the Nicod Rule to lines 3 and 4.

Deriving all tautologies from a single axiom (axiom form) by a single rule in
terms of a single operator is thus seen to be possible. But it is a tedious business.
As Dr. Samuel Johnson is said to have remarked in quite another connection, it
is ., like a dog’s walking on his hind legs. It is not done well; but you are
surprised to find it done at all.’
 io
A First-Order Function
Calculus
10.1 The New Logistic System RS1

In Chapter 4 and the first four sections of Chapter 5, we used logical

principles governing the quantification of individual variables to prove the
validity of arguments and to demonstrate logical truths. An axiomatic devel¬
opment of those principles is called a ‘first-order function calculus\ or, alter¬
natively, a lower* or ‘restricted* function calculus.1 In this chapter we con¬
struct such a logistic system, develop some of its theorems, and prove that it
has such (desirable) characteristics as consistency and completeness. Again our
metalanguage will be ordinary English plus some elementary arithmetic, and
some special symbols that will be introduced and defined as needed. Our object
language or object logic is the new system RSls which will now be described.
The logistic system RS1 contains infinitely many primitive symbols of the
following categories.
1. Infinitely many capital letters from the first part of the alphabet, with and
without subscripts:

A, C, Aj, B|, Cj, A2, ••*

These are propositional constants, and on the system’s intended interpretation

will express noncompound propositions.
2. Infinitely many capital letters frqjn the middle part of the alphabet, with
and without subscripts:

P, Q, R, P|, 01, p2> 02> ^2’ ■ • •

These are propositional variables, and on the system’s intended interpretation

will be statement variables of the kind discussed in Section 2.3 of Chapter 2.
Symbols of the first two categories are propositional symbols.
3. Infinitely many capital letters from the first part of the alphabet, with and
without subscripts, having right-hand superscripts T, ‘2*, ‘3’-

1 Also a ‘first-order’, ‘lower’, or ‘restricted’ predicate calculus.

290
Sac. 10.1J The New Logistic System RS, SSI
A1. B\ C\ AJ, B\, CJ, A\, B\, C\, ...
A2, Bl Cl Af, Bf, Cl Al Bl Cl-..
Al Bl Cl A\, Bl Cf, Af, Bl C\, ...

These are predicate constants, and on the system’s intended interpretation each
of them will designate a particular attribute or binary relation or ternary
relation... or n-ary relation according as its right superscript is ‘1’ or ‘2’ or ’3’
or ... or Vi\
4. Infinitely many capital letters from the middle part of the alphabet, with
and without subscripts, having right-hand superscripts T, ‘2^, ‘3\ ...

P\ Q\ R\ Pi Ql R\, Pi Ql Rl...
P2, Q\ R\ Pi Ql Rl Pi Ql Rl...
P\ Q*> Rl Pi Ql Rl Pi Ql Rl ...

These are predicate variables, and on the system’s intended interpretation will
be symbols for which names of particular attributes, binary relations, and so
forth, can be substituted. Symbols of the third and fourth categories are
predicate symbols.
5. Infinitely many lower-case letters from the first part of the alphabet, with
and without subscripts:

a, b, c, Oj, fc,, Cj, a2, b2, c2, ...

These are individual constants, and on the system’s intended interpretation

will be proper names of individuals.
6. Infinitely many lower-case letters from the latter part of the alphabet,
with and without subscripts:

X, y, Z, Xf, J/i> *2’ ?/2’ z2> •

These are individual variables, and on the system’s intended interpretation will
be individual variables of the kind discussed in Chapter 4. Symbols of the fifth
and sixth categories are individual symbols.
7. Just four additional symbols complete the list of primitive symbols of RS1;
these are the tilde, the dot, and left- and right-hand parentheses:

■, (. )

In addition to the primitive symbols we introduce some defined symbols into

our object language RSr Before doing so, however, we must indicate the use of
certain special symbols of our metalanguage. . As in Chapter 7, we introduce the
special symbols ‘~\ \\ and ')’ to denote the symbols ‘(\ and J of
292 A First-Order Function Calculus [Ch. 10

the object language, and we shall also use brackets and braces to denote the
object language’s parentheses when that is conducive to easier reading. Capital
letters, with and without subscripts, will be used in* the metalanguage as
syntactical variables, that is, as symbols for which designations of any symbols
or sequences of symbols of the object language can be substituted. Lower case
letters, with and without subscripts, will be used as syntactical individual
variables, that is, symbols for which designations of individual symbols of the
object language can be substituted. Finally, we adopt the convention that the
juxtaposition of two symbols of the object language will be denoted in the
metalanguage by the juxtaposition of their names. Thus in any context where
4F’ denotes ‘A1’ and V denotes V, 4F(x)’ will denote *A1(jc)’. We shall also find
it convenient to insert commas into any sequence of symbols of the meta¬
language that designate individual symbols of the object language. Thus where
*F' denotes 4B3’ and4x1’,4x2\ and x3’ denote a2\ and 4a3\ respectively, we
shall use *F(xv x2,x3)’ to denote 'B*(axa2az)\
We introduce the symbols V, ‘D\ * = ’, and 43’ into the object language by
definition, and denote them in the metalanguage by the symbols V,4 D\ 4=\
and 43\ The new symbols of the object language are introduced as definitional
abbreviations:

Df, PvQ is defined to be an abbreviation of

Df. P 3 Q is defined to be an abbreviation of — (P—()).
Df. P=: Q is defined to be an abbreviation of (P 3 Q)*(Q 3 P).
Df. (3 x)P is defined to be an abbreviation of ~~(x)~~P.

We shall feel free to drop parentheses (in the metalanguage), retaining only
those required to avoid ambiguity, or to make for greater ease of interpreta¬
tion. We shall also, on occasion, write as *PQ\ Although we shall not
always take advantage of the following convention by dropping unnecessary
parentheses, we set up the following order of precedence among the symbols of
our metalanguage, each symbol having precedence over any that lies in a
column to its right:

= D v • —

(x) where not immediately preceded by a predicate symbol

(3x)

This convention dictates that an expression such as

P=(x)Qv~~RS D (3x)T'U

is understood to denote the same formulas of the object language that are
denoted by ’.
Sec. 10.1] The New Logistic System US, 203
ip)= {{[(*K>M(~fl)-(S)]} D {[(3jt)T]•[[/]}}

Our final convention is that of association to the left, which means that where
the convention of order of precedence does not suffice to remove the ambiguity
of an expression, its parts should be grouped by parentheses to the left. That is,
when an expression contains two (or more) occurrences of the same connective,
and their relative scopes within the expression are not otherwise indicated, the
occurrence to the right shall be understood to have the wider (or widest) scope.
We define formula of RSt as any finite sequence of symbols of RSP Among
these are included such sequences as

)~)(
(x)(Al(x))
(&y)(B2(xy)))D(Cl(a))
(3 v ~ =
Q3(ab)

of which we shall want lo include only the second and third as well formedf,
that is, as meaningful on the system’s intended or normal interpretation.
Now we define well formed formula of RS { by the following recursive rules:

(a) 1. If F is a propositional symbol, then F is a wff.

2. If F is an n-adic predicate symbol and xv x2, ..., vM are n (not
necessarily distinct) individual symbols (where n = 1, 2, 3,...),
then F(xv x2,... ,xn) is a wff.
(/?) 1. If F is a wff\ then ~~(F) is a wff.
2. If F is a toff and G is a wff, then (F)*(C) is a wff.
3. If F is a wff and x is an individual variable, then (x)(F) is a wff.
No formula of RSj is a wff unless its being so follows from these rules, or from
these rules together with the definitions of the defined symbols that have been
stated.

Now that we have an effective criterion for wff\ we shall restrict our
discussion in the remainder of this chapter to wffs, which alone interest us.
It is convenient at this point to introduce and define some additional special
terms. (There is no loss of generality in phrasing our definitioas in terms of
undefined symbols only, for defined symbols are always eliminable.)

Df. If x is an individual variable, then (x), when not immediately preceded

by a predicate symbol, is the universal quantifier of x (or on x).
Df. If Q is a wff that occurs in a wff F, then Q is a well formed part of F.
Df. An occurrence of a variable x in a wff P will be called a bound
occurrence of x in F if it is in a well formed part of F of the form (x)Q.
294 A First-Order Function Calculus [Ch. 10

Df. An occurrence of a variable x in a wff P will be called a free occurrence

if it is not bound.

We assume infinitely many postulates for our object logic. Every wff of any
of the five following patterns is a postulate:

PI .PD (P-P)
P 2. (P-Q) D P
P3. (PDQ)D[-~(Q-R)D~(R-P)]
P 4. (x)(P D Q) D [P D (*)£>], where x is any individual variable, P is any
wff containing no free occurrences of x, and Q is any wff.
P 5. (x)P D Q, where x is any individual variable, y is any individual
variable or constant, P is any wff, Q is the result of replacing each free
occurrence of x in P by y, and if y is a variable, then it must occur free in Q
at all places where x occurs free in P (that is, no bound occurrence of y in Q
is the result of replacing a free occurrence of x in P by y).

The restrictions on P 4 and P 5 serve to prevent the inclusion of such

manifest falsehoods (on the intended interpretation) as

(*)[(* = 1) D (x + x = 2)] D [(% = 1) D (x)(x + x = 2)]

this statement has a true antecedent and a false consequent, but it is not an
instance of P 4, since x occurs free in (x = 1) Also prevented is such a patent
falsehood as

(x)[(ly)(y ? x)] D [(3y)(y # y)]

which is not an instance of P 5, since there is a bound occurrence of the

variable y in (3y)(y ^ y) which is the result of replacing a free occurrence of
the variable x in (3y)(y ^ x) by y.
We assume two rules of inference for RS1:

R 1. From P and P D Q to infer *

R 2. From P to infer (x)P.

Next we define ‘demonstration* for RS1. Formally,

^i> P2> • ■ •» Pn 1“ Q

asserts that there is a finite sequence of wffs Sv S2,..., St such that for every
S,- (1 < / < t) either:

(a) is one of the Postulates P i-P' 5; or

(b) is one of the premisses P4 (1 < i < n); or
Sac. 10.1]
The New Loglallc System RS, 200
(c) S, is the result of applying R 1 to two earlier S's of the sequence, sav &
and Sk where i < / and k < /; or
(d) S, is the result of applying R 2 to an earlier S, of the sequence, so that S#
is (x)Si where i < /; and S, is p.

Informally, we regard P,, P2,..., Pn b p as asserting that O is validlv inferred

from P,, P2, ..., Pn in RSr
Next we define ‘theorem of RS,’ as any wjf p such that b p. It should be
noted that b is a special symbol of our metalanguage, and does not occur in
RS, itself.
We can now establish the consistency of RS,. We begin by stating the
following definition:

Df. Where P is any wff of RS,, its associated propositional formula (abbre¬
viated to a.p.f. and symbolized F°) is the formula that results from F by first
deleting all occurrences of quantifiers in F, together with all parentheses that
were required by those quantifiers, and then replacing every well formed part
of F of the form Pn(xv x2,..., xw) and every propositional symbol in F by the
propositional constant A.

As an example, where F is

((*1)((^2)(A2(^1r2)))) D (((xl)(B1(xl)))9((x2)(Bl(x2))))

it’s a.p.f. F° is

(A) D ((A)-(A))

Next, we state and prove the following:

Lemma: If b F, then F° is a (truth table) tautology.

Proof. 1. First we show that every postulate of RS, has only tautologies for
its a.p.f.’s. An a.p.f. of any instance of P 1 is of the form P° D (P° *P°), which
is a tautology. Similarly, for P 2 and P 3. The a.p.f. of any instance of P 4 has
the form (P° D p°) 3 (P° D p°), which is a tautology. In P 5, since p differs
from P only in the individual symbols they contain, their a.p.f.’s are identical,
so the a.p.f. of any instance of P 5 has the form P° D P° which is a tautology.
2. Next, we show that any wff obtained by applying the rules of RS, to wffs
that have tautologies for their a.p.f.’s has a tautology for its a.p.f. By R 2 we
derive (x)Pfrom P. But (x)Pand Phave identically the same a.p.f. Therefore, if
the a.p.f. of P is a tautology, then the a.p.f. of (x)P is also. By R 1 we derive p
from f and P D p. So if P° and P° D Q° are tautologies, then p° is also.
Since all theorems of RS, follow by the rules from the postulates, all
theorems of RS, have tautologies as their a.p.f.\s. This completes our proof of
the lemma. The consistency of RS, will be stated as
296 A First-Order Function Calculus [Ch. 10

Metatheorem I. RSX is consistent.

Proof: (Here we use Post’s criterion for consistency: a system is consistent if

it contains a wff that is not a theorem.) The formula ‘(A)'{'-'(A))’ is a wffot RSX
that has an a.p.f. (itself) which is not a tautology. Hence, by our lemma, it is
not a theorem of RS1# so RSX is consistent. The consistency of RSX also follows
from its analyticity, which is remarked (in passing) in Section 10.6.

10.2 Development of RS1

In developing the present system, we shall borrow freely from the results
obtained in Chapter 8, in which it was established that every tautology can be
demonstrated as a theorem in the Propositional Calculus based on P 1, P 2, P 3,
and R 1. This is noted in the following theorem and derived rule.

Theorem O. All tautologies are theorems, and if in any tautology T that

contains the propositional symbols Pv P2, • • •, Pn we replace all occurrences of
Pv P2, ..., Pn, respectively, by any wffs Fv F2, ..., Fn of RSV the result is a
theorem F of RSV

Proof: By the completeness of the Propositional Calculus based on P 1, P 2,

P 3, and R 1, there is a demonstration in RS x for every tautology. If in every step
of that demonstration, we replace all occurrences of Pl> P2’ ■ ■ ■ > Pn> by FV p2’
. .., Fn, we obtain a demonstration in RSX of F, since making the indicated
substitutions in the postulates P 1, P 2, P 3 gives us simply other instances of
those postulates.

Derived Rule O. All tautologically valid argument forms are de¬

monstrable in RS1? and if in any tautologically valid derived rule S2, ...»
Sm h T that contains the propositional symbols Pv P2, ..., Pn, we replace all
occurrences of Pv P2, ..., Pn, respectively, by any wffs Fv F2, ..., Fn of RSt,
the result is a demonstrably valid derived rule G1# G2, ..., Gm h IJ of RSr

Proof. Exactly parallels the proof of Theorem O.

In any proof in RSls any step that is justified by Th. O or DR O will be noted
simply by ® (for ‘Propositional Calculus’).
At this point we proceed to establish the first few derived rules and theorems
of RSX.

DR 1. If P contains no free occurrence of x, then P D Q\- P D {x)Q.

Demonstration: Sx: PD Q premiss

S2: (x)(P D (?) . R 2
S3: (x)(P D Q) D [P !D (x)Q] P 4
S4: P D (x)Q R 1
Sec. 10.2] Development of RS, 297
DR 2. (x)[F(x) D G(x)], (x)F(x) h (x)G(x)

Demonstration: S, (x)[F(x) D G(x)] premiss

52 (x)[F(x) D G(.v)l 3 [F(x) D C(x)} P 5
53 F(x) D C(x) R I
54 (x)F(x) premiss
S, (x)F(x) D F(.v) P 5
S« F(x) R 1
S7 G(x) R 1
S. (*)C(*) R 2

DR 3. (*)(P-p) [-(*)/’•(*)?

Proof: (x)iP-Q) premiss

ix)iP-Q) D P-Q P 5
P-Q R 1
P ©
ix)P R 2
Q ©
(*)Q R 2
(x)P-(x)V ©

DR 4. ix)iP=Q)\-ix)P = i.x)Q

Proof: (x)(P = Q) premiss

(x)(P=Q) D (P^V) P 5
P=Q R 1
P DQ ©
(x)P D P P 5
(x)P => Q ©
ix)r D (x)Q 1)R 1
Q J P ©
P 5
ix)V O Q
ix)Q D P ©
DR 1
(x)Q 3 (x)P
[ix)P D ix){)\-[ix)Q D ix)P| ©
ix)P=ix)Q df.

Th. r: h(x)(P-0 = (x)P'ix)Q

Proof: First wc establish A: b (x)iP-(J) D (x)P-(x)()

h ix)iP-Q)DP-Q P 5
t-P-QD P P 2
\-ix)(P-Q) DP ©
I- ix)iP-Q) D ix)P DR 1
298 A First-Order Function Calculus [Ch. 10

By steps similar to the above we obtain:

i- (*xp-0 =>(*)<?
whence I- (x)(P’Q) D (x)P’(x)Q ®

Next we establish B: h (x)P‘(x)Q D (x)(P'Q)

h (*)P DP P '5
I- (x)Q DQ P 5
I- {x)P-(x)Q D P-Q ®
\-(x)P-(x)Q D (x)(P-Q) DR 1

Now from A and B,

h[(x)(P.g) D(x)P-(x)Q]-
[(*)?•(*# D (x)(P-Q)] ®
h(*xp-p) = (*)p-(*)g df.

T h . 2. h(x)(PD(?)D[(x)PD(x)g]

Proof: I- [(x)(P D Q)-(x)P] D (x)(P D Q) P 2

I- [(x)(PD0•(*)?] D(x)P ®
I- (x)(P DQ)D(PDQ) P 5
■ I- (x)P DP P 5
f- [(x)(P D Q)-(x)P] D [(P D g)-P] ®
h [(P D Q) *P] D Q ®
\-[(x)(PDQ)-(x)P]DQ ®
I- [(x)(P D g)-(x)P] D (x)Q DR 1
t- (x)(P D Q) D [(x)P D (x)Q] ®

Next we state and prove the Deduction Theorem for RSt, which corresponds
to the strengthened method of Conditional Proof used in Section 3.8 and
Chapters 4 and 5.

Metatheorem II. (Deduction Theorem—D.T.) If there is a demon¬

stration that Pv P2,, Pn_1, Pn\- Q in which no variable occurring free in Pn
is quantified by R 2 in any step, then there is a demonstration that Pv ***>
Pn-1 h Pn D 9<n which exactly those variables are quantified by R 2 that were
quantified by R 2 in the original demonstration?

Proof: We assume that there is a sequence of wjfs Sv S2,..., Sf such that

every Sf (1 < ; < 0 is (a) one of the Postulates P 1-P 5, or (b) one of the P/s
(1 < i < n), or (c) results from applying R 1 to two earlier S’s of the sequence,
2The restriction that no variable that is free in Pn is quantified by A 2 parallels exactly the
restriction on UG in Chapter 4, because Pn here has the role of an assumption within whose scope
lie subsequent lines of the first demonstration.
Sec. 10.2] Development of RSj 299
or (d) results from applying R 2 to an earlier S; and is Q. Now consider the
sequence of wffs: Pn D Sv Pn D S2,..., Pn D St. If we can ‘fill in wffs before
each Pn D S;- in such a way that the resulting total sequence is a demonstration
from Pv P2, ..., Pn_1, so each line of the resulting total sequence is either (a)
one of the Postulates P 1-P 5, or (b) one of the P/s (1 < i < n - 1), or (c)
results from applying R 1 to two earlier lines of the sequence, or (d) results
from applying R 2 to an earlier line of the sequence, then since the last line
Pn ^ St is Pn 3 Q, we shall have a demonstration that Pv P2,..., Pn_1 h Pn 3 Q.
The proof proceeds by weak induction on the number of lines (t) in the
original demonstration.
(a) In case t = 1 we have only the formula Pn 3 Sx to consider. We wish to
show that Pv P2, ..., Pn_x I- Pn 3 Sv By assumption, St is either a Postulate
P 1-P 5 or a Pj (1 < i < n).

case 1. S1 is a Postulate. Here we fill in with the demonstration of

Sx D (Pn D Sx) and Sx itself. From these we have Pn 3 Sx by R 1, so the total
sequence up to and including Pn 3 Sx is a demonstration that h Pn 3 Sx and
hence that Pv P2, .. ., Pn_x h Pn D Sr

case 2. Sj is a Pt (1 < i < n). Here we fill in with the demonstration of

Sx D (Pn D Sx) and Sx itself, from which we have Pn D Sx by R I. Here the
total sequence is a demonstration that Sx h Pn DSX, and since Sx is a Pt
(1 < i < n — 1) it is a demonstration that Pv P2, ..., Pn_ j h Pn D Sv

case 3. Sx is Pn. Here we fill in with the demonstration of Sx D Sx, which is

a demonstration that h Pn D Sx, and hence that Pt, P2, ..., Pn_x h Pn D Sx.

(/?) Now suppose we have filled in for all lines Pn D Sv Pr D S2,..., up to

and including Pn D Sfc_1? so we have a sequence of wffs, which is a demon¬
stration that Pj, P2,..., Pn_x h Pw 3 S,_r Under this assumption, we show how
to fill in to include Pn D SA in the sequence. This will then be a demonstration
that Pv P2, ..., Pn_i h Pn 3 Sfc. By assumption, S* is either a postulate, a
(1 < i < n), resulted in the original demonstration from applying R 1 to two
earlier S’s, or resulted in the original demonstration from applying R 2 to an
earlier S.

case 1. Sk is a Postulate. Fill in as in the a-case.

case 2. Sk is a Pi (1 < i < n). Fill in as in the a-case.

case 3. Sfc results from applying R 1 to two previous S's, say S4 and Sj where
S. is St D Sk. By the /8-case assumption, since i<kt j <k we already have
Pn D Si and Pn D Sjt which is Pn D (Sf D Sk). Here we insert the demonstra-
tion of’[PB D (St D SJ] D [(Pn 3 S() D (Pn 3 SJ] (a tautology), and PB 3 Sk
follows by two applications of R 1.
300 A First-Order Function Calculus [Ch. 10

case 4. Sk results from applying R 2 to an earlier S, say Sj where / <*.By

the /i-case assumption we already have Pn D Sj. Sk-is-(x)Sj where x, by our
original assumption, is not free in Fn. Hence by R 2, (x)(Pn D S;), and by P 4,
(x)(Pn D Sj) D (Pn D (x)Sj). Now R 1 will give Pn D (x)Sj which is Pn DSfc.

From a, [i by weak induction we can thus fill in for any number of steps S;. of
the original demonstration. Moreover, no variable is quantified in the ‘filled in’
sequence which was not quantified in the original sequence. This proves
Metatheorem II (D.T.).

MT II. Corollary: The D.T. as stated above holds for any system of
logic which has only rules R 1 and R 2 and which contains demonstrations for

P D P, P D (Q D F), [F D (Q D fl)] D [(F D Q) D (P D fl)], and

(x)(P D ()) D (P D (x)Q) where no free x’s occur in P.

Proof: Obvious.

We can illustrate the use of the D.T. by using it to prove DR 5:

DR 5. (x)[F(x) D C(x)] h (x)F(x) D (x)G(x)

Proof: (x)[F(x) D G(x)], (x)F(x) h (x)G(x) DR 2

' (x)[F(x) D G(x)] h (x)F(x) D (x)G(x) D.T.

Some additional theorems which can be established quite easily using the
Deduction Theorem are

Th. 3. h (x)(P=Q)D[(x)P=(x)Q]
#T h . 4. h (x)(P D Q) D [(3x)P D (3x)Q]

Their proofs will be left as exercises for the reader. We now proceed to some
theorems that state equivalences. Since (3x) was introduced as an abbreviation
for ~~(x)~~, the next theorem,

T ii . 5. h(3x)P= ~~(x)~~P

follows immediately from I-(x)~~Fee^ —(x)^F (®) by definition.

The following theorems, however, require proofs that are simple but not
quite so simple as that of Th. 5.

Th. 6. h(x)P= ~~(3x)~~P

Th. 7. h ~-(x)P=(3x)~~P

Th. 8. h ~~(lx)P=(x)~~P
Sec. 10.2] v««wv|viiieiii vi nvj wv

The proofs of these will also be left as exercises.

The next theorem provides for the permutation of universal quantifiers:

T H • 9- 1-(x)(y)P=(y)(x)P
Proof: b (y)P D P P 5
\~(x)[y)PDP] R 2
h (x)(y)P D (:x)P DR 5
b (x)(y)P D (y)(x)P DR 1

We obtain b (y)(x)P D (x)(y)P in the same fashion, and then have

b (*)(«f)P=(y)(x)P by®

Another theorem that can be proved quite simply is

Th. 10. b[(x)Pv(*)£>] D (x)(PvQ)

Proof: b (x)P DP P 5
b (x)Q D Q P 5
b [(x)Pv(x)Q]D(PvQ) ®
b [(r)Pv (x)Q] D (r)(/’v Q) DR 1

Having derived a number of equivalences as theorems, it will be convenient

to establish a rule of replacement which will permit the interchange of
equivalent formulas in any context. This will be proved as the next Meta¬
theorem.

MetatheoremIII (Rule of Replacement—R.R.). Let Pl% P.it..., Pn, be

any wffs, let A be any wff that does not occur in any Pi$ and let W be any wff
that contains no symbols other than the Pfs (1 <i < n), A, •, (x), and
parentheses.3 Let W* he the result of replacing any number of occurrences of
A in W by B. Then A = B b W = W*.
Proof: We use strong induction on the number of symbols in W, counting
each occurrence of P{ (1 < i < w). A, B, •, (x), as a single symbol.

(a) In case W contains a single symbol, W is either a Pi or A.

CASE 1. W is Pi. Here W* is P{, and since h P{ = l\, we have b W= W*,

whence A = B h W =. W*.

case 2. W is A. Here W* is either A or B.

3Tlie last restriction does not limit the generality of the Metatheorem, since all occurrences of
the defined symbols v, D, =, and 3 can be replaced bv undefined symbols.
302 A First-Order Function Calculus [Ch. 10

Subcase A: W* is A. Since h A = A, we have h We W*9 whence

A = Bh W= W*.
Subcase B: W* is B. Since A = B b A = B, we have AeBH We W*

(/8) Now suppose the Metatheorem true for any wff W that contains k or
fewer symbols, and consider a tvff W that contains k + 1 symbols. W must be
^L, (x)L, or MmN.

case 1. W is L must contain just & symbols, so by the /8-case assump¬

tion A = B b L = L*. But (.L = L*) D = ~~I*) is provable, so
A = BI-L = ~~ L\ which \sA = B^W=W\ Note that (— Lf = ~~(L*).

case 2. W is (x)L. Again L contains just Je symbols, so by the /8-case

assumption A = B b L = L*. Hence by R 2 we have A = B b (x)(L = L*), and
by DR 4 we have (x)(L = L*) b (x)L = (x)L*, whence AeBI- We W*. Note
that (x)(L*) = ((x)(L))*.

case 3. W is M'N. Since M and N each contains fewer than k symbols, by

the /8-case assumption A = B\~M = M* and A = B\-N=N*. Since
h [(M=M*)-(N=N*)] D [M-N=M*-N*], and every M*-N* is a W\ we
have A Bb Wee W*.

This completes the induction and finishes the proof.

MT III. Corollary: If W and W* are as in MT III, then A = B,

W b W*. The proof is quite obvious.

10.3 Duality

We begin our discussion of duality with a quite complicated definition:

Df. Let W be any wff that contains no occurrences of D or = (any wff can
be made into such a W by rewriting every well formed part of the form
as — Pv Q and every well formed part of the form P=Q as
(—P v ^)) v ?)). Where Pl9..., are propositional symbols or composed
of predicate symbols followed by the appropriate number of individual sym¬
bols, W will be constructed out of Pv P2,..., Pn, •, v, (x), (3x) exclusively.
Then the dual of W (written WA) is formed by

replacing every occurrence of Pi in W by ^P{ ,4

replacing every occurrence of ^Pi in W by -Pi9
replacing every occurrence of (x) in W by (3 or),
replacing every occurrence of (3x) in W by (x),
replacing every occurrence of • in W by v,
replacing every occurrence of v in W by •.

4Except those occurrences of Pi in well formed parts of the form — P{.

Sec. 10.3)

Examples (where P, Q, R, S are P/s):

1. W: P-Q
WA: •—P v —f)
2. W:(x)(PvQ)
WA: (3x)(~P*~p)
3. W: (y)(3z)[Pv(~£>vR-S)]
WA: (3t/)(z)[~P-(>(~R v ~S)]

There are several immediate consequences of our definition: First, where W is

any wff that contains no part of the form ~~Pf:

1. W = WAA

Where W and U are any whatever:

2. (W-l/)A = WAv t/A

3. (Wv I/)A = WA-(/A
4. ((x)W)A = (3x)WA
5. ((3x)W)A = (x)WA
fi t~WiA f a- If Wis a Pj, then (~W)A is Pv
' ' \ b. If Wcontainsat least two symbols, then (~W)A = —(WA).

We can now establish a general duality result.

Metatheorem IV. (Duality Theorem) If WA is the dual of W, then

\- —W = WA.
Proof: We use strong induction on the structure of W (i.e., the number of
symbols it contains, counting each I] as a single symbol).

(a) If W contains just one symbol, W is a Pr Here WA is —Pi% and since

I- —P{ eh —Pi by ®, we have I-W E= WA.
(/3) Assume the Metatheorem true for any W containing k or fewer symbols.
Now consider any W containing k + 1 symbols.

1. W is ~R.
case
Subcase A: R contains more than one symbol. Then (~R)A is ~(RA). By
the /8-case assumption, h ~~R = RA, hence by ® I- ~~R = ~(RA), or
|-R = (~R)A which is h ~W = WA.
Subcase B: R contains just one symbol, that is, R is a P{. Then W is ~P,
and WA is Pv By ® b ~~P( = Pt, whence h ~W = WA.

case 2. W is (x)R. By fi we have I- ~R ~ RA. We also have, by ® and R.R.,

I- ~(x)R = ~(x)~~R, so, by R.R., b ~(x)R = ~(x)~flA, which (by defini¬

tion) is I- ~(x)R = (3x)fiA or I- ~ W= W\
304 A First-Order Function Calculus [Ch. 10

case 3. W is (3x)R. By /?, b — R = BA. We also have (Theorem 8)

b ~~(3x)R = (x)^R. By R.R. we have b '-~(3jc)R = (x)BA, which is
b ~~W= W\

CASE 4. W is A-B. By /3, b^A = AA and b = By ®,

b —(A *B) = v — B, so by R.R., b ~~(A*B) = AA v BA, which is
b ^W= WA

CASE 5. W is AvB. By /?, I-A = AA and b = BA. By ®,

b —(A vB) = ~~A• —B, so by R.R., b ^(A vB)~ AA*BA, which is
b — Wee Wa

MT IV. Corollary: b (W= U) D (WA = t/A)

The proof of the corollary is obvious.

Duality has many uses. The traditional ‘Square of Opposition'5 which

displays I and O propositions as the negations of E and A propositions,
respectively, is clearly a special case of the duality result established above:

The familiar theorems of De Morgan are also special cases of the duality
theorem. And the general duality theorem permits us to formulate negations
with ease. For example, the proposition

There is a course that all students take.

may be symbolized as

(lx){C\x)-(y)[S'(y) D T*(y,x)])

and may be rewritten without using the implication sign to become

(3*){C1(x)*(i/)J~S1(i/) v T2(y, x)]}

5 Discussed in Chapter 4, page 68.

Sec. 10.3] Duality 305
The dual (and hence the negation) of this formula is

W{-C%v) v (3y)[S\y)-~~T*(y, *)]}

which is most ‘naturally’ written using the implication sign as

%i)]}
and is the symbolic translation of the English sentence

For every course there is some student who does not take it.

Having established the Rule of Replacement and the Duality Theorem, some
additional theorems are very easily proved.

Th. 11. b (3x)(P v Q) EE (3.v)P v (3jc)£)

Proof: b (x)(~P—p) = (x)^P-(x)~~Q Th. 1
b (3x)(Pv Q) == (3x)Pv (3*)^) Duality Theorem Cor.
and R 1

The proofs of the next two theorems are equally simple, and may be left as
exercises.

T if. 12. b (3x)(3y)P= (ly)(3x)P

*T h. 13. b(3x)(/’*(» ^ (*x)P-(lx)Q

Now that we have the Rule of Replacement, it is desirable to augment our

supply of equivalences. For the next ten theorems (Th. 14 through 1 h. 23) and
the next derived rule (DR 6), we make the blanket assumption that there are no
free occurrences of x in (). We shall give proofs of some of these next results,
and leave others as exercises for the reader:

Tii. 14. b (x){) EE ()

Proof: h (x)Q D Q P 5
\-QDQ ©
H Q => MV DR 1
©
i-(*)e=p

Th. 15.
Th. 16. h (x)(P-Q) = (x)P-Q

Proof: h (x)(P-Q) = (x)P-(x)Q Th. 1

i-(*)(? = <? Th. 14
I-(x)(P‘Q) = (x)P‘Q R.R.
306 A First-Order Function Calculus [Ch. 10
Th. 17. h(3*XFvg)s(3*)Pvg
Proof: h (x)(—P*—Q) = (x)—P-~~Q Th. 16
h (3i)(Pv Q) = (3x)P\ Q ‘ -Duality Theorem Cor.
and R 1

Th. 18. h (3x)(P DQ)= (x)P D Q

Th. 19. h (3x)(C D P) = Q D (3x)P

Th. 20. h(x)(() D P) = <? D (x)P

*Th. 21. h (x)(P D 0 = (3x)P D Q
Tii. 22. h(x)(PvQ) = (x)PvQ

Th. 23. h(3x)(P-Q) = (3x)P-Q

* D R 6. P D Q h (3x)P D <?

The next theorem requires no such restriction on free occurrences of

variables in the formulas it contains:

Th. 24. h(3x)(y)PD (y)(3x)P

1- (x)~P D ~F P 5
h-P D ~(x)~P ®
h P D ~(x)~P ®
1-P D (3x)P df.
!-(«/)/’DP P 5
\-(y)PD(3x)P ®
\-(y)PD(y)(3x)P DR 1
h(3x)(.,)PD(y)(3x)P DR 6

For the next two theorems we make the general assumptions that x and y are
any two individual variables, and F(x) and F(y) are exactly alike except that x
is free in F(x) at all and only those places that y is free in F(y):

Th. 25. h(x)F(x) = (y)F(y)

Proof:h (x)F(x) D F(y) P 5
\-(x)F(x) D (y)F(y) DR 1
We obtain h (y)F(y) D (x)F(x) the same way, and then have
b (x)F(x) = (y)F(y) by®

Th. 26. \-(3x)F(x) = (3y)F(y)

Proof: h(*)~F(*) = («/)~F(t/) . Th. 25
I- ~(x)~F(x) = ~(y)~F\ y)' ®
\-(3x)F(x) = (3y)F(y) df.
Sec. 10.4] RS, and the ‘Natural Deduction’ Techniques 307
Theorems 25 and 26, together with the Rule of Replacement, permit the
interchange of bound variables’—subject to the restrictions mentioned in the
statement of the theorems.
We end the present group of theorems with

T h . 2 7. If P and Q are as in P 5, then I- Q D (3x)P

Proof: I- (x)~P D ~Q P 5
h D ~(r)~P ®
h()D ~(x)~P ®
h Q D (3x)P df.

10.4 RSj and the ‘Natural Deduction’ Techniques

We wish now to show that RS, is an axiomatization of the logic we used in

validating inferences by our list of nineteen Rules of Inference, the strength¬
ened method of Conditional Proof, and the Quantification Rules set forth in
Chapter 4. That RS, contains all the nineteen Rules of Inference has been
shown by our completeness proof for the system of logic based on P 1, P 2, P 3,
and R 1, and our discussion in connection with Th. 0 and DR 0. In a sense, the
Deduction Theorem for RS, corresponds to the strengthened method of
Conditional Proof.
In the presence of R 1, P 5 corresponds to UI. The restrictions on UI are
exactly the same as the restrictions on P 5. We may establish UI as a derived
rule of RS,:

DR 7. If P and Q are as in P 5 then (x)PH Q.

Demonstration: (x)P premiss
(x)P D Q P 5
Q R 1

Again in the presence of R 1, EG follows from P 5 via Th. 27. The restric¬
tions on EG are exactly the same as those on P 5. We may establish EG as a
derived rule of RS,:

DR 8. If P and Q are as in P 5, then Q h (3x)P.

Proof: Q premiss
Q D (3x)P Th. 27
(3 x)P R 1

The rule UG corresponds roughly to R 2, although it seems from one point of

view weaker than R 2, and from another point of view stronger. The second
restriction on UG prevents its use to infer .((j,)(4>g) from Qv, where v is a
variable that occurs free in an assumption within whose scope Qv lies. But
308 A First-Order Function Calculus [Ch. 10

there is no such restriction on R 2. That restriction on UG, however, serves

only to limit its application in the context of a Conditional Proof. In RSt the
same limitation is imposed by placing the restriction on the rule of Conditional
Proof itself, in our statement of the Deduction Theorem—which does not
permit going from P h Q to h P D Q if R 2 is used on a variable occurring free
in P. Hence in the context of RS1 the second restriction on UG applies to R 2,
although it is stated for the Deduction Theorem rather than for R 2 itself. So
UG is not really weaker (more restricted) than R 2.
The rule UG seems stronger than R 2 in that it permits a change of variable
in its application, which R 2 does not. But in the presence of the rest of the
system RSX, UG is easily proved as a derived rule, combining the power of R 2
and Th. 25. The general convention governing <t>/z and in Chapter 4, and the
first restriction on UG (that v does not occur free in (/i)(<1>/a)), are together
equivalent to the conditions stated for Th. 25, that <t>/x and <J>r are exactly alike
except that /z is free in at all and only those places that v is free in Or. We
have F(x) h (x)F(x) directly by R 2, and hence by Th. 25 we have

DR 9 . If F(x) and F(y) are exactly alike except that x is free in F(x) at all
and only those places that y is free in F(t/), then F(x) h (y)F(y).

We can do everything in RSt by DR 9 that we could do earlier by UG.

In Th. 0 and DR 0, and in DR 7, DR 8, DR 9, and the Deduction Theorem,
the system RSt is seen to comprise all of the deductive strength available in our
nineteen Rules of Inference, Quantification Rules UI, EG, UG, and the
strengthened principle of Conditional Proof.
Instead of attempting to formulate a derived rule in RS1 corresponding to
El, we prove a Metatheorem that legitimizes any use of El in a proof subject
to the restrictions involved in its statement.

Metatheorem V. (The legitimacy of El—perhaps better, the elimina-

bility or avoidability of El.)
If there is a sequence of to ffs which would constitute a demonstration that
Pv P2,..., Pn b Q except that some toffs of the sequence are derived from
preceding ones by El (here construed simply as the rule that if R can be derived
in a given context from F(y), then it can be derived in that context from
(3x)F(x)) such that:

1. there are m uses of El of which the ith use is to infer Rt from

(3 xJFjiXj) on the basis of its having been derived from F^yJ, where the
free xi’s in F|(xt) correspond exactly to the free y{'s in F^yfj, and
2. no yi occurs free either in Q, or in any premiss Pp or in any toff of the
sequence that precedes F^yJ, or in the wff R{ derived from Fi(yi), and
3. R 2 is not used to infer (yt)S from any S;

then there is a genuine demonstration that Pv P2,... ,Pn hQ which makes no

use of El. And in the latter (genuine) demonstration there are no uses of R 2 to
Sec. 10.4] RS} and the ‘Natural Deduction’ Techniques 300
quantify variables free in Pv P2,..., Pn which were not quantified by R 2 in the
original sequence or ‘quasi-demonstration \

Proof: In any such quasi-demonstration every wff Fi (ty.) is preceded by a wff

(3o:.)/\(xi) and followed by a wff Except for the m formulas h\(yfj,
F2({/2), • • •» KiilJm) every wff of the indicated sequence is either an axiom or a
Pj or follows from preceding wffs by R 1 or R 2. Suppose there are exactly t
uses of R 2 of which the c/th •
use is to infer S. ( = (.r.
v
)S,)
Ai
from S,A; .

We define S' to be (r-) ... (.r. )(*vi ) • • • )S.. , which is (x-. ) .. . (X: )
,) • • • (xj')(XjJS^. Since S£ b by / — 1 uses of P 5 and R 1, it is obvious
from the existence of the original quasi-demonstration that there exists a
genuine demonstration that

(1) p„P2,...,Pn,S',Si,/-,(«/,),/•;(!/,),...,/-;„(!/,„)I-(,)

which involves no use of either El or R 2. (Where /A(f/;) occurred without

justification in the quasi-demonstration, here it is a premiss; and where was
inferred by R 2, here it is inferred from the premiss S'k bv / — 1 uses of P 5 and
R 1.)
Since no use is made of R 2 in the demonstration that (1), we can use the
Deduction Theorem to obtain

O') .
si,, /',(</,). /',< </,).!/,„ ,) i /■;„(*/„,) => y

Now by successive uses of R 2, Theorems 21 and 20, and the Rule of Replace¬
ment, we obtain from those same premisses first

then

I- (3?3 y

and finally

I- (3xm )PJrj D <>

Since FJijJ is preceded by (3xjFJxj in the original quasi-demonstration,

we have also

and finally by R 1

h Q
310 A First-Order Function Calculus [Ch. 10

Thus we have a demonstration that

(2) Pi. P*.P„. Si,, S'k.S^FM F2(y2),..... Fm_1(ym_1) I- Q

in which R 2 is not used to quantify any variable free in any of the premisses
Sfc2> • • •, S^, F2(y2),..., The foregoing argument can
be used m times to obtain a demonstration that

(m) F1,P2,...,Pn,S'i,S'2,...,S;<hp

in which R 2 is not used to quantify any variable free in any of the premisses
S'ki, Sko, ..., Skt. Hence we can use the Deduction Theorem again to obtain

r-r>..s;,.,Ks;,Dt>
Since preceded in the original quasi-demonstration, we obtain also from
those same premisses

Now by t uses of R 2 we obtain

and finally by R 1

Thus we have a demonstration that

(m + 1) pv *2> • • •*Pn> • • * > Q

in which R 2 is not used to quantify any variable free in any of the premisses
Ski, S;2,..,, S;( t. The foregoing af^ument can be used t times to obtain a
genuine demonstration that

(m+t) PvP29...,Pn\-Q

in which there are no uses of R 2 to quantify variables free in P1, P2,..., Pn

which were not quantified by R 2 in the original quasi-demonstration. This
completes our proof of Metatheorem V.

The preceding discussion suffices to show that our first-order function

calculus RSj is an axiomatization of ftKe logic used in Chapter 4 and the first
three sections of Chapter 5. That it contains no more than was used there can
Sec. 10.5] Normal Forms 311
be shown by giving demonstrations (in the earlier sense of ‘demonstration’) of
the postulates of RSr
The question naturally arises as to how much logic is contained in RSV We
shall present a proof that RS1 is deductively complete in the sense that all
logically true propositions involving quantification of individual variables only
are provable as theorems within it. This sense will be specified more precisely
in Section 10.6. Before proving the completeness of our first-order function
calculus, however, we shall develop some notions and Metatheorems concern¬
ing the subject of ‘Normal Forms.’

10.5 Normal Forms

We begin by defining the notion of ‘prenex normal form’. A wff is in prenex

normal form if and only if it has the structure (^):v,)(^).v2).. . ({)xn)C where
xv x2,..., xn are distinct individual variables, ({fy) is either (3jc,) or (x)), and G
is a quantifier-free wff containing at least one occurrence of each of the xfs.
Some samples of formulas of RSj that are in prenex normal form are

(x) (A'(x))
(3 x)((B'(x))D(C*(xy)))
(y) ((3 z)(~(A3(xyz))))

We now define the term ‘scope’ as follows: If ((i)x)B is a well formed part (i.e.,
a wff that is a part) of a wff A, then the scope of that particular occurrence of
(Qx) in A is the particular occurrence of B which immediately follows ({)*).'*
And we define the phrase ‘initially placed’ in this way: A quantifier is initially
placed in a wff F if it*either stands at the beginning of For is preceded (except
for parentheses) only by other quantifiers, and its scope extends to the end of
F. It should be clear that an equivalent definition of ‘prenex normal form’ can
be stated as: A ivff is in prenex normal form if and only if all its quantifiers are
initially placed, no two quantifiers are on the same variable, and every variable
occurring in a quantifier occurs also at least once within the scof)e of that
quantifier.
We can now state and prove our next Metatheorem.

Metatheorem VI. Given any wff F of RSV a formula P in prenex

normal form can be found such that r F =E F.
Proof: If F is in prenex normal form, P is F and b F = P by ®.
If F is not in prenex normal form, it must either: (a) contain a quantifier that is
not initially placed, or (b) contain two quantifiers on the same variable, or
(c) contain a variable that occurs only in a quantifier.

6This definition provides a more precise formulation of the notion of scope that was discassed
informally in Section 4.4. See pp. 86-87.
312 A First-Order Function Calculus [Ch. 10
case 1. All quantifiers in F are initially placed.
Subcase Li. At least two quantifiers are on tbe.same variable. Here F is
of the form

(QxJ(Qx2)...(Qxi)...(Qxi)...(Qxn)G
If we denote all of F that immediately follows the first (QxJ by A, then xt does
not occur free in A, and by Th. 14 or Th. 15, h (QxJA = A. Hence by the Rule
of Replacement the first of any initially placed quantifiers on the same variable
can be simply dropped out to yield an equivalent formula. So if all quantifiers
of F are initially placed, then there is an equivalent formula all of whose
quantifiers are initially placed and which has at most one quantifier on any
variable.
Subcase l.ii. There is a variable in F that occurs only in a quantifier.
Here we have F of the form

(Qxl)...(Qxi)...(Qxn)G

where x{ does not occur free in G. If we denote all of F that immediately

follows (QxJ by A, then xt does not occur free in A and by Th. 14 or Th. 15,
h (Qx^A = A. Hence by the Rule of Replacement any initially placed quanti¬
fier on a variable which has no occurrence outside of that quantifier may simply
be dropped out to yield an equivalent formula. So if all quantifiers of F are
initially placed, then there is an equivalent formula, all of whose quantifiers are
initially placed and that contains no variables whose only occurrences are in
quantifiers.

case 2. Some quantifiers in F are not initially placed. In this case it is

convenient to write F in unabbreviated form, so the only propositional calculus
symbols in it are • and Next we consider a wff G such that no two
quantifiers in G are on the same variable, no variable in G has both free and
bound occurrences, and b F = G. If F is such a wff, let F be G; otherwise G is
obtained from F by interchanging bound variables according to Theorems 25
and 26 and the Rule of Replacement.
Apart from parentheses, each symbol other than a quantifier that occurs to
the left of a given quantifier (QXj) in a wff G can be counted as a “reason” why
(Qxj) is not initially placed in G, and each symbol that occurs to the right of
(Qx:•) in G and does not lie within its scope can also be counted as a “reason”
why (QXj) is not initially placed in G. It is obvious that any quantifier (Qx^)
occurring in a ivff G is initially placed in G if and only if there are no “reasons”
why it is not initially placed in G.
Given any such wffG, in which the quantifier on is not initially placed and
all quantifiers in G preceding the quantifier on xj are initially placed in G, we
can construct a wff G1 such that: every initially placed quantifier in G is
initially placed in Gx also, h G = Gv and there are fewer “reasons” why the
quantifier on xf is not initially placed in Gj than there were “reasons” why the
Sec. 10.5] Normal Forms 813

quantifier on was not initially placed in G. The construction of Gj proceeds

as follows.
K (Qxj) is the first quantifier in G that is not initially placed in G, then it can
occur in only one of the following wf parts of G: ((Qxf)(A))-(B) or (B)-((Qx-)(A))
or
In the first case, since x- has no free occurrence in B, we replace the wf part
(((?*;( A)) •(#) by (Qx;.)((A)*( B)), using Theorem 16or Theorem 23 and the Rule
of Replacement, to obtain the wff Gv
In the second case, we first use ® and the Rule of Replacement to replace
the wf part (B)u((Qxj)(A)) by ((Qx})(A))'(B), and then replace that as in the
preceding case by ((&,■)(( A)-(B)) to obtain the wffGv
In the third case we obtain the wffG1 using Theorem 7 or Theorem 8 and
the Rule of Replacement to replace the wf part ^((^)(A)) by (A)),
where (@*Xj) is either (xj) or (3.^) according as (Qxf) is (3jq) or (x}).
If the quantifier on xj is not initially placed in Gv we repeat the construction
just described to obtain G2, in which there are still fewer “reasons” why the
quantifier on Xj is not initially placed. Since each wff contains only finitely
many symbols, in any wff, there can be only finitely many “reasons” why any
quantifier in it is not initially placed. Hence the construction described need be
iterated only finitely many times to produce a wff Gk in which the quantifier on
Xj is initially placed.
A new series of such constructions can be used for each quantifier that is not
initially placed in the original wff/1', until finally a wffGK is obtained in which
all quantifiers are initially placed and such that \- F= GK. And to GK there is
an equivalent wff in prenex normal form by Case 1.
This completes the proof of Metatheorem IV.

We define two new terms at this point: In any wff in prenex normal form

the group of quantifiers (Qxx)(Qx2)... ({)xn) is the prefix and the quantifier-free
formula G is the matrix.

EXERCISES

Find a prenex normal form for

1. (x)(3y)(3x)(y)[(z)G(z)-F(x) D (Iz)tllz)-F(z)]
*2. [(x)G(x) v H(y)] = [(«/)(*)*’(!/> *)]
3. (x)[F(x) D (J(x)] D [(x)F(x) D (x)C.(x)]
*4. (y)(3x)F(x, y) O (3x)(y)F(x, y)
5. (xtK3*2)(x1)(3x4)(x5)[P6(x1.xs) 3 (s1)(3x)(3s)ff(x], s)]

A wff in prenex normal form may contain free variables. We wish to be able,
without any loss in generality, to confine our attention to wffs containing no
314 A First-Order Function Calculus [Ch. 10

free variables. These will be referred to as closed toffs (abbreviated cwffs). To

legitimize our concentrating on cwffs, we state and prove the following:
*

Metatheorem VII. For any wff F a cwff G may be found in prenex

normal form such that I- F if and only if b G.

Proof: By Metatheorem VI we can always find the prenex normal form P

for any wff F such that b F = P.

case 1. If P is a cwff, then G is the same as P, and because b F = G, b F if

and only if b G.

case 2. If P contains n free variables xv x2, ..., xni then G is the cwff
(Xi)(x2) ... (xn)P. For if b F, then b P\ and by R 2, b (xn)P,\ and by n uses
of R 2, we have b G. And if b G, that is, b (xl)(x2) ... (xn)P, then since
b (xjixz) ... (xn)P D (x2) ... (xJP by P 5, R 1 gives b (x2) ... (xJP, and by n
uses of P 5 and R 1, we have b P, and hence b F.

We prove now a further result that gives for any wff a. still more specialized
normal form wff that is a theorem if and only if the original wff is a theorem.

Metatheorem VIII. For any wffFa cwffR may be found which is

in prenex normal form and begins with an existential quantifier, such that b F if
and only if b R.

Proof: By Metatheorem VII, for any wff F, there is a cwff G in prenex

normal form such that b F if and only if b G. Let D(t) be a function of one
variable t, such that neither D nor t occurs in G. Then by ®, b G =
{G-[D(t) D D(t)]}-
Hence

b G D {G-[D{t) D D(t)]} by ®
b(3t){C D {G-[D(t) D D(t)]}) by DR 8, (EG)
hGD(3t){G‘[D(()DD(t)]} by Th. 19 and R.R.

We also have

l-{G*[D(0 DD(()]) DG P 2
Kt){{G-[D(l)DD(f)]}DG) R 2
b(3t){G‘[D(t) DD(()]) DG Th. 21 and R.R.

Hence

h(3t){G-[D(i)DD(f)]} = G by ®
See. 10.S]
Normal Forms SIS
Now the prenex normal form of (3f) {G • [ D(t) D D(t)]} is the formula oftypeR
that was desired. It is dosed, it is in prenex normal form, it begins with an
existential quantifier, and 1- F if and only if I- R. Where G is (p*j) (Ox,) ...
(Qxn)G', then

I-(3t){G*[f)(f) D D(t)]) or
h (3tMQ*i)(Qx2) • • • (<?xn)G']-~[D(t)-~D(f)]}

has as its prenex normal form

(3t)(Q*d(Qx.i)... «,\){G'-~~[D(t)-~D(t)].}

by repeated uses of Th. 16 or Th. 23 and the Rule of Replacement.

By definition, any cwff in prenex normal form that begins with an existential
quantifier will be said to be of type R.

EXERCISES

For each of the following wffs, construct a formula of type R that is a theorem if
and only if the given wff is a theorem:
*1. (x)(y)(z){\H(x) v H(y)\ v II(z))
2. F(xvx»...,xn) D G(z)
*3. (3x)G(xyy)vF(z)
4. (3z)F(z,z) D (x)G(x1 y)
5. (x)F(x, y) D G(z)

Next, we present three new definitions, of which the third defines the
‘Skolem normal form’, which is a still more specialized type of well formed
formula, about which we shall establish a further Metatheorem.

Df. The rank of an occurrence of an existential quantifier in a formula of

type R is the number of universal quantifiers which precede that existential
quantifier in the prefix.
Df. The rank of a formula of type R is the greatest of the ranks of the
occurrences of the existential quantifiers of its prefix.
Df. A wffF is in Skolem normal form if and only if it is of type R and of
rank 0.

Metatiieorem IX. Given any wff F, we can find a formula F0 in

Skolem normal form such that h F if and only if b Ft).

Proof; For any wff F we can construct a formula of type R, call it Fv such
that b F if and only if b Fv (If F is already of type R, then F = Fv) If l<\ is of
rank 0, then it is in the desired Skolem normal form. If Fx is of rank k > 0, then
316 A First-Order Function Calculus [Ch. 10

we show how to construct a formula of type R of rank less than fc, call it F2.
such that h Fl if and only if h F2. We thus embark on a process that will yield
formulas F3, F4, f5, .. . all of type R and of decreasing rank. Ultimately, then,
we arrive at a formula F0 of type R and of rank 0 such that h F if and only
ifhF(,

If F\ is not in Skolem normal form, then its rank is greater than 0, that is, Fx
has the structure

(3x1)(3x2) ... (3xn)(y)(QZl)(Qz2)... (QzJG

where {Qzfj(Qz^j, ..., (Qzm) are quantifiers, of which at least one is an

existential quantifier. We now introduce the following notations:

1. B(xv x2,..., xn, y) = df (Qz^iQz.f)... (Qzm)G. This we shall abbreviate

as A.
2. H(y) = df H(Jt],x2y... ,xn, y) where H is any (n + l)-place predicate
variable (with arguments xv x2, ..., xn, y) such that H does not occur
in G.
3. H(t) = df H(Xj, x2t..., xn, t), where t is a variable that does not occur
in G.
4. C = df (3*j)(3At2) ... (3x„){(!/)[(M()z2) ... (QzJG D U(y)} D
W(t)}

First, we want to show that b F, if and only if b C. The two implications are

(i) If b Fv then b C. Here we prove that b i\ D C.

b A D {[A D H(y)] D H(y)} ®

b (y){A D {[A D H(y)] D H(y)}} R2
b (y)A D (y){[A D H(y)] D H(y)} DR 5
b (y)A D {(y)[A D H(y)] D (y)H(y)} Th. 2, ®
b (y)A D {(y)[A D H(y)] D (t)H(t)} Th. 25, R.R.
b (xn){(y)A D {(i/)[A D H(y)] D (t)ff(t)}} R 2
b (3xn)(y)A D (3xJ{(y)[A D H(y)] D (t)H(t)} Th. 4, R 1

b (hxjilxj ... (3xn)(y)A D (S^XSx.,)... (3x„){(y)[A D H(y)] D (t)H(t)}

by n uses of R 2, Th. 4, R 1
b Fj D C df.

Hence, if b Fv then b C.

(ii) If b C, then b F,. Here we must .assume that b C and derive b Fv

Sec. 10.5] • vi lira

H (3xt)... (3in){(!/)[(^,)... (p-jc D U(y)] D (t)ll(t)) C

b(3x,)...(3xj{(i/p(x,,...,.vni!/) D
i/(xi, -..,xB, i/)] D(()%,..,V|)) df.
h(3i2)... .D
W(*'i.•'„.'/)] 3(t)//(x',.vB,0} El

h (i/)[B(x;, (/) D //(x'„ ..., (/)] D ..., x'n,t) n uses of El

Now we substitute B(x.... x', y) for the predicate variable II{x\.x', r/)7

b (y)[B(x\, D B(x'„.... x'„,y)} D (t)B(x’v ■■■> <■ 0

h , x'n, y) D B(x',.x'n, I/)] D {y)B(x\,x:;,!/) Th. 25. R.R.
t- (3//){ [«(*;,..., x;, y) d B(x\,.... a-;, , v)] £>
(</)«(*',.•<„!/)} Th. 18, R.R.
I- [B(xi, jf) 3 B(x'„ y)l d (y)«(.v'„.... x'„, i/) El
I- B(xi, ...,x'„,y)D B( x\,x'„, ij) ®
\-(y)B{x\,...,x'n,y) 111
b(3xB)(!/)/J(xj,...,A„,!,) Dr 8, (EG)

h (3x,)... (3xn)(y)B(xvxn,y) ti uses of EG

b (3xt)... (3x„)(;/)(();,)... (QzJC df.
df.

Hence if h C, then h i\.

C is not in prenex normal form. Its prcnex normal form is constructed as

follows, using the notation ()* defined as follows:

If ({ht) is (a,), then (^)is (3s,)

If (Qzf) is (3a,). then (()%) is (zt)

C= (3a,) ... (3A„){(y)[(^,)(^2)... (QzjC D ll(y)] D (l)H(t)}

Since H{y) does not contain any free ~,'s,

b (Qzx) ... (QzJG 3 U(y) = ((>*-,) • • ■ ((>%,)[£ 3 //(»)]

by m uses of Th. 18 or Th. 21. Hence by the Rule of Replacement,

C=(3xt) ... (3a:nm(V'zl) •.. (<?%)[<:: 3 U(y)\ 3 (t)H(t)}

7The rule governing such substitution (rule of substitution for functional variables) is difficult
and complicated to express. Since this is our only use of it, we shall not state it here. A discussion
of alternative formulations can be found in A. Church, Introduction to Mathematical Logic /,
Princeton University Press, Princeton, N.J., 195b, pp. 193 ff., 289, and 297.
318 A First-Order Function Calculus [Clt. 10

Since (t)H(t) does not contain any free z/s,

I- {(yxe**,) • • • => H(y)] D (t)H(t)) = ... •

{(3</)(pZl)... (gzm)[[G D H(y)] D (it)H(t)]}

by m + 1 uses of Th. 18 or Th. 21. Hence by the Rule of Replacement,

c= (3xt)... (3xn)(3yXQzJ ... (QzJ[[G D H(y)] D (t)H(t)]

Since t is not free in G D H(y), by Th. 20,

h [[G D H(y)] D (t)H(t)] = (t)[[G D H(y)] D H(t)],

hence by the Rule of Replacement,

C = (3xx)... (3xn)(3y)(QZl)... (QzJ(t)[[G D H(y)] D H(t)]

The right-hand formula is the formula F2 of type R, such that h F if and only
if h F2. Now the rank of F2 is lower than the rank of Fv for their prefixes are the
same except that F2 has an additional universal quantifier in the final position.
This does not affect the rank. Another exception is that one of the universal
quantifiers ((y)) which preceded an existential quantifier has been replaced by
an existential quantifier, which lowers the rank by one. If F2 has rank 0 then it
is the Skolem normal form of F. If it is not, then the argument can be repeated
to obtain a formula F3 of type R of still lower rank. Hence we can ultimately
arrive at a formula of type R of rank 0, which is the desired Skolem normal
form.

EXERCISES

Find a Skolem normal form for

*1. (3*)(y)(3x)[F(x#y)vG(*)]
2. (x)F(x) D \{3y)G(y) D (z)H(z)]
*3. (x)(3y)F(x, y)
4. (x)(y)(3z){3x){[[F(x)-G(y)]v H(z)] D I(w)}
5* (3xj)(x2)(x3)(3x^)F(x^, x2, Xft, xf)

10.6 Completeness of

We turn now to the problem of proving our first-order fimction calculus

complete. There are several different sefises of the term ‘complete' which must
be distinguished. A very strong kind of completeness is possessed by a system
all of whose well formed formulas are provable as theorems. But this kind of
completeness amounts to inconsistency, and is of course undesirable. Our first
Sec. 10.6]

Metatheorem, which established the consistency of our system, proved it to

lack this kind of completeness. A somewhat weaker kind' of completeness is
possessed by a system each of whose wffs is either provable as a theorem or else
has its denial provable as a theorem; that is, for every wff F either h F or I-F.
This kind of completeness is not desirable either, for on their normal interpre¬
tations, each of the following wffs

Nr (3x)F(x) D (x)F(x)
N2: (3x)(3y){F(x)-F(y)-[G(x) = ~~G(y)]} D (x)F(x)
N3: (3x)(3y)(lz){F(x)-F(y)-F(z)-[G(xj= ~G(i/)]-
[H(y) = *H(z)]-[l(x) = —/(«)]} ^ (x)F(x) .

asserts (respectively) that there is at most one individual, that there are at most
two individuals, that there are at most three individuals,... But if we want our
logical system to be applicable to any possible nonempty universe, regardless
of the exact number of individuals it contains, there is no Ni such that we want
either Nj or ^'Ni to be provable as theorems. A different kind of completeness
was proved for the propositional calculus presented in Chapter 8. That logistic
system was proved to be complete in the sense that every wff that (on its
normal interpretation) is a truth table tautology is provable as a theorem in the
system. That kind of completeness is admirably suited for a propositional
calculus, but it will not do for a functional calculus. A first-order function
calculus that did not have the wff

(3x)(y)F{x9 y) D (y)(3x)F(x, y)

provable as a theorem w ithin it would be woefully unsatisfactory and ‘incom¬

plete’, even though the formula in question is not a truth table tautology. The
reason for regarding as unsatisfactory a logic in which the given expression is
a wff but not provable is that (on its normal interpretation) the formula
expresses a logical truth—where a logical truth’ is a proposition that is true in
(or of) every possible nonempty universe. 1 he kind of completeness desired for
our first-order function calculus can be expressed loosely by saying that every
logical truth that can be expressed in the system is provable as a theorem
within it. This notion, however, must be expressed much more precisely before
it can be used in any proof of completeness.8

8 The following discussion and proof of completeness is an adaptation to RSj of the completeness
proof for an alternative first-order function calculus given by Leon Henkin in his article The
Completeness of the First-Order Functional Calculus’, Vie Journal of Symbolic Logic, vol. 14
(1949), pp. 159-166. Professor Henkin simplified the published proof and he communicated it to
this writer in the spring of 1951. It is reproduced here with his kind permission.
The first proof of completeness for a first-order function calculus was published by Kurt Godel,
‘Die Vollstandigkeit der Axiomc der logischen Funktionenkalkuls, Monalshefte fiir Mathematik
und Physik, vol. 37 (1930), pp. 349-360.
320 A First-Order Function Calculus [Ch. 10

Instead of continuing to speak of ‘possible universes’, we shall talk about

models, where a model is any (nonempty) collection, of elements each of which
is thought of as an individual. And instead of speaking of our system of logic as
being ‘applied’ to a ‘possible universe’, we shall speak of a model as constitut¬
ing an ‘interpretation’ of our formal system. This last term will be given a
precise meaning presently. With any given set of individuals that may be
intended to serve as a model, we assume that we are given also the attributes
that may belong to the individual elements, and the relations (binary, ternary,
and so on) that may hold between (or among) them. Now we define ‘interpre¬
tation of a wff S with respect to a given model ’ as an assignment of meanings
such that

1. To each propositional symbol in S we assign a truth value, either T or F.

2. To each individual constant in S, and to each variable with free occur¬
rence in S, we assign an element of the model.
3. To each predicate symbol in S we assign an attribute or a binary or
ternary or n-ary relation according as it is of degree (has right super¬
script) 1, 2, 3, or n.

This notion of interpretation of a wff with respect to a given model is incom¬

plete, however, because no mention is made of what to do with the logical
symbols —, •, and the quantifier symbol (x). Our intended or normal interpre¬
tation of a wffS with respect to a given model is defined as an interpretation of
S with respect to that model, subject to the following conditions:

(a) Any well formed part — W is assigned the truth value T or F according
as the well formed part W is assigned the truth value F or T.
(b) Any well formed part of the form X* Y is assigned the truth value T if
and only if both X and Y are assigned the truth value T.
(e) Any well formed part (x)R is assigned the truth value T if and only if
R is assigned the value T, regardless of which element of the model is
assigned to all free occurrences of x in R.
(d) Any well formed part x2,. .., xn) is assigned the truth value T if
and only if the elements of the model assigned to xv x2,..., xn, in that
order, stand to each other in the n-ary relation assigned to Pn.

So far we have defined ‘interpretation’ and ‘normal interpretation’ only for a

wjf with respect to a specified model. An interpretation of a system with respect
to a given model is an assignment of meanings that provides interpretations for
all wffs of the system, and a normal interpretation of a system with respect to
a given model is an assignment of meanings which provides normal interpre¬
tations for all wffs of the system.
There are many interpretations, and even many normal interpretations of a
formal system with respect to a given model. On some of these the value T may
be assigned a wff that on another interpretation is assigned the truth value F.
Sec. 10.6]
Completeness of RSt 321
It is obvious, however, that the postulates of our first-order function calculus
will be assigned the truth value T on any normal interpretation of the system
with respect to any model. A particular normal interpretation of a wff S with
respect to a specified model will be said to satisfy S with respect to that model
if by it S is assigned the truth value T. A wff S will be said to be sctisjiable with
respect to a model if there is a normal interpretation of S with respect to that
model which assigns the truth value T to S. And a w ff S will be said to be
satisfiable if there is at least one model and a normal interpretation of S with
respect to it which assigns the truth value T to S. ‘Valid’ is another term that it
is important to introduce in this connection. A wff S is valid with respect to a
model if every normal interpretation of S with respect to that motjel assigns the
truth value T to S. It follows immediately from the definitions that the wffS is
valid with respect to a given model if and only if every normal interpretation
of S with respect to that model satisfies S. Finally, we define the term ‘valid’ by
itself: a wff S is valid if it is valid with respect to every model. It follows
immediately from the definitions that for any wff S, either S is valid or — S is
satisfiaUe.
In the notion of a model, we have a precise formulation of our informal
notion of a possible universe, and normal interpretation is what ought to be
signified by the looser phrase ‘intended application’. Consequently, the validity
of a wff is a precise notion that serves the purpose suggested by the phrase
‘logically true’. The completeness we wish to prove for our first-order function
calculus RSj can now be defined precisely. A first-order function calculus is
complete if and only if every valid wff is provable in it as a theorem.
As for a propositional calculus, so for a first-order function calculus: the
converse of completeness is analyticity. An analytic first-order function calcu¬
lus is one all of whose theorems are valid. It is obvious that all the postulates of
RS1 are valid, and that any wffs derived from them by R 1 and R 2 are valid
also. Clearly RSt is analytic. The consistency of RSj is a consequence of its
analyticity, just as the consistency of R.S. was shown to be a consequence of the
analyticity of R.S. in Chapter 8.
In establishing the completeness of RSa, it will be sufficient to confine our
attention to wffs containing no free individual variables. It is easily demon¬
strated that to every wff S which contains free individual variables, there
corresponds a wff Sc (called the closure of S) containing no free individual
variables, such that b S if and only if b Sr, and such that S is valid if and only if
Sc is valid. Where S contains exactly n free variables, say xv x2,..., xn_v xn (in
the order of their occurrence), then S' will be (Xj)(x2) ... (xn_i)(xn)S. Now if
b S then b (xn )S by R 2, and another application of R 2 gives b fo-^xJS.
Finally, we have b Sr by n uses of R 2. Conversely, if b S', that is, b (xj)(r2)...
(x Mx )S then applying R 1 to it and the instance of P 5 which is (xt)(x2)...
(Cows 3 M (*„->„)* will give h (*2)... (V-i)WS; so by » uses of
R 1 and P 5, we obtain I- S. If S is valid, every normal interpretation of S with
respect to any model assigns to S the value T regardless of which elements of
the model its free individual variables Xj, x2> • • • taken to denote. From
322 A First-Order Function Calculus [Ch. 10

this it follows that (xn )S is valid, and (xn_ )S is valid,..., and finally that Sc
is valid. On the other hand, if Sc is valid, (x2)(x3)... (xH )S is assigned the value T,
regardless of which element of the model x1 is taken to denote, from which it
follows that (x3)... (xn )S is assigned the value T regardless of which elements xx
and x2 are taken to denote, from which it follows ... finally that S is assigned
the value T regardless of which elements of the model the free individual
variables xv x2,... ,xn are taken to denote, which means that S is valid. Hence
with no real loss in generality we can confine our attention to closed tvffs (tuffs
containing no free variables), which we shall refer to as ctvffs.
To prove our system complete, it is useful to introduce the notion of a set of
cwffs. We shall use the capital Greek letters Gamma and Lambda, with and
without subscripts, to denote sets of cwffs of RSt. We shall also make use of
braces to represent sets of cwffs: for example, the set whose only member is the
cwffS will be written {S}, the set whose only two members are the cwffs Sx
and S2, will be written {SA, S2), and so on. To assert that there is a demonstra¬
tion of the wff(? from the closed well formed premisses P,, P2,..., Pn, we may
write either

Pv P2.P„ h Q

or

{PvP2.?„}!-<?
If F is a set of cwffs and S is a cwff not belonging to F, then the new set
containing all members of V and S will be written as F U {S}. To assert that
there is a demonstration of Q from premisses belonging to V U {S} we may
write either

F U {S}\-Q or r, S\~Q

An alternative statement of the Deduction Theorem makes use of this new

notation:
jm

Iff, P\~Q then Fh PDQ

If there is a demonstration of the wff Q from the premisses Pv P2,..., Pn, then
there is, of course, a demonstration of Q from any set of cwffs F which contains
{Pv P2,..., Pn } as a subset (which will be written {Pv P2,..., Pn } C T), no
matter what other cwffs F may contain. The set F may even be infinite: There
is nothing in our proof of the Deduction Theorem that requires any revision to
accommodate this possibility. Of course, in any demonstration of Q from an
infinite set of premisses only a finite number of them will actually be used, for
a demonstration is always a finite sequence of wffs.
We defined a first-order function calculus to be complete if and only if every
Sec. 10.6] Completeness of RSt 329
valid wff is provable in it as a theorem, and then showed that the first-order
function calculus RSj is complete if and only if every valid cwff is provable in
it as a theorem. To say that every valid cwff is provable in RSt as a theorem is,
by transposition, to say that for any cwff S, if S is not a theorem, then S is not
valid.
In defining the terms ‘valid' and ‘satisfiable’, we remarked that for any wffS,
either S is valid or is satisfiable. Hence to say that S is not valid is to say
that —S is satisfiable. And so we can say that RSt is complete if and only if, for
any cwff S, if S is not a theorem, then — S is satisfiable. We can establish this
result by introducing a characteristic (p such that both

(1) If S is a cwff th'&t is not a theorem, then S is a cwff having characteristic

qp; and
(2) If S is a cwff having characteristic <p, then — S is satisfiable;

for from (1) and (2) the completeness of RSj will follow bv a Hypothetical
Syllogism.
We first define two attributes of sets of cwffs. A set A of cuffs is inconsistent
if A I—-PmP, otherwise A is consistent. It is obvious that any wff can be
inferred from an inconsistent set, for where S is any wff* b (~~P'P) v S by
®, which is logically equivalent to — PmP\- ~~(~~P*P) D S. Since b — (^PmP)
by ®, we have by R i that — P-P b S, whence A b S where A is any inconsist¬
ent set of cwffs and S is any wff whatever. Now the characteristic (j> referred to
above can be defined as follows: A cwff' S has characteristic q> if and only if
{—S} is a consistent set.
And we demonstrate (1) above as the following:

Lemma: If S is a’ cwff that is not a theorem, then {—S} is a consistent set.

Proof: We show that for any cw ff S, if {—S} is not a consistent set then S
is a theorem. Let S be any cwjf such that {} is not consistent. If —^S) is
not consistent, then it is inconsistent, which means that

~Sh -~P-P

Because S is a cwff, we can use the Deduction Theorem to obtain

h D ^P-P

from which

b ~~(*P-P) 3-S

follows by ®. We already have

324 A First-Order Function Calculus [Ch. 10
by ®, so by R 1 we obtain

which by ® gives

bS

which is the desired result.

Instead of demonstrating (2) above, simply, as

If S is a cwff such that {~~S} is a consistent set, then — S is satisfiable,

we prove a somewhat more general result, for the statement of which the term
‘simultaneously satisfiable' must be defined. The definition is this: A set A of
cwffs of a formal system is simultaneously satisfiable if and only if there is a
model and a normal interpretation of all formulas of A with respect to that
model which satisfies all the wffs of A. Now we state and prove

Me tat he orem X. Any consistent set of cwffs of RSX is simultaneously

satisfiable.

Proof: Our proof of the simultaneous satisfiability of any arbitrary consistent

set A of cwffs of RSj proceeds by constructing a larger set of civffs which
includes A as a subset, and then proving that this larger set is simultaneously
satisfiable. This will, of course, establish the desired result for A.
We begin by considering a set of distinct symbols uv u2, . • • which are
different from any of those contained in RS ,. We denote by ‘RS/’ the formal
system obtained from RSt by adding to it the symbols uv u2, w3, ... which
serve as individual constants. The only difference between RS! and RS/ is that
the latter has a larger number of individual symbols. In every other respect the
two systems are identically the same. We assume—for definiteness—that all
cwffs of RS/ have been ordered in a sequence, so we can refer to them as the
first, the second, the third, and so oiT(there are many alternative ways of
accomplishing this, of course).
Starting with any arbitrary consistent set A of cwffs of RSP we enlarge it by
adding cwffs to it one at a time. The first cwff to be added is D p/,
where (3x)Pi is the first cwff of RS/, which begins with an existential
quantifier; and Q* is the result of replacing all free occurrences of x in pt by
where is the first ut which has no occurrence in Qv We may call the
resulting set ‘A/, where Ax = df A U {(3x)p! D Q*}, and can easily
prove it to be consistent. For if A^ were inconsistent, we should have A,
(3x)Pj D Q* I-P'P, and by the Deduction Theorem, A h ['BxJpj 3
P/] D —P*P. Since A contains no occurrences of nor any free occurrence
of any individual variables, in the derivation of [(Bx)^ D p/] D —P-P from
Sac. 10.6]
Completeness of RS, 329
A we could replace all occurrences of u, | by the free individual variable x to
obtain A h D p,] D From that we could successively obtain

A b (*){[(3*)P,. 3 Pi] 3 by R 2
A b (3*)[(3*)p, 3 ^,] D ~~P'P by Th. 21 and R.R.
A h [(3*)p, 3 (3x)pt] D ~P*P by Th. 19 and R.R.
A h [(3*)p, D (3x)Pj] by ®
and finally
A h -f’-P

which is contrary to the assumption that A is consistent* lienee Aj is a

consistent set.
The next cwjf to be added is (3r)()2 D Q2*, where (3x)^)2 is the second cwjf
of RS/ which begins with an existential quantifier, and ^)2* is the result of
replacing all free occurrences of ji* in by tii r where uj t is the first u- with no
occurrence in either ()2 or A,. We now define A2 as A] U {(3x)()2 D <j2*},
and it can be proved to be consistent in exactly the same way that A Y was. We
continue to add cwffs one at a time: to Ai ^ j we add (3 jc)^ D {)* where (1x)(Ji
is the jth cwjf of RS/ which begins with an existential quantifier, and ()/ is the
result of replacing all free occurrences of x in ()■ by f#4., where u, is the first ut
with no occurrence either in Q- or in A^_ Each A • is consistent, each of them
contains A as a subset, and each set is contained in all of the later sets of the
sequence.
Next, we define the set Aw as the sum or union of all sets in the preceding
sequence. (The existence of the set Au, is guaranteed by the Union Axiom. See
page 184.) Any cwjf of RS/ is a member of Au, provided that it is a member of
any of the sets A* of the sequence. It is clear that A C Au, and easily proved
that A^ is consistent. For if A^ were inconsistent, there would be a demon¬
stration of ~~P*P from a finite number of formulas of Aw. Each of these would
have to occur in A or some A, for the first time and in every A, thereafter, for
Au contains no formulas except those which occur in A or At or A2 or A3 or
.. ., and we have remarked that A C A x C A2 C Aa C . . . Since there can be
only a finite number of premisses used in any demonstration, there would have
to be a first set A* which contains all the premisses actually used in the alleged
proof that A^ b ~~PmP. But that proof, then, would equally well be a proof that
A-I—-P'P, contrary to the fact that each A* is consistent. Hence Aw is
consistent.
Now we proceed to enlarge Aw until we obtain a maximal consistent set T
of cwffs of RS,*. Our method of ‘enlarging A* is by successively adding to it
each cwff of RS,* that is consistent with the formulas of Au, together with
those previously added to it. The resulting set F will be consistent, since any
demonstration that F I-P‘P would make use of only a finite number of
formulas of I\ the last of which (in the order of their addition) could not have
been added if the directions for adding formulas had been followed. From our
definition of F, it follows for any cwff of RS,* which is not in F, that its
326 A First-Order Function Calculus [Ch. 10

addition to F would result in an inconsistent set; hence we call T a maximal

consistent set. . .
It is obvious that T contains all the postulates of RS^ that are closed and the
closures of those that are not, and also all theorems of RS^ that are closed,
because if S is a ctvff such that F b S, then S is contained in F. For since T is
maximal, the addition to it of any cwff S not already contained in it results in
an inconsistent set, that is, for any cwff S not belonging to F, we have T,
S b ~~P'P. From this, by the Deduction Theorem, we have F b S D ^P*P.
Now if we were also to have F b S, then by R 1 we should have T b ~~P*P,
contrary to the fact that F is coasistent. It should also be observed that where S
is any cwff of RS*t either F b S or T b ^S. This follows from the fact that
every cwff S of RSt* either is or is not contained in T. If S is in F, then
obvioasly, F b S. While if S is not in F, then F, S b ~~P*P and by the Deduction
Theorem, FbS D ~P*P. From this, by ®, we have V b —(— P-P) D —S, and
since we also have by ® that F I-(— P*P), by R 1 we have T I-S. From the
preceding remarks it follows that for any cwff S, either S belongs to F or
belongs to F.
To show that the formulas of T, and hence certainly those of A, are simulta¬
neously satisfiable, we take as our model the set I of all individual constants of
RS *, regarded as individuals. These are all the individual constants of RSV a,
h, c9 av bv cv a2> &2, c2,..., together with all the symbols uv u2, u3, n4.
Our interpretation of the cwffs of F consists of the following assignment of
meanings:

1. To each propositional symbol of RS* we assign the truth value T if it

belongs to F; otherwise it is assigned the truth value F.
2. To each individual constant of RST* (considered as a symbol in an
interpreted system) we assign itself (considered as an individual which
belongs to the model I). There are no free individual variables in any wff
of F, since these are all cwffs; consequently no attention need be paid to
them here.
3. To each predicate symbol Pn we assign the n-ary relation (or attribute,
if n = 1) among n individuals av a2,..., an of I of having their names,
properly parenthesized, adjoined in that order to Pn, constitute a cwff of
F, or Pn(al9 a2,... ,an)c F.

The interpretation is completed and made normal by adding these conditions:

(a) Any cwff is assigned the truth value T or F according as the cwff S
is assigned the truth value F or T.
(b) Any cwffX • Y is assigned the truth value T if both X and Y are assigned
the truth value T; otherwise it is' assigned F.
(c) Any cwff (x)R is assigned the truth value T if the truth value T is
assigned to every result of replacing all free occurrences of x in R by an
individual constant of RS*; otherwise it is assigned F.
Sec. 10.6] Completeness of RS1 327
(d) Every cwff Pn(av a2,..., an) is assigned the truth value T if the
elements of the model assigned to av «2,..., an, in that order, stand to
each other in the n-ary relation assigned to PM, otherwise it is as¬
signed F.

We now can prove that all wffs of I' are satisfied on this interpretation. What
we shall actually prove is the stronger result that every cwff S of RSt* is
assigned the truth value T or F according as S belongs or does not belong to F.
Our proof is by strong induction on the number of symbols in S, counting each
propositional symbol, each cwff of the form Pn(av a2,..., an), each occurrence
of a universal quantifier (x), and each occurrence of • and — as a single symbol.
(a) In case n = 1, S is either a propositional symbol or a cwff of the form
Pn(ai, a2,..., an). If S is a propositional symbol, by paragraph 1 of the original
assignment of meanings, S is assigned the truth value T or F according as it is
contained or not contained in I". If S is the cwff Pn(al,a2.an) then by
paragraph 3 and condition (d), S is assigned the truth value T or F according as
it is contained or not contained in V.
(/3) Here we assume that every cwff of RSj* which contains fewer than k
symbols is assigned the truth value T or F according as it is contained or not
contained in T. Now we consider any cwff S of RS/ that contains exactly
k > 1 symbols. Here S must be either — W or X• V or (x)R.

case 1. S is —W.
If — W is in T, then since V is a maximal consistent set, W is not in F. Here,
by the /3-case assumption, since W contains fewer than k symbols, W is assigned
the truth value F. By condition (a), — W must be assigned T. Hence if S is in
r, it is assigned the truth value T.
If — W is not in T, then W is in F. Here W is assigned T, so by condition (a),
—W must be assigned F. Hence if S is not in l\ it is assigned the truth value F.

case 2. S is X-Y.
If X • y is in F, then by ® T I- X and V b y, so X and y are both in F. Since
each contains fewer than k symbols, by the /3-case assumption they are both
assigned the truth value T, and by condition (b), S is also.
If X* y is not in F, then not both X and y can be in V (for if they were, then
by ® T h X • y and X • Y would have to be contained in F), so not both X and y
are assigned the truth value T, and by condition (b), S is assigned the truth
value F.

case 3. S is (x)R.
If (x)R is in F then T h (x)R, and where Q is any result of replacing all free
occurrences of x in R by an individual constant, h \ x)R 3 ^ by P 5, so by R 1,
F h Q, Since F is a maximal consistent set, every such Q is contained in T, and
since each of them contains fewer than k symbols, they are all assigned the
328 A First-Order Function Calculus [Ch. 10

truth value T by the /3-case assumption. Now by condition (c), (x)R is assigned
T also. Hence if (x)R is in F, then (x)R is assigned.the truth value T.
If (x)R is not in F, then ~~(x)R is in F, whence F b ^(x)R, and by Th. 7 and
the Rule of Replacement, F h (3x)-^R. Since F is a maximal consistent set, it
must contain (3x)~~R. But (3x)-— R is a cwff of RS,* which begins with an
existential quantifier, and so the formula (3 x)~~R D ~~R* must have been
added to some to form A;-—where ~~R* is the result of replacing all free
occurrences of x in ^R by u{ . Since the cwff (3x)~~R D —R* belongs to
some such A;., it must belong to F. Hence, by R 1, F b — R*, so ^R* belongs to
F also. Since F is consistent, R* does not belong to T, and since it contains
fewer than k symbols, by the /i-case assumption, it is assigned the truth value F.
Hence by condition (c), (x)R is assigned the truth value F.

This completes the induction, and proves that all formulas of F, and hence all
those of A, are simultaneously satisfiable. Metatheorem X, together with the
discussion which preceded it, completes our proof of

Metatheohem XI. RSj is complete.

In the preceding section we showed that RS1 is logically equivalent to the

natural deduction techniques developed in Chapter 4. Hence our proof of
Metatheorem XI validates the claim made at the end of Chapter 4 that the
natural deduction techniques there developed ‘permit the demonstration of all
logically true propositions constructed out of truth-functional connectives and
the quantifications of individual variables'.

10.7 RS1 with Identity

We obtain a first-order calculus with identity by adding to RSa a new binary
relation symbol, the predicate constant 72’, and additional axioms governing
the new symbol to impose those conditions on it that serve to characterize
identity. Our metalanguage expression for wffs containing the new identity
symbol, for example, ‘I2(acy)’, will be*x = ;/’. Although it is customary to use
two or more special axioms or axiom schemata for identity, a single axiom
schema is sufficient.9 It is

P 6. (y){Fy = (3x)[(x = y)-Fx]}

From this new postulate, in the presence of the other postulates, rules, theo¬
rems, derived rules, and metatheorems of RSX already at hand, we can derive
the following results concerning identity.

9Thc idea of using this one axiom schema is credited to Hao Wang by W. V. O. Quine in Set
Theory and Its Logic, Harvard University Press, Cambridge, Mass., 1963, p. 13.
Sec. 10.7] RSj with Identity 329
Theorem I — 1: h [(* = y)*Fx] D Fy

Proof: (x = y)-Fx\-(y){Fy = (3x)[(x = y)-Fx]} P 6

(x = y)'Fxh Fy = (3x)[(x = y)‘Fx] DR 7
(x = y)’Fx\-(3x)[(x = y)-Fx] D Fy ©
(x = y)'Fx\- (x = y)*Fx premiss
(X = y)-Fx\-(3x)\(x = y)’Fx] DR 8
(x = y)‘Fx h Fy R 1
b [(* = y)"Fx] D Fy D.T.

Theorem 1-2: \- y = y *

Proof: We begin with the instance of P 6 in which *Fz' is ‘~~(z = i/)’:

(y)[~(.V = y) = (3x)[(x = y)’~~(x = !/)]] P 6
~(y = y) = (3x)[(x = y)—(x = y)] DR 7
(y = y)= ~(3x)[(x = y)‘—(x = y)j ©
~[(* = y)-~(* = !/)] ©
x

R 2
II

II

~(3x)[(x = y)’~~(x = y)] Th. 8 & R.R.

y = y ©

Theorems 1-1 and 1-2 are the usual axioms for identity.

Theorem I - 3: h (x)(i/)[(a* = y) D (y = .v)J

Proof: We begin with an instance of Th. 1-1 in which bz is z = x :

[(* = y)’(x = ■ »')] 3 (y = •»’) Th. 1-1
(x = x) D [(x = !/) D (y = x)j ©
JC = JC Th. 1-2
(x = y) => («/ = x) R 1
(y)l(x = y) ^ (y = x)] R 2
(x)(«/)[(x = y)o(y = x)] R 2

Theorem 1-4: t-(x)(y)(z){[(x = !/)*(*/ = *)] D (* = *)}

Proof: We begin with an instance of Th. 1-1 in which ‘Ft/ is *^(u = *)’
n
7T

Th. 1-1
II
I
X
II
H

I
II

(x = y) O [~(x = z)D ~(y = *)] ©

(x = y)D [(y =z)D(x = *)] ©
[(x = y)*(y = Z)1 D (* = z) ©
(z){[(* = y)*(!/ = 2)] D (*=*)) R 2
R 2
(yX*){[(* = »)■(» = z)l D (* = *)>
(x)(y)(z){[(x = y)-(y = z)]D(x=z)} R 2
330 A First-Order Function Calculus [Ch. 10

Theorem 1-5: h (x = y) = (y=x)

(*)(«/)[(* = y) => («/ = *)] Th. 1-3

(«/)[(* = y) ^ (y = *)] DR 7
(x = y)D(y = x) DR 7
(*)[(* = z) D (z = *)] DR 9
(m)(z)[(to = z) D (z = in)] DR 9
(*)[(«/ =z)D(z = y)] DR 7
(y = x) D (x = y) DR 7
[(* = y) => (y = *)]•[(«/ =x)D(x = y)] ®
df.
III
K

H
II

II

The four ‘Principles of Identity’ stated in Section 5.4 can be established here as
derived rules of inference.

DR I - 1. x = y\- y = x
Proof: x = y premiss
y =* Th. 1-5 & R.R.

DR I - 2. Fx, y = x\- Fy

Proof: y =x premiss
x = y DR l-l
Fx premiss
(x = y)‘Fx ®
[(* = y)’Fx1 Fy Th. 1-1
Fy R 1

DR I - 3. Fx, -Fy h —(y = x)

Proof: [(x = y)'Fx] D Fy Th. 1-1

(* = !/) 3 (Fx D Fy) ®
Fx premiss
—Fy premiss
—(Fx D Fy) ®
-(x = y) ®
~(y = x) Th. 1-5 fit R.R.

The fourth principle, p H v = v, is already given by Theorem 1-2.

A more complicated logistic system is. required for the formalization of the
logical principles involved in appraising arguments that concern attributes of
attributes or of relations, or relation's _ among attributes and relations, or the
notions of all or some attributes or relations. Such logistic systems are generally
Sec. 10.8] First-Order Logic Including ZF Set Theory 331

called extended function calculi, and if they are consistent, they are demon¬
strably incomplete.10 But these more advanced parts of symbolic logic lie
beyond the scope of the present book.

10.8 First-Order Logic Including ZF Set Theory

The ZF Set Theory described in Chapter 7 can be formalized as a first-order

theory. The only special mathematical symbol required is that for set mem¬
bership. Or rather, since the first-order theory, as a logistic system, is uninter¬
preted, we can specify that the new and special symbol is the binary relation
symbol £, which we intend subsequently to interpret as set membership.
Indeed, if our purpose is simply to develop set theory, and thereby mathe¬
matics, in our first-order logic, our notations can be much more parsimonious.
Let us begin again to develop a first-order logic with formalized Set Theory as
our goal. Instead of the elaborate panoply of primitives enumerated in Section
10.1, let us begin with the one two-place predicate symbol £, and a denumer-
ably infinite list of individual variables

u, v, w, x, y, z, ul9 vv wv xv t/j, zl9 u2, v2i ...

Of course, we shall need the four logical symbols —% *, (, and ). Those are all
the primitive or undefined terms we need.
We can define the additional logical symbols v, D, =, and 3, for our object
logic, as in Section 10.1. Then we can define wff for our system recursively as
follows:

(a) If x and y are individual symbols, then x £ y is a wff.

(ft) 1. If F is a wff, then —(F) is a wff.
2. If F and G are wffs, then (F)-(G) is a wff.
3 if p is a wff and x is an individual variable, then (x)(t) is a wff.

We assume the familiar five postulates of RSt together with the axiom
schema for identity, P 6 of Section 10.7. We assume the two familiar rules of
inference R 1 and R 2 of Section 10.1, and the definition of demonstration. It is
easy enough to define = in terms of £ as in ZF-1, and then to define * and*
as obvious abbreviations. All of the ZF axioms can be symbolized m this
system. Then, we can either add the ZF axioms as additional axioms or else
take them all together as premisses to every argument whose conclusion is
specifically set theoretic or mathematical.

"“’T”1' N.gd.ndI,mo B.Mminm ™»b, tand in I. M. Copi .nd

JTS*Ufc *• M«»n-. P,«. ** .m
332 A First-Order Function Calculus [Ch. 10

So far so good.
There is a new paradox lurking, however. We have 6een, in our completeness
proof for RSj, how that system is satisfied by a denumerable model. On the
assumption that the ZF Set Theory is consistent, it, too, has a denumerable
model. This was proved by Thoralf Skolem. But the ZF system contains both an
axiom of infinity and a power set axiom, which permits proof that there are
nondenumerable sets. Here is the Skolem Paradox: A system which has a
denumerable model contains a theorem that asserts the existence of nonde¬
numerable sets.
How shall we understand this situation? It cannot be said that there is
general agreement about what has gone wrong. Here, for what they may be
worth, are two of the proposed resolutions.
One thing that can be said is that the set membership relation simply has not
been (cannot be?) completely described or defined formally. So the binary re¬
lation in our model, which serves as interpretation for £, is not set member¬
ship but something only formally parallel to it. In short, this proposed resolu¬
tion says that the possibility of a denumerable model for a nondenumerable
theory has shown only that our model is ‘nonstandard’, and is not what we were
trying to talk about or specify.
Another way out of the Skolem Paradox is based on the realization that
equinumerousness or equivalence of sets is established only by actually pro¬
ducing a function that is or effects a one-to-one correspondence between them.
It may be that the function needed to map a set onto its power set simply
cannot be defined in the system. If the set is denumerable, then its power set
can be proved nondenumerable in the system only because the function that
would enumerate its members does not exist in the system. It is as though the
power set, seen from outside the system, can be proved equinumerqus with the
denumerable set, but to specify the one-to-one correspondence requires a
different, more powerful system than ZF. This proposed resolution has more
adherents than one might have thought. Intuitively, sets are the size they are,
quite independently of what functions are available to map one onto another.
This view seems to me to be an extreme kind of operationalism—which
appears to suggest that if you canno^t measure something, then it does not exist!
It is heartening to be able to end a textbook that has no prerequisites with a
problem that has not yet been solved.
 APPENDIX

Incompleteness of the
Nineteen Rules
The nineteen Rules of Inference presented in Sections 3.1 and 3.2 of Chapter
3 are incompletewhich is to say that there are truth functionally valid
arguments whose validity cannot he proved using only those nineteen Rules. To
discuss and establish this incompleteness, it is useful to introduce the notion of
a characteristic that is ‘hereditary with respect to a set of rules of inference’.
We offer this definition: A characteristic {\> is hereditary with respect to a set of
rules of inference if and only if whenever <1> belongs to one or more statements,
it also belongs to every statement deduced from them by means of those Rules
of Inference. For example, truth is a characteristic that is hereditary with
respect to the nineteen Rules of Inference presented in the first two sections of
Chapter 3. As we have already said, any conclusion must be true if it can be
deduced from true premisses bv means of our nineteen Rules of Inference.
Indeed, we should not want to use any rules of inference with respect to which
truth was not hereditary.
Nowt, to prove that a set of rules of inference is incomplete, we must find a
characteristic d> and a valid argument a such that

(1) <l> is hereditary with respect to the set of rules of inference; and
(2) <I> belongs to the premisses of (x but not to the conclusion of (x.

The characteristic truth is hereditary with respect to any set of rules of

inference in which we may be seriously interested and, therefore, satisfies
condition (1) above. But where « is a valid argument, it follows immediately
from our definition of validity that truth can never satisfy condition (2) above.
Hence to prove the incompleteness of our nineteen Rules we must find a
characteristic other than truth that is hereditary with respect to our nineteen
Rules, and can belong to the premisses but not to the conclusion of some valid
argument a.

iThe following proof of incompleteness was devised by my friend Professor Leo Simons of the
University of Cincinnati. See Leo Simons, ‘Logic Without Tautologies’, Notre Dame Journal of
Formal Logic, vol. 15 (1974), pp. 41J-431, and More Logics Without Tautologies’, ibid., vol. 19
(1978), pp. 543-557. For some discussion of the matter, see Hubert L. Armstrong, ‘A Question
about Completeness’, ibid., vol. 17 (1976), pp. 295-296, and Paul J. Campbell, ‘An Answer to
Armstrong’s Question about Incompleteness in Copi\ ibid., vol. 18 (1977), pp. 262-264.

333
 To obtain such a characteristic, we introduce a three-element model in terms
of which the symbols in our nineteen Rules can be, interpreted. The three
elements are the numbers 0,1, and 2, which play roles analogous to those of the
truth values true (T) and false (F) introduced in Chapter 2. Every statement
will have one of the three elements of the model assigned to it, and it will be
said to assume, take on, or have one of the three values 0, 1, or 2. Just as in
Chapter 2 the statement variables p, </, r,..., were allowed to range over the
two truth values T and F, so here we allow the statement variables p, q, r,... to
range over the three values 0, 1, and 2.
The five symbols *•’, V’, ‘ D’, and that occur in our nineteen Rules
can be redefined for (or in terms of) our three-element model by the following
three-valued tables:

p ~P P 9 p-q Pv9 P -> 9 P=9

0 2 0 0 0 0 0 0
1 1 0 1 1 0 1 1
2 0 0 2 2 0 2 2
1 0 1 0 0 1
1 1 1 1 1 1
1 2 2 1 1 1
2 0 2 0 0 2
2 1 2 1 0 1
2 2 2 2 0 0

Alternative (but equivalent) analytical definitions can be given as follows,

where ‘min(x, y)' denotes the minimum of the numbers x and i/, and ‘max(x, y)’
denotes the maximum of the numbers x and y.

—p =2 — p
p-q = max(p, q)
pvq = min(p, q)
p D q = min(2 — p, q)
p=q = max(min(2 — p, q), min(2 — q, p))

The desired characteristic 0 that is hereditary with respect to our nineteen

Rules of Inference is the characteristic of having the value 0. To prove that this
characteristic is hereditary with respect to the nineteen Rules, it will suffice to
show that it is hereditary with respect to each of the nineteen Rules. This can
be shown for each rule by means of a three-valued table. For example, that
having the value 0 is hereditary with respect to Modus Ponens, can be seen by
examining the table above, which defines the value of p D q as a function of
the values of p’ and of ‘q’. The two premisses p’ and ‘p D q both have the
value 0 only in the first row, and there the conclusion tq1 has the value 0 also.
Examining the same table shows that having the value 0 is hereditary also for
Simplification, Conjunction, and Addition. Filling in additional columns for
 Incompleteness of the Nineteen Rules 33S

anf Wi? S!*0W that havin8 the value 0 is hereditary with respect to
0 us o ns an disjunctive Syllogism also. That it is hereditary with respect
to Hypothetical Syllogism can be shown by the following table:

P 9 r p Da t/Dr n ~
0 0 0 0 0 0
0 0 1 0 1 1
0 0 2 0 2 2
0 1 0 1 0 0
0 1 1 1 1 1
0 1 2 1 1 ' 2
0 2 0 2 0 0
0 2 1 2 0 1
0 2 2 2 0 2
1 0 0 0 0 0
1 0 1 0 ] 1
1 0 2 0 2 1
1 1 0 I 0 0
1 1 1 1 1 1
1 1 2 1 1 1
1 2 0 1 0 0
1 2 1 1 0 1
1 2 2 1 0 1
2 0 0 0 0 0
2 0 1 0 1 0
2 0 2 0 2 0
2 1 0 0 0 0
2 1 1 0 1 0
2 1 2 0 1 0
2 2 0 0 0 0
2 2 1 0 0 0
2 2 2 0 0 0

Only in the first, tenth, nineteenth, twenty-second, twenty-fifth, twenty-sixth,

and twenty-seventh rows do the two premisses ‘p D q' and *q D r both have
the value 0, and in each of them the conclusion p D r has the value 0 also.
Even larger tables would be ilecdcd to show that having the value 0 is
hereditary with respect to Constructive Dilemma and Destructive Dilemma.
Larger tables can be constructed, but they are not absolutely necessary,
because the alternative analytical definitions on page 334 can be used to show
that having the value 0 is hereditary with respect to the Dilemmas, as on
page 336.
When we construct three-valued tables to verify that having the value 0 is
hereditary with respect to replacement of statements by their logical equiva¬
lents, we notice that although the biconditionals themselves need not have the
336 Appendix A

value 0, the expressions flanking the equivalence sign necessarily have the same
value. For example, in the table appropriate to. the first of De Morgan’s
Theorems, * '

the equivalent expressions (p*r/)* and p v ~~ q' have the same value in
every row even though the statement of their equivalence fails to possess the
value 0 in rows two, four, and five. It should be obvious, however, that having
the value 0 is hereditary with respect to the replacement of all or part of any
statement by any other statement that is equivalent to the part replaced.
Alternative proofs that having the value 0 is hereditary with respect to the
nineteen Rules make use of our analytical definitions of the logical symbols. For
example, that having the value 0 is hereditary with respect to Constructive
Dilemma can be shown by the following argument. By assumption,
‘(p D q)'(r D s)’ and ‘p v r both have the value 0. Hence both p D q and
V D s’ have the value 0, so either p = 2 or q = 0 and either r = 2 or $ = 0.
Since p v r has the value 0, either p = 0 or r = 0. If p = 0, then p ^ 2,
whence q = 0, and if r = 0, then r ^ 2 whence s = 0. Therefore, either q = 0
or s = 0 whence ‘q v s’ has the value 0, which was to be shown.
Once it has been established that the characteristic of having the value 0 is
hereditary with respect to the nineteen Rules, to prove the incompleteness of
those Rules, one need only exhibit a valid argument whse premisses have the
value 0, but whose conclusion does not have the value 0. Such an argument is

ADB
.‘.A D (A*B)

whose validity is easily established by a truth table. Where ‘A’ has the value 1
and ‘B’ has the value 0, the premiss ‘A D B’ has the value 1 D 0 = 0 but the
conclusion ‘A D (A#B)’ has the value 1 D (1*0) = 1 D 1 = 1. Therefore, the
nineteen Rules of Inference are incomplete.
 APPENDIX

Normal Forms and

In this Appendix we develop an alternative method for recognizing valid

truth-functional arguments. Given a complicated statement form, one can
often find a simpler one logically equivalent to it, to which the given form can
be ‘reduced’ by algebraic operations. In treating the logic* of compound
statements and statement forms from the algebraic point of view, it is conven¬
ient to use a different notation for denial or negation. Here we symbolize the
negation of an expression by drawing a bar above it. In this notation, the
familiar truths of De Morgan are symbolized as p-q = (p v q) and
p v q = (p9q). All of the statement forms treated here are truth-functional, so
their status as tautologous, contradictory, or contingent remains unchanged
when any part is replaced by an expression that is logically equivalent to the
part replaced. Thus p v p is a tautology and remains one when p is replaced in
it by p, since p and p are logically equivalent by the principle of Double
Negation.
Since p •(</•/*) and (p-q)-r are logically equivalent, they may be indifferently
written as p • q • r. Similarly, p v (q v r) and (p v q) v r may be written as pvqvr.
The logical equivalences involved are principles of Association. The conven¬
tion of dropping unnecessary parentheses permits us to enunciate generalized
statements of De Morgan’s Theorems as pl 'p.j’Pa* • ■ • *Pn = (P1 v Viv Pa v
... v pw)andp1vp2vp;jv...v^ EE (fVfVPa- • • • -Pn)*The two tautologies
[p-(q v r)] EE [(p-q) v (p-r)] and [p v (q-r)\ = [(p v q)-(p v r)] are principles of
Distribution. Generalized principles of Distribution may be expressed as

[p-(qi v q2 v ... v qn)\~ \\p-qj v (p-q2) v ... v (p-qn)]

and
[pv(<h-<l2- ■■■ •</.)]= l(PVfll)*(PV</2)* -(PV'/»)]

The capital Greek letters pi and sigma are often used to express generalized
logical products (conjunctions) and generalized logical sums (disjunctions). We
define

n Pi=df(pi*P2‘ ••• 'P.)

i= l
337
and
n

2>i =df(Pivp2v---vPrt) '

i=1

Given these notations we can express our generalized De Morgan Theorems as

n n n n

n>-s^ and i>=n^

i=l i-1 i=l i=l

and the generalized Distribution principles as

p‘S^=and pvii^=n^v^)
i=l i= l i=l i=l

It is convenient to have four other principles of logical equivalence available

for purposes of algebraic manipulation and transformation.
First, the principles of Commutation expressed as

(p-</) = (<7*p) and (pv<i) = (qvp)

Second, the principles of Tautology, which, stated as

p = (pvp) and p = (p'p)

assure us that any statement, wherever it may occur, is replaceable by the

disjunction (conjunction) both of whose disjuncts (conjuncts) are the same as the
given statement, and vice versa.
Third is the principle that any statement p is logically equivalent to the
conjunction of itself with any tautology of the form q v q9 that is

p=fp-{q\q)]

Our fourth and final principle is the logical equivalence

p=[pv(q-q)]

which permits us to interchange p and p v (q*q) wherever either of them

occurs.
It is clear that by invoking the defining equivalences

(pDq)=(p\q) and. [p = q)=[(p-q)v(p-q)}

 Normal Forms and Boolean Expansions 339

material conditionals and biconditionals can always be eliminated in favor of,

or expressed in terms of, conjunctions, disjunctions, and negations. Moreover,
by repeated uses of De Morgan’s Theorems and Double Negation, any formula
can be replaced by a logically equivalent one in which no negation symbol
applies to a compound part. Thus

P'[(q v r)*(s*t)]

is transformed by repeated applications of De Morgan’s principles and Double

Negation as follows:

pv[(qvr)-(s-t)]

P v [(q v r) v (s • £)]
p v [(q v r) v (sv t)]

pv [(qvr) v(sv f)]

which becomes, by the principle of Association,

pvqvrvsvt

In this expression, the negation symbol is applied only to the single variables q
and s.
The principle of Distribution permits us to change any given formula into an
equivalent one in which conjunction symbols occur—if at all—only between
single variables or their negations, or else into a different equivalent form in
which disjunction symbols occur—if at all—only between single variables or
their negations. Thus the expression

(1) (P v t/)*(rvs)

in which the conjunction symbol occurs between disjunctions, reduces by

repeated applications of the principles of Distribution, Commutation, and
Association into

(2) . {p‘r)v (q‘r)\(p's)\(q-s)

in which conjunction symbols connect single variables only. And the expression

(3) (p-q)v(r-s)

in which the disjunction symbol occurs between conjunctions, reduces by

repeated applications of the other parts of the principles of Distribution,
Commutation, and Association into

(4) (pvr)-(qfvr)'(pvi‘)'(qfvs). '

in which disjunction symbols connect single variables only.

Two ‘normal forms’ can be defined. A formula is in conjunctive normal form
when, in addition to statement variables, it contains no symbols other than
those for conjunction, disjunction, and negation; negation symbols apply only
to single variables; and no disjunct is a conjunction, that is, disjunction symbols
occur only between single variables or their negations. Thus (1) and (4) above
are in conjunctive normal form. A formula is in disjunctive normal form when,
in addition to statement variables, it contains no symbols other than those for
conjunction, disjunction, and negation: negation symbols apply only to single
variables; and no conjunct is a disjunction, that is, conjunction symbols occur
only between single variables or their negations. Thus (2) and (3) above are in
disjunctive normal form.
Let us examine the moderately complicated formula (p D q) D (p = q) to
see what gain in perspicuity is achieved by reducing it to disjunctive normal
form. By replacing the material implication symbols by their definitions, the
initial expression is reduced first to

(p v 7) ^ (p = <7)
and then to

(pvq)v(p = q)

Replacing the material equivalence symbol by its definition, we obtain

(pv<7)v[(p.<j)v(p^)]

Applying De Morgan’s principle, we obtain

(p-q)v[(p-q)v(p-q)]

which, by the principles of Double Negation and Association, becomes

(p-q) v(p-q) v(p-q)

Although it is already in disjunctive normal form, it can be further simplified by

the following transformations. First, we apply the principle of Tautology to
replace the first disjunct by its ‘double’, obtaining

(P’</) v (pmq) v (p*g) v (puq)

 Normal Forms and Boolean Expansions 341

Then we rearrange the terms by simply commuting or interchanging the

second and third disjuncts, to get

(p-q) v (p-q) v (p-q) v (p-q)

Using the principle of Distribution on the first pair of disjuncts, we obtain

[p-(q v q)]v (p*q) v (p-q)

and using it and Commutation on the two right-hand disjuncts, we get

[p*(f/ v q)] v [(p v /!)•</]

Now, by the principle that the conjunction of any statement p with a tautology
of the form qvq is logically equivalent to the statement itself, we obtain first

pv[(pvp)-q\
and finally
pvq

which is logically equivalent to, but much simpler than the form with which
we began.1
The term ‘normal form’ is sometimes reserved for more specific types of
expressions. These more specific types are also called ‘Boolean Expansions’ or
‘Boolean normal forms’, after the British logician George Boole (1815-1864). A
formula containing the variables p, </, r,... is said to be in disjunctive Boolean
normal form, or a disjunctive Boolean Expansion, when it is in disjunctive
normal form, when every disjunct contains exactly one occurrence of every
variable (either the variable or its negation), when the variables occur in
alphabetical order in each disjunct, and when no two disjuncts are the same.
The disjunctive Boolean Expansion of pvq is formed by first replacing p by
p*(qv q) and q by (p v p)*q to obtain the equivalent expression p• (q v q) v
(p v p)-c/, then using the Disjunction and Association rules to obtain (pmq) v
(p9q) v (puq) v (pmq), and finally cancelling out any repetitions of the same
disjunct by the rules of Commutation and Tautology, which results in
(p*q)v(puq)v(p9q). The disjunctive Boolean Expansion of any noncontra¬
dictory formula can be obtained by the same general method.
A formula containing the variables p, q, r, ... is said to be in conjunctive

1 An application of normal forms to electrical circuits (parallel connections are represented by

V and series connections are represented by *■*) has been made by Claude E. Shannon, ‘A
Symbolic Analysis of Relay and Switching Circuits’, IVansactums of the American Institute of
Electrical Engineers, vol. 57 (1938). pp. 713-723.
An informal account of the above and several other interesting applications can be found in
John E. Pfeiffer’s article ‘Symbolic Logic’, in Scientific American, vol. 183, no. 6 (December 1950).
See also Martin Gardner, ‘Logic Machines’, ibid., vol. 186, no. 3 (March 1952).
Boolean normal form, or a conjunctive Boolean Expansion, provided that it is
in conjunctive normal form, every conjunct contains exactly one occurrence of
every variable (either the variable or its negation), the variables occur in
alphabetical order in each conjunct, and no two conjuncts are the same. The
conjunctive Boolean Expansion of puq is formed by first replacing p by
p v (qmq) andq by (p*p) v q to obtain [p v (qmq)]m[(pmp) v q], Then, using the
Distribution and Association rules to obtain (pv q)*(pv q)*(pv q)9(pv q), and,
finally, cancelling out any repetition of the same conjunct by the rules of
Commutation and Tautology results in (p v q) • (p v q) *(p v q). The conjunctive
Boolean Expansion of any nontautologous formula is easily obtained by the
same general method and can be used for deciding whether the original
formula is contradictory.
Any conjunctive Boolean Expansion containing n variables and 2n conjuncts
is reducible to an explicit contradiction. Thus the conjunctive Boolean Expan¬
sion (p v q) *(p v q) *(p v q) *(p v q) is equivalent by the rule of Distribution to
[p v (q •q)] • [p v (q •</)], which is equivalent to the explicit contradiction p *p. It
is equally obvious that the conjunctive Boolean Expansion of any contradiction
will contain 2n conjuncts if it involves n variables. Hence the general rule is
that a statement form containing n variables is contradictory if and only if it
has a conjunctive Boolean Expansion that contains 2n conjuncts.
The negation of a conjunctive Boolean Expansion is reducible by repeated
applications of De Morgan's Theorem and Double Negation to a logically
equivalent disjunctive Boolean Expansion that contains the same variables and
has the same number of disjuncts as the original conjunctive Boolean Expansion
has conjuncts. (Thus the conjunctive Boolean Expansion (pvqvr)•
(pvqvT)'(pvqvr) has its negation (pvqvr)'(p\q\r)*(pvqvr) logi¬
cally equivalent to the disjunctive Boolean Expansion (p-g-r) v (p•{/•**) v
(p • g •?).) Since the negation of a contradiction is a tautology, and a conjunctive
Boolean Expansion containing n variables is a contradiction if and only if it
contains 2n conjuncts, it follows that a disjunctive Boolean Expansion contain¬
ing n variables is a tautology if and only if it contains 2n disjuncts.
That such a disjunctive Boolean Expansion must be tautologous can perhaps
be seen more clearly by the following considerations. In the first place, since
we are concerned with truth-funcffonal compounds only, to speak of substi¬
tuting statements for the statement variables of a statement form is equivalent
to speaking of an assignment of truth values to the variables of that statement
form. Now just as each row of a truth table with n initial or guide columns
represents a different assignment of truth values to the n statement variables
involved, so each disjunct of the Boolean Expansion represents a different'
assignment of truth values to the variables they contain. And just as the 2n rows
of a truth table having n initial or guide columns represent all possible
assignments of truth values to its variables, so the 2W disjuncts of a disjunctive
Boolean Expansion represent all. possible assignments of truth values to its
variables. Since the 2n disjuncts represent all possible assignments of truth
values to its variables, at least one of them must be true. And since a disjunctive
 Normal Forms and Boolean Expansions 343

Boolean Expansion asserts only that at least one of its disjuncts is true, any
disjunctive Boolean Expansion containing n variables and 2n disjuncts is
tautologous. This point is made in somewhat different terms in Chapter 8,
Section 8.6, and again in Chapter 9, Section 9.2.
It was pointed out in Chapter 2 that a truth-functional argument is valid if
and only if its corresponding conditional statement (whose antecedent is the
conjunction of the argument’s premisses and whose consequent is the argu¬
ment s conclusion) is a tautology. Since counting the number of disjuncts of its
disjunctive Boolean Expansion permits us to decide whether or not a given
form is a tautology, this provides us with an alternative method of deciding the
validity of arguments. Thus the argument form p v qt ~**p . \ ^is proved valid
by constructing the disjunctive Boolean Expansion of its corresponding condi¬
tional [(p vq)-p] D q, and observing that the number of its disjuncts is 22.
Since the negation of a tautology is a contradiction, an argument is valid if
and only if the negation of its corresponding conditional is a contradiction.
Hence another method of deciding the validity of an argument is to form the
conjunctive Boolean Expansion of the negation of its corresponding conditional
and count the number of its conjuncts. If it contains n distinct variables and has
2n conjuncts, then the argument is valid; otherwise it is invalid.

EXERCISES

For each of the following, find as simple an equivalent as you can:

*1. p = p 9. pv(q-r-q)
2. pv(q-p) *10. p D [(p D q) D </l
3. pv(p-q) 11. (pvq)-(qvr)-(pvr)-r
4. q -(p v q) 12. (p-q) v (q-r)v (p-r) v r
*5. p = (q-p) 13. p • {pD [</•(</ => r)\)
6. p-(p v q) 14. p v {p D [q v (q D r)l>_=
7. p-q-p *15. {pvq) 3 [Ip 3 q) D (</ v*/)l
8. (pv q)*q 16. (p-q) D [(p ^ q) ^ (p*q) 1
17. [p'[(q-r) vp]}v {(p-f/)-[r D (p-F)|) v (p-[p D (p-q^F)]}
18. (p-q) v (p-q-r) v (p-q) v (p-q-r) v (p-q)v (p-q-r) v (p-q-r)
 APPENDIX

c
The Ramified Theory
of Types1
The simple theory of types, which was expounded briefly on pages 153-154,
suffices to eliminate such logical paradoxes as that of the alleged attribute
impredicable. But there is another kind of paradox or contradiction that is not
prevented by the simple theory of types. One example of this other kind is the
paradox of the liar, which was discussed on page 214. These two kinds of
paradoxes were first explicitly distinguished by F. P. Ramsey in 1926.2 Since
then those of the first kind have been known as ‘logical paradoxes’, the second
kind either as ‘epistemological paradoxes’ or, more usually, as ‘semantical
paradoxes’.
A singularly clear example of a semantical paradox is due to Kurt Grelling.3
The Grelling paradox can be stated informally as follows. Some words desig¬
nate attributes that are exemplified by the words themselves, for example,
‘English’ is an English word, and ‘short’ is a short word. Other words designate
attributes not exemplified by the words themselves, for example, ‘French’ is not
a French word, and ‘long’ is not a long word. We shall use the word ‘hetero-
logical’ to designate the attribute a word has just in case it designates an
attribute not exemplified by the word itself. Thus the words ‘French’ and ‘long’
are both heterological. The contradiction now arises when we ask if the word
‘heterological’ is heterological. If it is heterological, then it designates an
attribute that it does not exemplify, and since it designates heterological, it is
not, itself, heterological. But if it is not heterological, then the attribute that it
designates is exemplified by itself, so that it is heterological. Here the contra¬
diction is made explicit: if it is, then it isn’t, and if it isn’t, then it is.
A somewhat more formal derivation of the Grelling paradox, patterned after

'The reader who desires a more comprehensive discussion of this topic should consult Chapter
3 of I. M. Copi, The Theory of Logical Types, Routledge & Kegan Paul Ltd., London, 1971,
pp. 76-114.
2F. P. Ramsey, ‘The Foundations of Mathematics’, Proceedings of the London Mathematical
Society, 2nd series, vol. 25 (1926), pp. 338-384. Reprinted in The Foundations of Mathematics,
Harcourt Brace Jovanovich, New York, and Kegan Paul, Trench, Trubner & Co. Ltd., London,
1931, pp. 1-61.
3K. Grelling and L. Nelson, ‘Bemerkursgen jpu den Paradoxien von Russell und Burali-Forti’,
Abhandlung der Fries*schen Schulet n.s. vol. 2 (1907-1908), pp. 300-324.

344
 The Rammed Theory of Types 34s

Ramsey s version of it,* is the following. Where Des’ designates the name
relation, so that s designates <p’ is symbolized as sDes</>’, we begin with the
definition

Ilet(s) _ df (3</>):sDes<;> vsDes^ = ^ 4, = ,

CASE 1
1. Het(‘Het’) D KSs^^Het’Desc/) • Tlet’Desi// = ^ $ = c/> • ~~<>(‘Het’)
2. D :*Het'Des^> • Tlet’Desi^ == ^ = </>• —</,(‘Het’)
3.
D :%flet’Des$• ‘Ilet’DesIIcl El Ilet = </>• -~<;>(‘Ilet’)
4. 3 :‘Het’DesHet EE Ilet = <>-^<;>(4 5Het’) «
5. D :Het = (/>• ^(‘Hct’)
6. 3 :-Het(‘IIef)

CASE 2
1. —Het(‘ilet’ D :(</>) -1‘Ilet’ 1 )es</>• ‘Het’Desi// 7= ^4, = —^(‘1 let')]
2. 3 :‘IIet’DesHet D ~[‘Het’Dea£ ~ v,
4' = 1 let • ~1 Iet(‘I let’)]
3. 3 :‘IIet’Des^ = ^ 4, = Ilet- D •Ilet('Het’)
4. D :‘I Iet’Desy = ^ y = Met (assuming ‘Ilet’ univocal)
5. D :Het(‘llet’)

Hence

llet(TLet’) ~ -Het(‘HeO

which is a contradiction.
The derivation of fhe preceding contradic tion did not violate any restrictions
imposed by the simple theory of logical types, so that theory does not prevent
such contradictions. Contradictions of this second kind can be eliminated,
however, by the adoption of a more complicated version of the theory of
logical types, known as the ‘ramified' (or ‘branching') theory of logical types.
Now, according to the simple theory of logical types, all entities are divided
into different logical types, of which the lowest contains all individuals, the
next lowest consists of all attributes of individuals (designated by functions of
individuals), the next, of all attributes of attributes of individuals (designated by
functions of functions of individuals), and so on. There are also hierarchies of
relations among individuals and attributes, and of attributes of relations, but
they need not be considered here. Where we use lower-case letters to denote
individuals and capital letters to designate attributes, then we can represent

4Op cit., p. 358 and pp. 369-372; reprint p. 27 and pp. 42-46.
5Where writing the psi as a subscript to the equivalence symbol abbreviates the universal
quantification of the equivalence with respect to psi, so that ‘sDesi! ~ ^ ^ = </>’ is an abbreviation
for 4(t//)[sDcs^ = $ = 4>]\
this hierarchy as follows, where the subscript attached to a function indicates
its proper level in the hierarchy: . .

type 3: F3, G3, H3,...

type 2: F2, G2, H2,...
type 1: Fv Gx, Hx> ...
type 0: a, ft, c,, x, y, z

Only a function of type 1 can significantly be predicated of an individual, and

in general, a function of type i can be significantly predicated of a function of
type / if and only if i = y + 1.
The preceding hierarchy presents a rough ‘picture' of the simple theory of
logical types. Now the ramified theory divides each type above level zero into
a further hierarchy. Thus all functions of type 1, which may be predicated of
individuals, are divided into different orders in the following fashion.
All propositional functions of type 1 that contain either no quantifiers or else
quantifiers on individual variables only are said to be first-order functions. For
example, Fx(x) and (y)[Fx(y) D G1(x)] are first-order functions of type 1. First-
order functions will have a left superscript 1 attached to indicate their position
in the hierarchy of orders. Thus all first-order functions of type 1 may be listed
as XFV 1GV 1H1, ... Next, all propositional functions of type 1 that contain
quantifiers on first-order functions, but no quantifiers on any other functions,
are second-order functions. Examples of second-order functions of type 1 are
WI1^*) = 1Fx(a)] and (31G1)(3y)[1G1(t/) D 1H1(x)]. Second-order functions
will have a left superscript 2 attached, and all second-order functions of type 1
may be listed as 2F1, 2GV 2H1,... In general, an nth order function of type 1
will contain quantifiers on functions of order n — 1, but will have no quanti¬
fiers on functions of order m where m > n.
The ramified theory of logical types can be described compendiously by
means of the following two-dimensional array:

Order 1 Order 2 Order 3

type 3: %, kfg,... ; 2F3, 2G3, 2H3. %, 3G3, 3H3..

type 2: % XG2, lH2,... ; %, 2G2, 2H2. %, 3G2> 3H2..
type 1: %, 'Gv 'Hv ... ; %, 2GV . %, 3Gj, 3HV ....;

Just as the simple hierarchy of types prevents us from speaking about all
functions or attributes and permits lis to speak only about all functions of
individuals or all functions of functions of individuals, and so forth, so the
 The RamHM Theory of Typos 347

hierarchy of orders prevents us from speaking about all functions or attributes

of a given type. Such ranking permits us to speak only about all first-order
functions of a given type, or all second-order functions of a given type, and so
forth. Thus we cannot, according to the ramified theory of types, say that Bob
has all of A1 s good qualities, which would ordinarily be symbolized as

miCUFJ-FM D FJjb)}

Instead, we can say that Bob has all of Al’s good first-order qualities, sym¬
bolized as

(%){[xG2(%y%(a)] D %(b)}

or that Bob has all of Al’s good second-order qualities, symbolized as

D%m
or that Bob has all of Al's good nth order qualities, for some specified n. It
should be noticed that the attribute of having all of Al\s first-order attributes,
symbolized

=> %(*)}
is a second-order attribute.
The preceding formulation of the ramified theory of logical types is crude,
but it suffices to eliminate the Grelling paradox. In Case 1, the step from 2 to
3 is rejected on the grounds that the function ‘Met’ is of higher order than the
function variable V/A and may not be instantiated in its place. In Case 2, the
step from 1 to 2 is likewise rejected on the grounds that the function ‘Het’ is
of higher order than the function variable ‘</>* and may not be instantiated in its
place.
Another feature of the ramified theory of logical types is that it divides
propositions into a hierarchy of propositions of different orders. Just as any
function (of any type) is a first-order function if it makes no reference to any
totality of functions of that type (i.e., if it contains no quantifier on any function
variable of that type), so any proposition is said to be a first-order proposition if
it makes no reference to any totality of propositions (i.e., if it contains no
quantifier on any propositional variable). In general, a proposition is of order
n -f 1 if it contains a quantifier on a propositional variable of order n but
contains no quantifier on any propositional variable of order m where m > n.®
The restriction here is that we can never refer to all propositions, but only to
all propositions of this or that specified order. Thus we cannot say that ‘None
6^ principle Mathematics propositions are also divided into different orders on the basis of
differences in the orders of the functions they contain. We shall ignore that subtlety in the present
exposition, however.
of the propositions uttered by Smith tends to incriminate him’, which we might
partially symbolize as

(p)[(Smith utters p) D ^(p tends to incriminate Smith)]

but can say instead either that ‘None of the first-order propositions uttered by
Smith tends to incriminate him’, or that ‘None of the second-order propositions
uttered by Smith tends to incriminate him’, and so forth. We would partially
symbolize the second of these alternative propositions as

(2p)[(Smith utters 2p) D ~-(2p tends to incriminate Smith)]

That proposition contains a quantifier on a propositional variable of order 2,

and is, therefore, a proposition of order 3.
By divijding propositions into different orders and permitting reference only
to propositions of some specified order or orders, the ramified theory of logical
types effectively prevents the paradox of the liar. Any version of that paradox
involves the assertion that all propositions satisfying a certain condition are
false, where the assertion, itself, is a proposition that satisfies that condition.
(The condition in question might be being asserted by the speaker, or being
written in a specified location, or being uttered by a Cretan, etc.) The paradox
is fully explicit when the assertion in question is the only proposition that
satisfies the specified condition, for in that case, if it is true, then it is false, and
if it is false, then it is true. Such a contradiction is prevented by the ramified
theory of logical types in the following way. The assertion can refer only to all
propositions of a certain order, so it can assert only for some specified n that all
nth order propositions satisfying a certain condition are false. But no paradox
can arise here, because the italicized sentence expresses a proposition of order
n + 1, and even if it satisfies the specified condition, it is not an nth order
proposition and, therefore, does not assert its own falsehood.
The ramified theory of logical types, including both the hierarchy of types
and the hierarchy of orders, was recommended by Russell and Whitehead not
only for ‘its ability to solve certain contradictions’, but also for having ‘a certain
consonance with common sense which makes it inherently credible’.7 They
claimed to have deduced the theory from what they called the ‘vicious-circle
principle’, of which one of their formulations was: ‘Whatever involves all of a
collection must not be one of the collection’. But that principle is not obvious,
and their putative deduction is not convincing.8 The chief merit of the theory

7 A. N. Whitehead and B. Russell, ‘Introduction to the first edition, Principia Mathematical

chap. 11.
8In his ‘Foundations of Mathematics', F. P. Ramsey remarked on the ‘rather sloppy way’ in
which the type theory was deduced from the vicious-circle principle’ (p. 356, p. 24 of reprint),
and challenged the validity of that principle itself, holding that .. we may refer to a man as the
tallest in a group, thus identifying him bv.means of a totality of which he is himself a member
without there being any vicious circle' (p. 368,* p. 41 of reprint).
 The Ramified Theory of Types 348

of logical types, at least of its ramified version, would seem to be its prevention
of the paradoxes.
There are difficulties as well as advantages with the ramified theory of
logical types. One of these concerns the notion of identity. The usual definition
of identity (of individuals) is

(x = y) =df(F1)[F1(x) = Fl(y)]

from which definition all of the usual attributes of the identity relation can be
deduced. But that definition violates the ramified theory of logical types, since
reference is made in it to all functions of type 1. Were we to replace it by the
definition

(x = y) = di('F1)['I<\(x)='Fl(,j)]

which provides that x and y are identical if they have all of their first-order
attributes in common, the possibility arises that x and y might be identical (in
this sense) and still have different second-order attributes. For any n, defining
identity as the sharing of all attributes of order n would permit “identical”
individuals to differ with respect to attributes of order m where m > n. If we
accept the restrictions of the ramified theory of logical types, then we cannot
define the identity relation, and even if we accept the relation of identity as
primitive or undefined, we could not state all the rules for its usage.
Other disadvantages of the ramified theory of logical types are more tech¬
nical, and will merely be mentioned. Mathematicians wish to establish their
theorems for all functions (of numbers), but they cannot do so if they are bound
by the hierarchy of orders. Moreover, certain existence theorems in analysis,
such as that of the Least Upper Bound, cannot be proved within the restrictions
of the ramified theory of types. The Cantorean theory of the infinite, which is
basic to nearly all of modern mathematics, cannot be established within the
rigid framework of the ramified type theory. And even the principle of
mathematical induction must be abandoned in its full generality, for its
complete statement is prevented by the ramified theory of logical types.
To relax the excessive restrictions of the hierarchy of orders, the ‘Axiom of
Reducibility’ was introduced. That axiom asserts that to any function of any
order and any type there corresponds a formally equivalent first-order function
of the same type (two functions are formally equivalent when, for any admis¬
sible argument, they are either both true or both false). In the presence of this
axiom, the identity relation can be satisfactorily defined in terms of first-order
functions and all of the disadvantages mentioned in the preceding paragraph
disappear. The question naturally arises: Does the Axiom of Reducibility relax
the restrictions of the ramified theory of logical types sufficiently to permit the
reintroduction of the paradoxes? If all of the semantical terms such as ‘true*,
‘false’, and ‘designates’ are added to a logical system like that of Principia
Mathematica, which contains both the ramified type theory and the Axiom of
Reducibility, and names for all functions are added also, then at least some of
the paradoxes reappear.9 On the other hand, if the names of some functions are
left out, then even in the presence of the semantical terms, the paradoxes do
not seem to be derivable even with the aid of the Axiom of Reducibility.10
It may not be out of place here to indicate briefly how the ‘levels of
language* method of avoiding the semantical paradoxes11 is remarkably similar
to the ramified type theory’s hierarchy of orders.12 Confining our remarks to
the Grelling paradox, we note that this paradox does not arise in an object
language (like the extended function calculus, for example) when we assume
that there are no symbols in it that designate symbols. Now does it arise in the
metalanguage of that object language. Since the metalanguage contains syno¬
nyms for all symbols of the object language and names for all symbols of the
object language as well as its own variables and the name relation (which we
write as ‘Des’), the symbol ‘Het’ can be defined in it. By definition:

Het(u)- = : *(3c/>):fDes0'i>Dest// = ^xp =

Yet the Grelling cannot be derived in the metalanguage, because although

there is a symbol for the function Het, there is no symbol for the name of that
function. In other words, in the metalanguage, although we could substitute
‘Het’ for ‘<J>\ we cannot substitute “Het” for V because “Het” is not a symbol
of the metalanguage.
So far, there are no complications. The paradox does not arise in the object
language because that structure contains no names for the symbols in it. Nor
does the paradox arise in the metalanguage, because there is no name for the
function symbol ‘Het* in that language. The threat of the Grelling arises only in
the meta-metalanguage. If certain safeguards or refinements of the levels of
language theory are ignored, the paradox seems to be derivable. Ignoring those
safeguards, we have, by definition, in the meta-metalanguage,

Het(u) • EE : "(3$):vDes$ • uDesi// = ^ xp = </> • ^<p(v)

Since the meta-metalanguage doef*contain a name for the function symbol

‘Het*, we substitute that name, “Het**, for the variable V and obtain the
contradiction, as we did in our first derivation.

9See the author’s article on 4The Inconsistency or Redundancy of Principia Mathematica’ in

Philosophy and Phenomenological Research, vol. 11, no. 2 (December 1950), pp. 190-199.
10See the review of the preceding article by Alonzo Church, The Journal of Symbolic Logic,
vol. 16, no. 2 (June 1951), pp. 154-155.
11 First suggested by Bertrand Russell in his Introduction to L. Wittgenstein’s Tractatus
logico-philosophicus, Harcourt Brace Jovanovicb, New York, and Kegan Paul, Trench, Trubner &
Co. Ltd., London, 1922, p. 23.
12The remainder of this appendix is reprinted by permission from The Inconsistency or
Redundancy of Principia Mathematica, Philosophy and Phenomenological Research, vol. 11,
University of Buffalo, Buffalo, N. Y., 1950.
 The Rammed Theory of Types 351

The way in which the safeguards of the levels of language theory serve to
prevent this contradiction makes it very similar to the theory of orders. The
definition of Het in the meta-metalanguage that was written above requires the
addition of subscripts to resolve its ambiguity. Once these ambiguities are
pointed out and resolved, the contradiction vanishes.
In the first place, the meta-metalanguage contains two symbols for the name
relation, DeSj and Des2 .13 The first of these is the meta-metalanguage’s
synonym for the name relation in the metalanguage. The full sentence

vDes^
4

asserts that the function symbol denoted by V is a symbol of the object

language and designates the function <f>. The second has no synonym in the
metalanguage. The full sentence

fDes2c/>

asserts that the function symbol denoted by V is a symbol of the metalanguage

and designates the function <f>. These are quite different.
And in the second place, the meta-metalanguage contains two symbols ‘Het/
and ‘Het2\ between whose meanings there is significant difference. The first of
these is the meta-metalanguage’s synonym for the function symbol ‘Het’ of the
metalanguage. The full sentence

Het/f)

asserts that the function symbol denoted bv V is a symbol of the object

language and designates in the object language an attribute that it does not
possess. The second has no synonym in the metalanguage. The full sentence

Het2(u)

asserts that the function symbol denoted by i» is a symbol of the metalanguage

and designates in the metalanguage an attribute that it does not possess. Their
definitions are different:

Het/c) = df (3<t>):vDesrf>-vDesx\p = ^

and

Het2(i>) = df (3<f>):uDes24> * i>Des2^ = + $ = <t>'

It is clear that we cannot define He^ in terms of Des2 because the values of the
arguments of the two functions are terms of different languages, of the object

13As was suggested by Ramsey in a somewhat different context, op. cit., p. 370, in reprint,
p. 43.
language for He^ and of the metalanguage for Dcs2. For the same reason, Het2
cannot be defined in terms of Desr
No version of the Grelling can be derived fron? the definition of Hetls
because the only values of its arguments are terms of the object language, and
there is no term of the object language analogous to either ‘Het^ or ‘Het2\ The
only possibility lies in the direction of deriving a contradiction from the
definition of Het2, and this is thwarted by something remarkably like Prin-
cipias theory of orders.
In the definition of Het2 we cannot substitute for V the name of the symbol
for that function, because although the function symbol 4Het2’ occurs in the
meta-metalanguage, no name of that function symbol occurs there. The best
we can do is to substitute the meta-metalanguage’s name of the function
symbol of the metalanguage that is synonymous with ‘Het1’, for which we do
have a name in the meta-metalanguage (call it “Het”). Making the substitution,
we have

Het2(4Het’)* = : •(3<^):‘Het,Des2<;)'<Het,Des2^ = ^ \p = • ^<f>(‘Het’)

If we endeavor to deduce a contradiction from this equivalence by an argu¬

ment parallel to earlier versions, we are unable to do so. There is a choice of
function symbol to substitute for the generalized function variable \f/\ for there
are two function symbols in our meta-metalanguage that look promising: ‘Het/
and 4Het2’.
If we substitute <Het1\ we obtain

Het2(‘Het’)* D -.‘Het’Des^-TIet’DeSgHetj = Ile^ = <j>*^<}>(TIet’)

Since “Het” is the meta-metalanguage’s name for the function symbol of the
metalanguage that is synonymous with *Het1’, we have

‘Het’DeSgHet^^et’DesgHeti = Hetj = <J>

and consequently

Het2(‘Het’fD -Het^Het )

But this is no part of any contradiction. It is antecedently known on inde¬

pendent grounds that if any term satisfies Het2, it is in the metalanguage and
not in the object language, whereas only terms of the object language satisfy
Hetr
On the other hand, if we substitute ‘Het2’, we obtain

HetgfHet’) D :‘Het’Des2<;> • ‘Het’Des2Het2 = Het2 = </>•—o(‘Het’)

From this, if “Het’Des^et^’ were true, we should indeed be able to obtain a

contradiction. But “Het’Des2Het2’ is not true, because the argument “Het”
 The Ramified Theory of Types 393

denotes a symbol of the metalanguage, whereas the attribute denoted by the

argument llet2’ is not denoted by any symbol of the metalanguage. In other
words, “Het’Des2Het2* is false because Tlet2’ is a symbol of the meta-
metalanguage which has no synonym in the metalanguage.
This is very like the theory of orders, because the contradiction is evaded by
arranging that certain symbols of the meta-metalanguage are defined over
certain ranges. Thus ‘Des/ is defined over a narrower range than 4Des2 , and
‘Ilet/ is defined over a narrower range than ‘Met./; Desl and Ilett are satisfied
only by symbols of the object language, Des2 and Het2 are satisfied only by
symbols of the metalanguage, which is a wider and more inclusive language.
Not only is the levels of language theory remarkably analogous to the theory of
orders, but where each metalanguage is conceived as actually containing the
object language with which it deals,11 it can be identified with the Russellian
theory of orders as applied to symbols rather than to the functions they denote.
In spite of the indicated similarities, there are fundamental differences
between the two theories. Most significant is that unlike the ramified type
theory, the levels of language device for avoiding paradoxes does not jeopard¬
ize the derivation of any parts of classical mathematics. 1 hus no need aiises for
any analogue to the rediicibility axiom.

uv a Tarski in his ‘Semantic Conception of Truth ,

14This conception is recommended by A. I ■ .375. Reprinted in II. Feigl and

pp. 52-84. See especially p. *50 in vol. 4 and pp. 00-61 m reprmt.
"*
 Solutions to

Exercises on pages 14-15:

I. 1. F 5. T 10. F 15. F
II. 1. F 5. T 10. T 15. F
III. 1. (A-B)vC 5. (CvD)--(OD)
IV. 1. S-W 5. G-S

Exercises on pages 18-19:

I. 1. T 5. F 10. T 15. T
II. 1. T 5. F 10. F 15. F
III. 1. (A-C) D -D 5. -A D -(CvD) 10. (A D -C)-(~~C D D)

Exercises on pages 25-26:

I. a. 4 is the specific form of a. e. 13 has e as a substitution instance,
j. 10 has j as a substitution instance, and 20 is the specific form of j.
II. 1. Valid 5. Invalid—shown by row 2. 10. Invalid—shown by row 1.
15. Valid.
)t III. 1. Invalid—shown by row 1. 5. Valid. 10. Invalid—shown by row 3.
15. Invalid—shown by row 12.

Exercises on pages 30-31:

I. 1. Contingent—F in row 1, T in row 2. 5. Tautologous.
II. 1. Not equivalent—shown by rows 2 and 3. 5. Equivalent.

Exercises on pages 34-38:

I. 1. Simplification (Simp.). 5. Hypothetical Syllogism (H.S.)
10. Constructive Dilemma (C.D.)
II. 1. 3. 2, Simp. 5. 4. 3, Simp. 10. 7. 1, 6, M.T.
4. 1, 3, M.P. 5. 2, 4, M.P. 8. 2, 6, M.T.
5. 3, Simp. 6. 4, Simp. 9. 7, 8, Conj.
6. 4. 5, M.P. 7. 6, Add. 10. 3, 9, M.P.
7. 6, Simp. 8. 1, 7, M.P. 11. 4, 10, C.D.
8. 7, Add. 9. 8, 5, H.S. 12. 5, 11, D.D.
395
3J6 Solutions to Selected Exercises on Peges 36-38

II. I. 1. A D B
2. C D D
3. Hv-D)*(-Av-B) /.*. —A v —C
4. (A D B)-(C D D) 1, 2, Conj
5. —B v 3, Simp.
6. ~A v —C 4, 5, D.D.

5. 1. (R D ~S)-(T d ~r/)
2. (V D ~W)-(X D ~~Y)
3. (T D W)-(t/ =3 S)
4. Vvfl /:,. —Tv — 17
5. V D ~W 2, Simp.
6. R D ~S 1, Simp.
7. (V D <—W)*(R D ~S) 5, 6, Conj.
8. ^Wv —S 7, 5, C.D.
9. ~~Tv 3, 8, D.D.

IV. 1. 1. (A vG)DS
2. A-T /.\s
3. A 2, Simp.
4. A v G 3, Add.
5. S 1, 4, M.P.

1. (—K mP) D (B v R)
. 2. —K D (B D D)
3. K v(R D J£)
4. — K-P /.*. DvE
5. —K 4, Simp.
6. B D D 2, 5, M.P.
7. R D E 3, 5, D.S.
8. (B D D)-(fl D E) 6, 7, Conj.
9. BvB 1, 4, M.P.
10. DvE 8, 9, C.D.

10. 1. Tv(EDD)
2. T D C
3. (E D G) D (D D /)
4. (r-T ~C) D (D D C)

/.'. -—Dv —•£

7. —T 2, 5, M.T.
8. ED D 1, 7, D.S.
9. 7, 5, Conj.
10. D D G 4, 9, M.P.
11. E D G 8, 10, H.S.
12. D D 7 3, 11, M.P.
13. (D D 7)-(E D G) 12, 11, Conj.
14. —D v —E 13, 6, D'.ip. *
 Solutions to Selected Exercises on Paget 43-45 SS7
Exercises on pages 43-47:
I. 1. Commutation (Com.) 5. Tautology (Taut.) 10. Commutation
(Com.) 15. Distribution (Dist.)

II. 1. 3. 2, Add. 5. 4. 1, Exp.

4. 3, De M. 5. 2, Impl.
5. 1, 4, M.T. 6. 5, Taut.
6. 5, De M. 7. 4, 6, M.T.
7. 6, Com. 8. 7, De M.
8. 7, Simp. 9. 3, 8, D.D.
10. 9, Impl.

III. 1. 1. -A /.*. A D B 5. 1. K D L /: . K D (LvM)

2. -AvB 1, Add. 2. vL 1, Impl.
3. AD B 2, Impl 3. (-K v L) v M 2, Add.
4. ~K v (L v M) 3. Assoc.
5. K D (LvM) 4, Impl.

10. 1. A 3 ~(B 3 C) 15. 1. (—1/v V)-((/v W)

2. (D-fl) IT) C 2. -X D -W /.• . V v X
3. D /.’ . -A 3. -l/vV 1, Simp.
4. D D (B D C) 2, Exp. 4. 1/ D V 3, Impl.
5. B D C 4, 3, M.P. 5. W D X 2, Trans.
6. D C) 5, D.N. 6. (U D V)-(W D X) 4, 5, Conj
7. -A 1, 8, M.T. 7. (UvW)-(^lJv V) 1, Com.
8. U v VV 7, Simp.
9. VvX 6, 8, C.D.

20. 1. OD(FDQ)
2. P D (Q D R) /:. OD(PD R)
3. (OP) 3 Q 1. Exp.
4. (P-Q) DR 2, Exp.
5. (Q-P) DR 4, Com.
6. QD (PDR) 5, Exp.
7. (O-P) D(PDR) 3, 6, Fl.S.
8. OD[PD(PD R)] 7, Exp.
9.0: [(P-P) D R] 8, Exp.
10. O D (P D R) 9, Taut.

25. 1. J-(K-L)
2. (K-J) D [Mv(NvO)]
3. ~M-(~P—0)/.\N
4. (J-K)-L 1, ASSOC.
5. J’K 4, Simp.
6. K-J 5. Com.
7. Mv(NvO) 2, 6, M.P.
8. ~M 3, Simp.
9. NvO 7, 8, D.S.
10. 0\N 9, Com.
11. (~M-~P)—O 3. Assoc.
Ml SotuHons to Saiectad Exercises on Page* 45-92

12. ~0-(~M-~P) 11, Com.

13. ~0 12, Simp.
14. N 10, 13, D.S.

IV. 1. 1. S D G 5. 1. (G-L)v(W-T)
2. ~S D E /.'.GvE 2. —G /.\T
3. ~G D ~S 1, Trans. 3. v ~L 2, Add.
4. ~G D E 3, 2, H.S. 4. ~(G-L) 3, De M.
5. --G v E 4, Impl. 5. W’T 1, 4, D.S.
6. GvE 5, D.N. 6. T>W 5, Com.
7. T 6, Simp.
10. 1. BvT 15. 1. (WDM)*(/DE)
2. (BvC)D (L-M) 2. Wv/
3.-l /.*.r 3. (W D -£)•(/ D ~~Af) /.’. E =
4. -Lv-Af 3, Add. 4. MvE 1, 2, C.D.
5. 4, De M. 5. —E v 3, 2, C.D.
6. -(BvC) 2, 5, M.T. 6. £ D -Af 5, Impl.
7. -B--C 6, De M. 7.--MvE 4, D.N.
8. ~~B 7, Simp. 8. ~~Af D E 7, Impl.
9. T 1, 8, D.S. 9. (E D D E) 6, 8, Conj
10. E=~~M 9, Equiv.
21. 1 .(TDE)-(ADL) /.'. (TvA) D (EvL)
2. TDE 1, Simp.
3. -TvE 2, Impl.
. 4. Kv£)vL 3, Add.
5. -Tv(EvL) 4, Assoc.
6. (EvL)v-T 5, Com.
7. (AD L)-(T D E) 1, Com.
8.ADL 7, Simp.
9. -AvL 8, Impl.
10. (-AvL)vE 9, Add.
11. —A v (L v E) 10, Assoc.
12. -Av(EvL) 11, Com.
13. (EvL)v-A 12, Com.
14. [(E vL)v-!]•[(£ vL)v —A] 6, 13, Conj.
15. (EvL)v(~~T-~A) 14, Dist.
16. (~~T-~A)v(EvL) 15, Com.
17. ~(TvA) v(£vL) 16, De M.
18. (Tv A) D (EvL) 17, Impl.

Exercises on page 49;

1. A B CP 5.T_U_V_W_X
F T F F T T T T F
or F F T F T
Exercises on page 52;
21. 1. (T D E)‘(A D L) /.’.(TvA) D (EvL)
2. Tv A /.’.EvL '.
3. EvL 1, 2, C.D.
 Solutions to Selected Exercises on Psges 54-88 358
Exercises on page 54:
3. 1. (77 3 1)-(J 3 K)
2. (7 v K) 3 L
3. /.\~(Wv/)
4. —(/v K) 2, 3, M.T. 4'. --(//v 7) I.P.
5. — /• — K 4, De M. 5'. // v / 4', D.N.
6. —I 5, Simp. 6'. 7 v A: 1, 5', C D.
1. H D1 1, Simp. 7'. Z. 2, 6\ M.P.
8. —// 7, 6, M.T. 8'. L-~~L 7\ 3, Conj.
9. (JDK)-(HDI) 1, Com.
10. JD K 9, Simp.
11. —X—I 5, Com. 4
12. -X 11, Simp.
13. -/ 10, 12, M.T.
14. —H—J 8, 13, Conj.
15. —(tfv/) 14, De M.

5. 1. (V3 ~W)-(X 3 Y)
2. (~W 3 Z)-(Y 3 ~A)
3. (Z 3 ~B)-(—A 3 C)
4. V-X /. ’. —8-C
5. V D — W 1, Simp. 5'. -(-fl-C) I.p.
6. V 4, Simp. 6'.-B v —C 5', De M.
7. —W 5, 6, M.P. 7'. -Z v-A 3, 6', D.D.
8. —W D Z 2, Simp. 8'. ——vv v — y 2, 7', D.D.
9. Z 8, 7, M.P. 9'. — V v —X 1, 8', D.D.
10. Z D —8 3, Simp. 10'. -(V-X) 9', De M.
11. —8 10, 9, M.P. 11'. (V-X)--(V-X) 4, 10', Conj
12. (X D Y)-( V D W) 1, Com.
13. XDY 12, Simp.
14. X-V 4, Com.
15. X 14, Simp.
16. Y 13, 15, M.P.
17. (Y D -A)*(-W D Z) 2, Com.
18. Y D -A 17, Simp.
19. —A 18, 16, M.P.
20. (—A D C)-(Z D ~B) 3, Com. -
21. -ADC 20, Simp.
22. C 21, 19, M.P.
23. — B-C 1J, 22, Conj.

Exercises on page 56:

I. 1. 1. P (C.P.)
2. Pv— Q 1, Add.
3. —p v P 2, Com.
4. Q D P 3, Impl.
5. 1. -J-P /•'P (C.P.)
2. P 1. D.N.
360 Solutions to Selected Exercises on Pages 56-70

10 . 1. ADB /.*. A D (A-B) (C.P.)

2. A /..A-B (C.P.)
3. B 1, 2, M.P.
4. A-B 2, 3, Conj.

15. 1. A 3 (B-C) /:. [B 3 (D-E)] 3 (A 3 D) (C.P.)

2. B 3 (D-E) /.\A 3 D (C.P.)
3. A /. .D (C.P.)
4. B-C 1, 3, M.P.
5. B 4, Simp.
6. D-fi 2, 5, M.P.
7. D 6, Simp.
II. 1. 1. ~[(A 3 .B)v(A 3 ~B)] /: (A DB)’ -B) (I.P.)
2. ' (A 3B)-~Mi 3 ~B) .
1 De M.
3. ~(~A v B)*~(.—A v —B) 2. Impl.
4. (——-A*~B)'(~.—A-■B) 3, De M.
5. (~~-A-~B)-(--> A) 4. Com.
6. K—-A- ~B)-~-~B]-~»~A 5. Assoc.
7. [-"A •( ~B-~ ~B)]— -A 6. Assoc.
8. [(~B-~ ~B) • ~—A]--A 7, Com.
9. (~B — ■'B),(—-—A • -A) 8. Assoc.
10. ^B ■ -B 9, Simp.

5. 1. ~[(A 3 B) v (~A D C)] /: (A 3 B) v (~A 3 C) (I.P.)

2. ~(A 3 B) (—A 3 C) 1, De M.
3. ~( —A v B)-—(-A v C) 2, Impl.
4. (~ —A*—B)*(-A C) 3, De M.
5. ~A-[~B-(~~~A-~C)] 4, Assoc.
6. —■~A-[( -A • —C)—B] 5, Com.
7. — ~-A-[-~A-(~C-~B)] 6, Assoc.
8. —A-A)*(~C B) 7, Assoc.
9. — A • — — —A 8, Simp.

Exercises on page 61:

1.1.ADB
2. B 3 [(C 3 C) 3 D] /.’.A 3 D
—>3. A
4. B 1, 3, M.P.
5. (CD- —C) 3D 2, 4, M.P.

7.
r
6. C

8. C 3
6, D.N.
6-7, C.P.
9. D 5, 8, M.P.
10.ADD 3-9, C.P.

Exercises on pages 69-71:

I. 1. (x)(Sx 3 Bx) 5. (x)(Sx 3 fix)
10. (x)(Dx 3 Bx) 15. (x)[Cx 3 (*. Wx v Sx)]
20. (x)[(/x• (/x) 3 (PxvFx)]
 Solutions to Selected Exercises on Pages 70-77 361
25. (x)[(Ax-Ox) 3 (Dx 3 -Rx)]
30. (x)[(tfx-Cx) 3 Tx] 35. (x)[(C.x-Hx) 3 Tx] 40. (x)[(7W/x) 3 Gx]
II. 1. (x)(Cx 3 Bx) 5. (x)[\tx 3 (Vx• Fx)}

Exercises on pages 76-77:

I. 1. 1. (x)(Ax 3 Bx)
2. ~Bf /.'. —At
3. At 3 Bt I, UI
4. —At 3. 2, M.T.
5. 1. (x)(Kx 3 Lx)
2. (x)[(Kx-/.x) 3 Alx 1 (x)(Kx 3 Six)
—* 3. Ky
4. Ky 3 Ly 1, UI
5. Ly 4, 3, M.P.
6. KyLy 3, 6, Conj.
7. (KyLy) 3 My 2, UI
8. My 7, 6, M.P.
9. Ky 3 My 3-8, C.l*.
]0. (x)( Kx 3 Alx) 9, UG
pH. 1. ‘ 1. (x)(Ax 3 Bx) 5. 1. {x)(Lx
(x)(Lx 3 Alx)
2. ~Br; /.'.-Ac
~Af 2.(3 /. . (3x)(N.vAfx)
x)(l.x-Nx) /.’.
2. (3x)(l.x-\x) (lx)(\xmMx)
3. Ac D Be 1, UI 3. Lt-.Vtt
L/i-.Vri 2, El
2, El
4. —At; 3, 2, M.T. Im 3 Air/
4. Lit Afn I, UI
I. UI
5. /.tt 3, Simp.
6. Afrt I, 5, M.P.
7. No ml id 3, Coni.
8. AV/ 7, Simp.
9. Mr M a 8, 6, Conj.
10. (3.v)(JV.v-A/*) 9, EG

10. 1. (x)(D.y D Ex)

2. (*)(E* D Fx)
3. (x)(Fx 3 Gx) (x)(D.v 3 G.v)
4. Dy 3 /it/ 1. UI
5. Faj 3 Ft/ 2, UI
6. Ft/ 3 Gy 3, UI
7. Dy 3 Ft/ 4, 5, II S.
8. Dy 3 Gt/ 7, 6, U S.
9. (x)(Dx 3 Gx) 8, UG
15. 1. (x)[(Bx y Vx) 3 (Ox-Dx;| . (x)( Bx 3 Dx)

p> 2. By
3. (By v Vy) 3 (OyDy) U UI
2, Add.
4. By v Vy
3, 4, M.P.
5. OyDy
5, Com.
6. DyOy
0, Simp.
7. Dy
2-7, C.P.
8. By 3 Dy
8, UG
9. (x)(Bx 3 Dx)
382 Sohittoiis to Selected Exercises on Poges 77-83
20. 1. (x){(Bx v Wx) 3 [(Ax v Fx) 3 Sx]} (x)[Bx 3 (Ax 3 Sx)]
-*2. By
—> 3. Ay
4. (ByvWy) 3 [(Ay v Fy) 3 Sy] 1, UI
5. ByvWy 2, Add.
6. (Ay v Fy) 3 Sy 4, 5, M.P.
7. AyvFy 3, Add.
| 8- Sy 6, 7, M.P.
9. Ay 3 Sy_ 3-8, C.P.
10. By 3 (Ay 3 Sy) 2-9, C.P.
11. (x)[Bx 3 (Ax 3 Sx)] 10, UG

Exercises on pages 82-83:

I. 1. (3x)(Ax-Bx) | g, c | (Aa>Ba) v (AcBc)
Ac Ac
. ‘. Be .'. Be

Aa Ac Ba Be
"T T T F~

5. (3x)(Kx-Lx) | a, b | (Ka'La) v (Kb^Lb)

(3x)(~Kx—Lx) (~~Ka-~~La) v (~~Kb-~~Lb)
(SxXXx-—Kx) (La—Ka) v (Lb*~~Kb)

Ka Kb La Lb
T F T F
or F T F T

II. 1. (x)(Ax D Bx) A; D Bj Aj Bj

Bj Bj F T
/.A; .*.A/
5. (x)(Kx 3 — Lx) (a (Ka 3 ~La)-(Kb 3 —Lb)
(3x)(Mx-Lx) (Ma-La)v (Mb-Lb)
(x)(Kx 3 —Mx) (&a 3 ~-Ma)-(Kb 3 —Mb)

Ka Kb La Lb Ma Mb
T F F T T T
or F T T F T T

III. 1. (x)(Ax 3 Bx) 7] A; 3 Bj Aj Bj

Bj Bj F T
••A/ ..A/
5. 1. (x)(Vx 3 Cx)
2. (3x)(Bx—Cx) (3x)(/lx*—V5t)
3. Ra—Ca 2, El’ . '
4. Va 3 Ca 1. UI
 Solutions to Selected Exercises on Pagan *3-101 303

5. ~~Ca-Ra 3, Com.
6. ~~Ca 5, Simp.
7. —Va 4, 6, M.T.
8. Ra 3, Simp.
9. Ra'-^Va 8, 7, Conj.
10. (3x)(Rx-~Vx) 9, EG
10. (*)[T* 3 (Ox v Ex)] B Ta 3 (iOa v Ea) Ta Oa Ea
(x)[(Ox-Tx) 3 ~£x] (Ofl'Ta) D — Ea "X F T"
(3x)(7x-£'x) Ta-Ea
.'. (3x)(71c*Ox) Ta-Oa
Exercises on page 88:
I. (3x)Dx 3 (3x)(Py-By)
5. (*)[Bx 3 (Yx 3 Rx)]
10. (*){(Ox-Px) 3 [(y)(Mj/ 3 —Py) v Afx]}
15. (*){[Sx-(t,)(Sy 3 Wy)] 3 Wx}
Exercises on page 92:
I. 1. d and /
5. none
II. 1. a, b, and c
2- g
Exercises on pages 100-101:
3. The use of UG at line 4 is wrong, because v (V) occurs free in the assumption
'Fx = Gy’ within whose scope the premiss Ok (‘Fx = Gy) lies.
6. There are three mistakes here:
a. The use of UG at line 5 is wrong, because v (V) does not occur free in
(‘(3x)[(Fx*Gz) D Hy]') at all places that ju (*y) occurs free in O/a
C(3x)[(Fx-Gy) D Hy]').
b. The use of UG at line 5 is wrong, because the variable v (%z) occurs free in line
2 within whose scope (‘(3x)[(Fx*Gz) D Hy]') lies.
c. The use of UI at line 7 is wrong, because v (V) does not occur free in Qv
(‘(3x)[(Fx-Gx) D Hx]’) at all places that /a (Y) occurs tree in 4>/a
(\3x)[(Fx-Gy) D Hy]').
9. There are three (or four) mistakes here:
Line 3, ‘Fx'Gy’, is claimed at line 5 to be a corresponding to the (‘Fx-Gx) in
line 1. This claim is false regardless of whether we take v to be Y or to be *y . So
in either case we note that
a. The use of El at line 5 is wrong, because v (x* or 'y') does not occur free in <S>v
(‘Fx'Gy’) at all places that /a (Y) occurs free in <!>/a (‘Fx-Gx’).
If we take v in $v ({Fx'Gy ) to be x , then we note that
b. The use of El at line 5 is wrong, because v (Y) occurs free in p (*Fx’)t the last
line lying within the scope of the assumption (‘Fx-Gy').
In any case, there are two additional mistakes present:
c. The use of El at line 8 is wrong, because v (Y) occurs free in line 5 which
precedes (4~Fx*Gx').
d. The use of El at line 8 is wrong, because v (Y) occurs free in p ( —Fx ), the last
line lying within the scope of the assumption <t>v (‘—Fx-Gx*).
364 Solutions to Selected Exercises on 103-104

Exercises on pages 103-105:

I. 1. 1. (x){Ax 3 Bx) /.*. (x)(J3x D Cx) Z> (A& Z3 CJfc)

I-* 2. {x)(Bx 3 Cx)
r- 3. Ak
4. Ak 3 Bk 1, UI
5. Bfc 4, 3, M.P.
6. Bk 3 CJfc 2, UI
1 7. Cfc a s, m.p.
8. Afc 3 Cfc 3-7, C.P.
9. (*)(Bx 3 Cx) 3 (Ak ~Ck) 2-8, C.P.

5. 1. (3x)Lx 3 (y)My /.*. (x)[Lx D (y)Myj

2. L*
3. (3x)Lx 2, EG
4- (_y)My 1, 3, M.P.

5. Lx D (y)Mij 2-4, C.P.

a (*)[f-x D (y)Mi/] 5, UG
10. 1. (3x)Ax D (y)(By D Cy)
2. (3x)Dx D (3y)By /.*. (3x)(Ax-/J>x) D (3y)Cy
3. (3x)(Ax-/>x)
4. Ax-Dx
5. Ax 4, Simp.
6. (3x)Ax 5, EG
7. ly)(Bij 3 Cy) 1, 6, M.P.
8. Dx 4, Simp.
9. (3x)Ux 8, EG
10. (3y)By 2, 9, M.P.
11. By
12. By 3 Cy 7, UI
13. Cy 12, 11, M.P.
14. (3i/)Cy 13, EG
15. (3y)Cy 10, 11-14, El
16. (3y)Cy 3, 4-5, El
17. (3x)(Ax-Dx) D (3y)Gy 3-16, C.P.
 Solutions to Selected Exercises on Psge 104 385
15. 1. (x){Ox 3 [(y)(Py 3 Qy) 3 fix]}
2. (x){Rx 3 [(y)(Py 3 St/) 3 Tx]} /.-.(y)[Py D (PySy)j D (x)(Ox D ft)
3- (y)[Py 3 (<?y*Sy))
4. Ox
5. Ox D [(y)(Py 3 Qy) 3 Rx] 1, UI
6. (y)(Py 3 Qy) 3 Rx 5, 4, M.P.
7. Py 3 (Qy Sy) 3, UI
8. (Py 3 Qy)’(Py 3 Sy) 7, Dist.
9. Py 3 Qy 8, Simp.
10. (y)(Py 3 Qy) 9, UG
11. Rx 6, 10, M.P.
12. Rx 3 [(y)(Py 3 Sy) 3 Tx] 2, UI
13. (y)(Py 3 Sy) 3 Tx 12, 11, M.P.
14. Py 3 Sy 8, Simp.
15. (y)(Py 3 Sy) 14, UG
16. Tx 13, 15, M.P.
17. Ox 3 Tx 4-16, C.P.
18. (x)(Ox 3 Tx)_ 17, UG
1T[y)[Py 3 (py-Sy)l 3 (x)(Ox 3 7x) 3-18, C.P.

II. 4. 1. (3x)Gx 3 (y)[Cy 3 Gy)

2. (3x)(Px-Tx) 3 (y)(Gy 3 Ty) /(3x)[Px*('Tx*Gx)] 3 (y)(Cy 3 Ty)
3. (3x)[Px-(Tx-Gx)l
4. Pu’(Tu’Gu)
5. Gu 4, Simp.
6. (3x)Gx 5, EG
7. Pu-Tu 4, Simp.
8. (3x)(Px-Tx) 7, EG
9. (yXCy 3 Gy) 1, 6, M.P.
10. (y)(Gy 3 Ty) 2, 8, M.P.
11. Cm 3 Gu> 9, UI
12. Gw 3 Tto 10, UI
13. Cw 3 Tto 11, 12, H.S.
14. (y)(Cy 3 Ty) 13, UG
15. (y)(Cy 3 Ty)_ 3, 4-14, El
16. (3x)[Px-(TrGx)] 3 (y)(Cy 3 Ty) 3-15, C.P.
366 Solutions to Selected Exercises on Peges 105-108

8. 1. (x)[Rx D (SxvMx)]
2. (x)[(Ux-Rx) D -Sx] /:.(x)(Ux 3 Rx) 3 (y)(Uy 3 My)
-> 3. (x)(Ux D Rx)
r-> 4. Uz
5. Uz D Rz 3, UI
| 6. Rz 5, 4, M.P.
7. Uz-Rz 4, 6, Conj.
8. (Uz*Rz) D —Sz 2, UI
9. 8, 7, M.P.
10. Rz D (Sz v Mz) I, UI
11. SzvMz 10, 6, M.P.
12. Mz II, 9, D.S.
13. Uz D Mz 4-12, C.P.
14. (yXUy D My) 13, UG
15. (x)(Ux D Rx) D (y)(Uy D My) 3-14, C.P.

12. 1. (x){Lx 3 [(y)(Py 3 Vy) 3 Mx])

2. (3z)(Pz-Vz) 3 (y)(Py 3 Vy) /.'. (3x)Lx 3 [(3a)(Pa-Va) 3 (3y)My]
-» 3. (3x)Lx
-> 4. (3z)(Pz-Vz)
5. (y)(Py 3 Vy) 2, 4, M.P.
f—> 6. Lu
7. Lu 3 [(y)(Py 3 Vy) 3 Mu] 1, UI
8. (y)(Py 3 Vy) 3 Mm 7, 6, M.P.
9. Mm 8, 5, M.P.
10. (3y)My 9, EG
11. (3y)My_ 3, 6-10, El
12. (3z)(Pz-Vz) 3 (3y)My_ 4-11, C.P.
13. (3*)/,* 3 [(3z)(Pz-Vz) 3 (3y)My] 3-12, C.P.

Exercises on page 108:

1, (*)(y)[A* 3 (By v Cy)] _
(z) {[(v (y)Cy] 3 Da} [oJT]
(3x)(3z)(Ax D Dz)
{[Aa 3 (BavCa)]-[Aa 3 (Bb v CbjJ] • {[Ab 3 (BavCa)]-[Ab 3 (BbvCb)]
{[(Ba-Bb) v (Co-Cb)l 3 Da) • {[(Ba-Bb) v (Ca-Cb)] 3 Db}
[(Aa D Da) v (Aa D Db)] v [(Ab D Da) v (Ab D Dfr)]

Aa Ab Ba Bb Ca Cb Da Dh
~~T T T F F T F F"
orTTFTTFFF

5. (*)(3y)(Fx=Gy) _
,’.(3y)(x)(Fx=Gy) [aj>\
[(Fa = Ga) v (Fa = Gb)] • [(Fb == Ga) y (Fb = Gb)]
.'. [(Fa = Ga)’(Fb = Ga)] v [(Fa = Gb) • (Fb = Gb)]
 Solutions to Mooted Exercises on Pages IOS-115 367
Fa Fb Ca Gb
T F T F
or T F F T
or F T T F
or F T F T

Exercises on page 115:

1. * 1. Fy
2. (3x)Fx 1, EC
3. Fy D (3x)Fx 1-2, C.P.
4. (y)[Fy D (3x)Fx] 3, UG

5. ->1. (3*)(Fx-Gx)
p* 2. FyGy
3. Fy 2, Simp.
4. (3x)Fx 3, EG
5. Gy 2, Simp.
6. (3x)Gx 5, EG
7. (3x)Fx*(lx)Gx 4, 6, Conj.
8. (lx)Fx-(3x)Gx 1, 2-7, El
9. (3x)(Fx-Gx) D [(3x)Fx-(3x)Gx] 1-8, C.P.

10. r *l- ~(3«/)P*)F* D Fyi

2. (y)~[(3x)F* 3 Fy] 1, QN
3. —[(3x)Fx 3 Fy] 2, UI
4. (3x)Fx—Fy 3, Impl.
5. —Fy 4, Simp.
6. (x)~Fx ■ 5, UG
7. ~(3*)F* 6, QN
8. (3x)Fx 4> Simp.
9. (3x)Fx v (3y)[(3x)Fx 3 Fy] 8, Add.
10. (3y)[(3x)Fx 3 Fy] _ 9. 7, D.S.
11. ~(3y)[(3x)Fx 3 Fy] 3 (3y)[(3x)Fx 3 Fy] 1-10, C.P.
12. (3y)[(3x)Fx 3 Fy] v (3y)[(3x)Fx 3 Fy] 11, Impl., D.N.
13. (3y)[(3x)Fx 3 Fy] 12, Taut.

15 . -* 1. (3x)Fx 3 (y)Cy
2. F*
3. (3x)Fx 2, EG
1, 4, M.P.
4. (y)Gy
4, UI
5. Gy
2-5, C.P.
6. Fx 3 Gy
7. ij/)(F* 3 Gy) 6, UG
8. (x)(y)(Fx 3 Gy)__ 7, UG

9. [(3x)Fx3 (y)Gy] 3 (*)(y)(Fx 3 Gy) 1-8, C.P.

3M Solutions to Selected Exorcises on Pages 115-132

-> 1. (x)(y)(Fx D Gy)

i—«■ 2. (3x)Fx . .
i—> 3. Fx
4. (y)(Fx D Gy) 1, UI
5. Fx D Gy 4, UI
6. Gy 5, 3, M.P.
7- (y)Gy 6, UG
8. (y)Gy 2, 3-7, El
9. (3x)Fx D (y)Gy 2-8, C.P.
10. (x)(y)(Fx D Gy) D [(3x)Fx D (y)Gy] 1-9, C.P.

Exercises on pages 127-130:

I. 1. Every dog has his day.
5. All things come to him who waits.
10. A person is judged by the company he keeps.
15. It’s an ill wind that blows nobody any good.
20. Nobody borrows everything from everybody.

II. 1. (x)[(Dx-Mx) D (yX’fy D -Txy)]

5. (x){[Bx‘(y)(Txy D Ly)] D (z)[Txz D (u)~Buz]}
10. (x){Px D [(3t/)Axt/ D (z)(Pz D £zx)]}
15. (x){Pz D {(ly){Dy-Hxy(z)[(Pz-Vzx) D Byz]) D (3u)(Nu-Mxu)])}
20. (3x){Sx-(y)[Py D (3z)Byzx]}
25. (x){Sx D (y){Py D (3z)[(3to)Bujzx*~Bwzx ]}}
30. (x){Px D (y)[Cy D (3z)(Bzx-~Dxzi/)]}
35. (x){Cx D (y)[(3u)(3v)Dyuv D (3z)(hv)(Dyzw■ —Dyzx)])
40. (3x){Px‘(3y){Cy(z)[(3u)Dxzu D Dxzy]))
III. 1. (x){{Px-(3u)[Mu-(3z)(Bz-Bzu-Sxz)]} D
(3u)[M« • (3«)(Bo • Bvx • Sue)]}
5. (x){{Mx’(y)[(RyByx) D Sxy]) D
(z)[(Mz-Pxz) D Hxz}}

Exercises on pages 132-134:

I. 1. 1. (3x)(y)[(3z)Ayz D Ayx]
2. (y)(3z)Ayz /:. (3*){y)Ayx
r—» 3. (y)[(3z)Ayz D Ayx]
4. (3z)Ayz D Ayx 3, UI
5. (3z)Ayz 2, UI
6. Ayx 4, 5, M.P.
7. (y)Ayx 6, UG
8- (3x)(;/)Ayx 7, EG
9. (3x)(y)Ayx 1, 3-8, El
 Solutions to Selected Exercises on Psge 133 369
5. 1. (3x)[Hx-(y)(Iy 3 Jxy)] /.*. (*)(//* D /x) D (3g)(/g-/yy)
► 2. (x)(llx 3 lx)
► 3. Hx-(y)(l,j D Jxy)
4. Hx 3, Simp.
5. Hx D lx 2, UI
6. lx 5, 4, M.P.
7. (y)(ly D Jxy) 3, Simp.
8. lx D /xx 7, UI
9. /xx s, a m.p.
10. Ix'Jxx 6, 9, Conj.
(*y)(iyJyy) 10, EG
12. (3y)(ly •Jyty) 1, 3-11, El
13, (x)(Hx D /x) D (Bi/K/iy-yt/t/) 2-12, C.P.
!• (x){(3y)(Byb-Lxyh) D Fx)
2. (3x)(Cxb-Lxab) /.'.(x)(Cxb D -Ex) D —Bab
3. (x)( Cxb D ~~Fx)
4. Cxb'Lxab
5. Cxb D ~~Fx 3, UI
6. Cxb 4, Simp.
7. ~~Fx 5, 6, M.P.
8. (3y)(Byb-Lxyh) D Fx 1, UI
9. —(3 y)(Byb-Lxyb) 8, 7, M.T.
10. (y)(~~Byh v ~~Lxyb) 9, QN, De M.
11. ^Bahv^Lmb 10, UI
12. Lxab 4, Simp.
13. —Bab 11, 12, Com., D.N., H.S.
14. -Bab 2. 4-13, El
15. 3 3 14 3 14, C.P.
1. (x){(3./)|(3z)(Cz-~Rz'•Sxztj)\ D Cx)
2. (z)[(Wz-Cte) 3 (SZzrv Smzr)] /.*. (3z)(Wz-Orz-Gz-~~Rz
^ [(«)(-
-* 3. (3z)( Wz • Orz • G; • —Rz)
-> 4. (u)(—Smur)
-> 5. Wz-Orz-Gz—Rz
6. Wz-Orz 5, Simp.
7. (Wz'Orz) 3 (Shr v Smzr) 2, UI
8. Slzr v Smsr 7, 6, M.P.
9. —Smzr 4, UI
10. Slzr 8, 9, Com., D.S.
11. Gz-~~Rz 5, Simp.
12. G*—Ks-Ster 11, 10, Conj.
13. (3z)(Gz-~Rz-Slzr) 12, EG
14. (ly)l(3z)(Cz—Rz-Slzy)] 13, EG
15. (3i/)[(3z)(Gz-~itz-SZz!/)] 3 Cl 1, UI
16. a 15, 14, M.P.

17. CZ 3, 5- 16, El
4-17, C.P.
18. 4 3 17
3-18, C.P.
19. 3 3 [18]
370 Solution* to Selected Exercises on Pages 134-139

9. 1. {x)[{Px‘~~Rxx) 3 (y)(Py 3
2. (y){Py D (x)[(Px- ^Ryx) D *Hyx]) /. .[x){[Px-(z)(Pz 3
. • D (y)(Py D
-> 3. Px-(z)(Pz D ~~Rxz)
r-» 4. Py
5. (z)(Pz D ~~Rxz) 3, Simp.
6. Px D ~~Rxx 5, UI
7. Px 3, Simp.
8. ~~Rxx 6, 7, M.P.
9. Px-~~Rxx 7, 8, Conj.
10. (Px—Rxx) D (y)(Py D ~~Ryx) 1, UI
11. (y)(Py => ~Ryx) 10, 9, M.P.
12. Py D — Ryx 11, UI
13. ~~Ryx 12, 4, M.P.
14. Px*~~Ryx 7, 13, Conj.
15. Py D (x)[(Px-~~Ryx) D ~Hyx\ 2, UI
16. (x)[(Px-~~Ryx) D Hyx\ 15, 4, M.P.
17. (Px'~~Ryx) D — Hyx 16, UI
18. ~Hyx 17, 14, M.P
19. 4 D 18 4-18, C.P.
20. (y)(19) 19, UG
21. 3 3 20 3-20, C.P.
22. (jc){21} 21, UG

Exercises on pages 139-140:

3. 1. (x)[Fx 3 (f/)(Sy 3 Oxy)] /:. (y)[Sy 3 (x)(Fx 3 ~Oyx)]
2. (x)(y)(Oxy ‘D ~~Oyx) (auxiliary premiss)
3. Sy
4. Fx
5. Fx D (y)(Sy D Oxy) 1, UI
6. (y)(Sy D Oxy) 5, 4, M.P.
7. Sy D Oxy 6, UI
8. Oxy 7, 3, M.P.
9. (y)(Oxy D ~Oyx) 2, UI
10. Oxy D ~~Oyx 9, UI
11. Oyx m 10, 8, M.P.
12. Fx D ~~Oyx 4-11, C.P.
13. (x)(Fx D ~~Oyx) 12, UG
14. Sy D (x)(Fx D ~~Oyx) 3-13, C.P.
15. (y)[Sy D (x)(Fx D -Oyx)] 14, UG
 Solutions to Selected Exercises on Pages 139-140 371
5. 1. (x){{Px• (ly)[Py(3z)(Cz • Cyz)-Nxy]} D Ux) /:.(x){\Px'(3z)
(Ifc-Cxs)] 3 Ux}
2. (z)(Bz 3 Cz) 1, ... .
3. (x)(Nxx) ] auxiliary premisses)
-> 4. Px’(3z)(Bz’Cxz)
5. (3z)(Bz'Cxz) 4, Simp.
[—> 6. Bz*Cxz
7. Bz 6, Simp.
8. Bz D Cz 2, UI
9. Cz 8, 7, M.P.
10. Cxz 6, Simp.
11. Cz'Cxz 9, 10, Conj.
12. (3z)(Cz-Cxz) 11, EG
13. Px 4, Simp.
14. Nxx 3, UI
15. Px*(3z)(Cz*Cxz)*Nxx 13, 12, 14, Conj.
16. (ly)[Py •(3z)(Cz'Cyz)*Nxy] 15, EG
17. Px-(ly)[Py(lz)(Cz-Cyz)-Nxy] 13, 16, Conj.
18. {Px-{3y)[Py(lz)(Cz-Cyz)-Nxy]} D Ux 1, UI
19. Ux 18, 17, M.P.
20. Ux 5, 6-19, El
21. [Px-(3z)(Bz-Cxz)\ D Ux 4-20, C.P.
22. (x){[Px*(3z)(Bz9Cxz)] 3 Ux} 21, UG

7. 1. (x)[V* 3 (y){Oyx 3 ~Ixy)]

2. (y)(x)[(Rx-Oyx) 3 Tyx] (y)\(3x)(Vx-Oyx) 3 (3x)(Tyx*~/x«/)]
3. (x)(Vx 3 Rx) (auxiliary premiss)
4. (3x)(Vx-Oyx)
r-> 5. Vx • Oyx
I 6. Vx . 5. Simp.
7. Vx 3 Rx 3, UI
8. Rx 7, 6, M.P.
9. Oyx 5, Simp.
10. Rx-Oyx 8, 9, Conj.
11. (x)[(R*-Oyx) 3 Tyx] 2, UI
12. (Rx-Oyx) 3 IV* 11, UI
13. Tyx 12, 10, M.P.
14. Vx 3 (y)(Oyx D ~/xy) 1, UI
15. (y)(Oyx 3 ~lxy) 14, 6, M.P.
16. Oyx 3 ~Ixy 15, UI
16, 9, M.P.
17. ~Ixy
18. Tyx • ~/xy 17, 13, Conj.
19. (3x)(Tyx- ~Ixy) 18, EG
4, 5-19, El
20. (3x)(Tyx‘~~lxy) _
21. (3x)(Vx-Oyx) 3 (3x)(Tyx-^-Ixy) 4- 20, C.P.
22. (y)[(3x)(Vx-Oyx) 3 (3x)(iyx-~Ixy)] 21, UG
372 Solutions to Selected Exercises on Pages 149-150
Exercises on pages 149-150:
2. 1. (3x){Px-Sx-(y)[(PySy) 3 x = y]-Lx} ///(*)[(!*-S*) D Lx]
— 2. Pz-Sz t
-* 3. Px-Sx-(y)[(Py-Sty) 3 x = y]-Lx
4- (y)[(Py*y) =>* = </] 3, Simp.
5. (Pz-Sz) D x.= z 4, UI
6. x — z 5, 2, M.P.
7. Lx 3, Simp.
8. Lz 6, 7, Id.
9. Lz 1, 3-8, El
10. (Pz-Sz) 3 Lz 2-9, C.P.
11. (x)[(Px-Sx) 3 Lx] 10, UG
4. 1. (3x)(Px-(y)[(Pyx y) 3 Fxy]-Sx} /:. (y)[(Py~Sy) 3 (3x)(Px-Fxy)]
—> 2. Py^Sy
-* 3. Px*(y)[(Pyx / y) 3 Fxy]-Sx
4. Sx 3, Simp.
5. —St/ 2, Simp.
6. x / y 4, 5, Id.
7. Py 2, Simp.
8. Pyx / ;/ 7, 6, Conj.
9- ({/)[(Fy*x ^ i/) 3 Fxij] 3, Simp.
10. (Pyx y) D Fxf/ 9, UI
11. Fxy 10, 8, M.P.
12. Fx 3, Simp.
13. Px-Fxy 12, 11, Conj.
14. (3x)(Fx-Fxi/) 13, EG
15. (3x)(Px- Fxy) 1, 3-14, El
16. (Py^Sy) D (3x)(Fx •/'*!/) 2-15, C.P.
17. (^[(Fy—St/) D (3x)(Px-Fxy)] 16, UG
6. 1. (x)(Fx D (u)UFu-Lxu) D Sxiv 11
/•'• (3x){Fx-(y)[(Fy-x ^ y) 3 Fxy]} 3 (3x){Fx*(y)[(Fyx ^ y) 3 Sxy]}
2. (3x){ Fx • (y)[( Fy x ^ y) 3 Fxy]}
_ 3. Fx-(y)[(Fyx ^ y) 3 Fxy]
4. Fx ^ 3, Simp.
5- (*/)[(Fy-a- # y) 3 Fxy] * 3, Simp.
6. Fx 3 (y)[(Fy-Fxy) 3 Sxy] 1, UI
7- (y)t(Fy-Fxy) 3 Sxy] 6, 4, M.P.
8. (FyFxy) 3 Sxy 7, UI
9. (Fy-x ^ y) 3 Fxy 5, UI
10. Fxy 3 (Fy 3 Sxy) 8, Com., Exp.
11. (Fy-x * y) 3 (Fy 3 Sxy) 9, 10, H.S.
12. (Fyx ^ yFy) 3 Sxy 11, Exp.
13. (Fy-x ^ y) 3 Sxy 12, Com., Taut.
14. (y)[(Fyx ^ y) 3 Sxy] 13, UG
15. Fx-(y)[(Fy-x # y) 3 Sxy] . 4, 14, Conj.
16. (3x){Fx-(y)[(Fyx # y) 3 Sxy]} 15, EG
17. (3x){Fx-(y)[(Fyx # y) 3 Sxy]} 2, 3-16, El
18. 2 D 17 2-17, C.P.
 Solutions to Selected Exercises on Pages 155-175 373
Exercises on pages 155-156:
I. 3. (*)(«/)[* ^yD (3 F)(Fx-~~Fy)]
0. (3x){Fxd-(y)(Fyd 3 t = y)-(G)[(Gx-VG) 3 Gd]-(H)[(Hx-FH) 3 -Hd]}
9. (t){[Mf(F)(VF 3 Ft)] 3 V*}*(3*)[MfVf(3F)(VF*— Ft)]
II. 2. -»1. (3t)(3F)Ft -♦ 1. (3F)(3t)Ft
-» 2. (3F)Ft -* 2. (3t)Ft
i—> 3. Fx r-» 3. Fx
4. (3t)Ft 3, EG 4. (3F)Ft 3, EG
5. (3F)(3t)Ft 4, EG 5. (3t)(3F)Ft 4, EG
6. (3F)(3t)Ft 2, 3-5, El 6. (3*)(3F)Ft 2, 3-5, El
7. (3F)(3t)Ft 1, 2-6, El 7. (3x)(3fTp> 1. 2-6, El
8. 1 3 7 1-7, C.P. 8. 1 3 7 1-7, C.P.

L (x)(y)(z)l(RxyRyz) 3 Rxz]-(x)—Rxx
2. (x)(y)(z)[(RxyRyz) 3 Rxz\ 1, Simp.
3. (y)(z)[(Rxy-Ryz) 3 Rxz] 2, UI
4. (z)[(RxyRyz) 3 Rxz] 3, UI
5. (RxyRtjx) 3 Rxx 4, UI
6. (x)—Rxx 1, Simp.
7. —Rxx 6, UI
8. -(Rxy-Ryx) 5, 7, M.T.
9. ~~Rxy v — Ryx 8, Dc M.
10. Rxy D ~~Ryx 9, Impl.
11. (y)(Rxy 3 -Ryx) 10, UC
12. (x)(y)(Rxy 3 —Ryx) 11, UG
13. {1} 3 12 1-12, C.P.
14. (fl){13) 13, UG

(x)(y)[(x = y) = (F)(Fx = Fi/)J /.'.

I!

1.
r>2. Fx
3. ^~~~Fx 2, D.N.
4. Fx 3, D.N.

5. Fx 3 Ft 2-4, C.P.
6. (Fx 3 Ft)-(Ft 3 Ft) 5, Taut.
7. Fx = Ft 6, Equiv.
8. (F)(Ft = Ft) 7, UG
9. ((/)[(* = y) = (F}(Fx=Fy)i 1, UI
10. (t = t) = (F)(Ft = Ft) 9, UI
11. [(t =t) 3 (F)(Ft = Ft)l-
[(F)(Ft = Ft) 3 (t = *)] 10, Equiv.
11, Simp.
12. (F)(Ft= Ft) 3(t = t)
12, 8, M.P.
13. X = X
13, UG
14. (t)(t = t)
Exercises on pages 175-176:
T h . 2. There is at most one entity 1 in C such that a H 1 = a
Proof: Suppose there are entities 1, and 12 such that « n 1, = a, a D 12 a.
Then 1, n 12 = 1, and 12 0 1, = 12. 1, H 12 = 1, n 1, by Ax « 6 .
Hence lt = 12 by two substitutions of equals for equals.
374 Solutions to Selected Exorcises on Pages 175-176
Th. 7. 0^1

Proof: Suppose 0 = 1. Then by Ax. 10 there is an « in C such that a ^ 0 and

a ^ 1. However,

«ni=an0 by assumption that 0 = 1.

a = a fl 0 Ax. 4
o=0 Th. 6

contrary to a ^ 0. Therefore, the supposition that 0=1 must be false, so 0 ^ 1.

Th. 12. a U (j8 U y) = (a U /)) U y

Proof: Lemma 1. a U (a fl ft) = a

Proof: a U (a fl ft) = (a D 1) U (a fl ft) Ax. 4

= a n (1 u ft) Ax. 8
= « n (ft U 1) Ax. 5
= a fl 1 Th. 5
= a Ax. 4

Lemma 2. «n(aUj8) = a

Proof: a H (at U /?) = (a U 0) H (a U ft) Ax. 3

= « u (0 n ft) Ax. 7
= a U (/? n 0) Ax. 6
= rt U 0 Th. 6
= tt Ax. 3

Lemma 3. If a D y = a and ft D y = tft then (a U ft) fl y = a U ft

Proof: (« U /]) fl y = y fl (« U /?) Ax. 6

= (y n «) u (y n 0) Ax. 8
= (a n y) U (/? n y) Ax. 6
= a U /? hypothesis
A. a H [a U (ft U y)] = a L. 2
p n [a U (P U yj] = (P n a) if\p n (P U 7)] Ax. 8
= (i8n«)ujS L. 2
= p u (P n a) Ax. 5
B. P n [a U (P U y)] = 0 L. 1
C. (a U 0) n [a U (0 U 7)] = a U P by L. 3 from A, B.
7 n [a U (/* U 7)] = 7 n [{P U 7) U a] Ax. 5
= [7 n {p u 7)] u (7 n a) Ax. 8
= I7 n (7 u p)] u (7 n a) Ax. 5
- 7 1) (7 n a) L. 2
D. 7 n [« U 03 U 7)] = 7 L. 1
E. [(a U ft) U y] fl [a U (ft U y)] = (a U ft) U y by L. 3 from C, D.
a 0 [(a U ft) U y] = [a fl (a U ft)) U (a fl y) Ax. 8
D

L. 2
s

?-
II
 Solutions to Selected Exercises on Pages 176-226 S7S
F. o D [(a U ft) U y] = a L. 1
P n [(« u ft) u y] = [ft n (« u ft)] u (p n y) Ax. 8
= [ft n (ft U a)] u (ft n y) Ax. 5
= pu(ftny) L. 2
G. ft n [(a U ft) U y] = ft L. 1
y ft [(a U /}) U y] = y n [y U (a U ft)] Ax. 5
II. y n [(a U ft) U yj = y L. 2
I. (ft U y) n [(« Uft)Uy\=ftL)y by L. 3 from G, H.
J. [a U (ft U y)] fl [(« U ft) U y] = n U (ft U y) by L. 3 from F, /.
« U (ft U y) = [a U (ft U y)] D [(« U ft) U yj J
= [(«Uj8)Uy]n[« U (ft U y)j Ax. 6
a U (ft U y) = (a U ft) U y E

T ii. 2 0. If a fl ft ^ 0 and p n - y = 0, then « f1 y # 0

Proof: (a n y) n ft = « n (y n ft) Th. 13
= <\ n (ft n y) Ax. (i
= (« n/ony Th. 13
= [(« n ft) n y] u o Ax. 3
= [(« n ft) n yj u (« n <>) Th. 6
= [(<* n ft) n y| u |« n (ft n -y)] hypothesis
= [(« n ft) n yi u [(« n ft) n -y] Th. 13
= « n ft Th. 11
(a n y) n ft 5* 0 hypothesis
« n y ^0 Th. 10

Exercises on pages 217-218:

5. Not a .«# 10. wff.

Exercises on page 220:

5. ~((i>)-(~(~((^)-(~(/,)»))>

Exercises on page 222:

1. f2(P,Q) is expressed as ~-(P*—(>>)
/5(P,y) is expressed as
fn(P,Q) is expressed as -~(PuQ)* ~^(Pm ~~Q)m

Exercises on pages 226-227

1. S(v,~~) is functionally complete, for it contains the same as does R.S., and the
‘•'of R.S. is definable in S(vs —> as P*(J = df ^(^Fv ^^)). Hence any truth
function expressible in R.S. is expressible in S(v,~~), whose functional completeness
now follows from that of R.S.

2. S(D,«) is functionally incomplete, for it contains no wff that can express a truth
function whose value is false when all its arguments have the value true. Proof by
strong induction on the number of symbols in the wff g(P, Q> R,...) of S( D,*).
a) If g(P, p, R,.. ) contains just one symbol, it is either P or Q or R or ...: If these
are all true, then g(P, p, R,...) cannot have the value false.
P) Assume that any wff containing <m symbols cannot be false when all its
376 Solutions to Selected Exercises on Pages 226-244

arguments are true. Now consider any wff g(P, Q, fl>...) containing exactly m
symbols (m > 1). It must be either gj(P, Q, R,.%..) D g2(Pf Q,R,...) or
g i{P, Q, K,.. .)-g2(P, QRt.. .).Butg x(Pt Qy R,.. .)andg2(P, Qy R,...) each contains
<m symbols, hence they must have the value true if all their arguments are true.
And because T D T and T*T are both T, g(P, Q,Ry...) cannot be false when
all of its arguments have the value true. Hence no wff of S( D, •) can be false when
its arguments are all truet so S(D,*) is functionally incomplete.
6. S(D, + ) is functionally complete, for the of R.S. is definable in S(D,+) as
~~P = df P -|- (P D P), and the *•* of R.S. is then definable in S(D, + ) as
P*Q = df ~~[P D (P 4- £)]. Hence any truth function expressible in R.S. is ex¬
pressible in S(D, + ), whose functional completeness now follows from that of R.S.

Exercises on page 237:

1. The same models that establish the independence of the H.A. postulates on pages
263-265 serve to establish the independence of the PN axioms.

Exercises on pages 243-244:

T h . 6. h (ft--P) D (PR)
Proof: h --P DP Th. 2
h (fl-P) D (PR) DR 3

DR 6. PDQ,QDR\-PD R
Proof PDP Th. 7
PDQ pr.
Q D R pr.
PDR DR 5

T h . 10. h (PQ)R D P(QR)

Proof: h (PQ)R D PQ Ax. 2
h PpDP Ax. 2
h (PQ)R D P DR 6
h PQDQP Th. 8
h QPDQ Ax. 2
\-PQDQ DR 6
h (PQ)R DQR DR 7, Cor. 1
h (PQ)R D P(QR) DR 8

DR 10. PDRyQDRh(PvQ)DR
Proof: PDR Pr-
QDR pr.
(PvQ) D (Rv R) DR 9
~~fl D '*-'R~*'R Ax. 1
[~~R D ~~R^R] D [~~(-^R~~R) D ~R] Th. 5
-(~~fl-R) D-~fl R 1
(fl v R) D-~R df.
-~fl D R Th. 2
(PvQ) DR DR 5
 Solutions to Selected Exercises on Pages 244*290 377
Th. 14. !-[Pp 3 R] 3 [P 3 (() 3 R)]

Proof: h ~~~~(Q~~R) 3 (Q~~R) Th. 2

h P-~(^~R) 3 P(Q~-R) DR 7. Cor. 2
1- P(()~fl) 3 (P())~R Th. 10. Cor.
^ (PQ)-R DR 6
h {P~-~~(Q~R) 3 (PQ)-~R} ^
HWHI => -[P-V-m Th. 5
^IWHI => ~[P—-«>~-R)| R 1
1- [Pp 3 fl] 3 [P 3 «? 3 R)| df.

Exercises on page 247: •f

Th. 17, Cor. \-PD(PvQ)

Proof: 1-PD ({) v P) Th. 17

f-(QvP) D(PvQ) Th. 11
\-P D (PvQ) DR 0

1)R 14. P, () h PQ

Proof: P 3 3 l\)) Th. 15

P Pi’¬
Q 3 0V) ll l

Q Pi¬
ll 1
‘V

Exercises on pages 249-250:

DU 20. (P 3 (J)‘(R 3 S), P v R h () v S'

7Voo/: (P D <;))•(« D S) pr.

P D {> DR 19
R D S DR 19, Cor.
(Pv R) D (pvS) DR 9
P\ R pr.
pv S R I

Th. .3 0. P(Qv R) ~ PQ v PR

Proof: h PQ 3 P Ax. 2
I- PR 3 P Ax. 2
|-(P<;)vPR) D P DR 10
h PP 3 ()P Th. 8
h PP 3 £> Ax. 2
1- P(1 3 0 DR 6
h PR 3 HP Th. 8
h RP 3 R Ax. 2
1- PR 3 R DR 6
f-(PpvPR) 3 (()vR) DR 9
h(PpvPH) 3 P(QvR) DR 8
l-P«DvR) 3 (^)v R)P Th. 8
378 Solutions to Selected Exercises on Pages 250-282

h (QvR)PDQPvRP Th. 18
hP(QvR) D QPyRP DR 6
\-QP=PQ Th. 21
hP(QvR) D PQvRP MT IV, Cor.
h RP=PR Th. 21
hP{QvR) D PQvPR MT IV, Cor.
hP(QvR) = PQvPR DR 14, df.

Exercises on page 276:

The H. A. system has already been proved to be a model system of logic. Because Pn
4as the same primitive symbols as H.A., it, too, is functionally complete. The
analyticity of PN is shown by using truth tables to show that all four of its axioms are
tautologies, and then showing that any wff that follows from tautologies by repeated
applications of the PN Rule must be a tautology also. Because its Axioms 1, 2, and 4
are the same as the Postulates 1, 2, and 4 of H.A., and because its Rule is the same as
the rule R' 1 of H.A., and because the primitives of the two systems are the same, we
can prove the deductive completeness of PN by deriving P 3 of H.A. from the four
Axioms of PN. First, we establish a derived rule for pN--

DR 1. PDQ, Q D R \prP D R

Demonstration: 1. (Q D R) D [(~~PvQ) D (~PvR)] Ax. 4

2. Q DR premiss
3. (-Pvp) D (—P v R) Rule
4. (PDQ)D(PD R) df.
5 .PDQ premiss
6. PD R Rule

rH. 1. |jr(Pv<?)D(evP)
Proof: 1. [Q D (pvP)] D {(PvQ) D [Pv(QvP)]) Ax . 4
2. Q D (QvP) Ax. 2
3. (PvQ)D [Pv(pvP)] Rule
4. [Py(QvP)] D [Qv(PvP)] Ax. 3
5. (PvQ) D [Qv(PvP)] DR 1
6. [(PvP) DP] D {[pv(PvP)] D {QvP)} Ax. 4
7. (PvP) DP Ax. 1
8. [Qv(PvP)] D (QyP) Rule
9. (PyQ) D(QyP) DR 1

Exercises on page 280:

1. 1. CAppp
2. CpApq
3. CApqAqp
4. CCpqCArpArq

Exercises on page 282:

2. PlP:l:PiP4.PlP:4:.PlP:l:PlP.l.PlP
6. P|e.|.P|e:|:P|P
8. PlP:l:PlP.l.PlP:.l:.PlP:l:PlP.l.PlP
 Solutions to Selected Exercises on Page* fto-343 379
Exercises on page 300:
Th. 4. I- (x)(P DQ)D [(3x)P 3 (3*)^]
Proof: (x)(P 3 Q) h (x)(P 3 Q) 3 (P 3 Q) P5
(x)(P D Q)h (x)(P 3 Q) premiss
(x)(PDQ)FPDQ R 1
(*)(PDg)h~gD~p ®
(x)(P DQ)f- (x)(~Q 3 ~P) R 2
(x)(P DQ) I- (x)-~Q 3 (x)-P DR 5
(x)(P DQ) h ~(x)~P 3 ®
(x)(P 3 p)h (3x)P 3 (3ac)() df.
I- (x)(P 3 Q) 3 [(3x)P 3 (3x)£>] D.T.

Exercises on pages 305-300:

Th. 13. h(3x)(P-Q) 3 (3x)P-(3x)p

Proof: I- [(x)~P v (x)~()] 3 (x)(~Pv ~~Q) Th. 10

I- ~(x)(~Pv -~Q) 3 ~[(x)~Pv ®
h ~(x-)~(P-p) D [~(i)~p—(x)~()] Duality
Theorem, ®, and R.R.
I- (3x)(P-Q) 3 (3x)P-(3x)() df.

Th. 21. h (x)(P 3 Q) = (3x)P 3 Q

Proof: h {x)(~~Q 3 ~P) = 3 (x)~-P Th. 20
h (x)(P 3 Q) = ~(x)~P 3 ® and R.R.
h (x)(P 3 Q) = ~(x)~P 3 y ® and R.R.
h(x)(P 3 Q) = (3x)P 3 Q df.

DR 6. P 3 £> h (3x)P 3 Q
Proof: P D Q premiss
(x)(P 3 (?) ' R 2
(3x)P 3 Q Th. 21 and R.R.

Exercises on page 313:

2. (3x)(to)(z)(3u)(3t>)(t){[F(u, v) 3 [(J(t)vW(y)]]-[[G(r)vH(i,)] 3 F(w,z)\)
4. (3y)(x)(3z)(to)[F(x, y) 3 F(z, u>)]

Exercises on page 315:

1. (3f)(x)(y)(z){{[H(x)vH(y))v H(z)}-~~\l)(t)-~D(t)]}
3. (3t)(y)(z)(3x){[G(x, y) v F(z)]-~[D(()-~D(«)]}

Exercises on page 318:

1. (3x)(3z)(y)[F(x, y)vG(z)]
3. (3u>)(3x)(3y)(t){{[F(x, j/),~[D(u)),~D(ui)]] D H(w,x)} D H(w,t))

Exercises on page 343:

l. p = p
(p.p)v(p*p)
p.p
380 Solutions to Selected Exercises on Page 343

5. p = (q*p)_ _
(P*9’P) vp-(q*p)=
(P'(I‘P)v [p*(q v P)]_
(P’9'P)v (P'9) v (p'p)
(p-q-p) v (p-q)
(r/'p-p) v(p-q)
(q-p-p^v (q-q)vjp-q)
q-[(p’p) vq]v(p-q)
q-[qv (p^pj\v (p-q)
(<T9)v(p-q)
(p-q)v(q-q)
p-q
10. p 3 Up 3 9) 3 9]
p v (p vq) v q
pv (p * q) v q
p_v(p*q)vq
[p*(q vq)] v(p-q) v [q-(p vp)J
(p-q) v (p-q) v (p-q) v (q -p) v (q-p)
(p-q) v (p-q) v (p-q) v (p-q) v (p-q)
(p-q) v (p-q) v (p-q) v (p-q) v (p-q)
(p-q)v(p-q)v(p-q)v(p-q)
[p-(q vq)] v [p-(q vq)]
pvp

15. (pvq) 3 [(p 3 q) 3(qVq)]

(p vq) v (p v g) v (q*q)
(p v q) v(p q)
(p-q) v (p-q)
(q-p) vjq-p)
q-(p v p)
<1
 SPECIAL SYMBOLS

PAGE PAGE

• 8, 216, 261, 276-279, 291 7 146

>"*W 10, 216 = df 152

V 11, 218, 261, 292 U 170

( ) 12, 215, 261, 291 n 170

D 17, 218, 261, 292 otfi 170-171, 206

= 28, 219, 261, 292 —nr, n 171

>
0, A, 0 172, 192
58, 97

1, v 172
(*) 65
c 175
(3*) 65
e 177
66
t 177
* 68
c 178
V 71, 91
{ } 178
f*
91
{*: l'x) 179
—-
141, 171, 328
<P 179
* 143, 173

MM
382
PAGE
PAGE

XsY, X ~ Y 203
ha, n<j4, n{,o£ A},
n{x: * € A}, 183
s{a) 204

UA, Uflj, U{ a: a £ A},

207-208
U{z: *€ A}, 184 £0

184 207
A - B <p(x, y)

185 q 209
A\B

210
<*>y> 185 «„ K

185 J>, <t 227

<P

186 1- 229, 294-295

A X B

188 s* 248, 271

f:X—*Y

261, 292
X~ Y 188, 215, 261, 291 V

hiA 262
u+ 192, 207

281
193 1
281
m <n 196 1

196 hi 283
Hn
D 292
m + n 196, 206, 227

198 = 292
311

199 F° 295
X Mj
296
199 ®
i

200 W* 301
B4
w4 302
R 201

311
« 202 (Q*i)

202 313, 317

< (Q\)
 an

PAGE PAGE

sc 321 345
<t>

345
K 325 t\ \
337 345
V =*

F,. *F. 346

n

f[Pi
337
i= 1 !P 348

338
i=l
INDEX

• systems and notations, see Chapter 9

Ambiguity, 12, 219
A proposition, 66-69, 170, 173, 174, 304 Analogy, logical, 19-21
a.p.f., 295 Analysis, 191
Abbreviation, 6, 15, 218, 292 Analytic, 230-231, 260, 284, 321
Abraham, 116 Analytical geometry, 164
Absorption, law of, 171 Analylicity, 230-231, 284, 296, 321
Abstract Set Theory, 176-212 And, 8, 12, 69, 116
Abstraction, 195 Anderson, Alan Boss, 260».
Aekerrnann, Wilhelm, 81 n., 260; see also And/or, 11
Hilbert-Ackcrmann system Antecedent, 16
Add., 34 Antisymmetric relations, 143, 202, 203
Addition Apodosis, 16
of cardinals, 196-197 Apparent variable, 84/i.
of classes, 175 Arbitrarily selected
of ordinals, 206 individual, 72, 84
principle of, 34, 42, 90, 102, 249 triangle, 72
Allinnative proposition Argument, 2-4, 5, 6, 162, 169, 228
particular, 66-69 and conditional statement, 29-30, 50, 53,
singular, 63 54-55, 343
universal, 66- 69 deductive, 3-4
Affirming the consequent, fallacy of, 24 extra-systematic, 160
Ajflnno, 66 inductive, 3-4
Aleph-null, L96 mathematical sense of, 220
Alexander, Linda, x Argument form, 19-25, 30
Algebra elementary valid, 33-34
Boolean, see Appendix B Aristides, 114
of classes, 170-176, 180, 182, 185 Aristotle, vii, 6, 7, 64, 117, 118
Algebraic operations on statements, see Arithmetization, 191
Appendix B Armstrong, Robert L., ix, 333u.
Alpha case, 223 Assignment of truth values, 48-49, 61-62,
Alternation, 10 78-81, 105-107, 253-257
Alternative, 10 Assoc., 39
denial, 231 Associated
interpretations, 163-164 conditional, 50
methods of proving completeness, 253- propositional formula, 295
257, 274-276, 321-328 Association
senses of completeness, 318-319 to the left, 50n., 219, 293

385
386 Index
Association (cont.) B
principles of, 39, 102, 171, 244, 249, 250-
252, 273, 337 Balboa, 27 * *
Assumption Baldwin, James Mgrk, In.
discharged, 58, 60 Ball, W. W. Rouse, 162n.
hidden, 167 Bar-Hillel, Yehoshua, 180n., 199n., 212
of limited scope, 57-59, 75, 90, 98 Bar symbol, 171, 337
Asymmetrical relations, 134-135 Base of a system, 166, 264
Atkin, Eugene, ix Bass, Walter A., ix
Attribute variables, 150 Bauer-Mengelberg, Stefan, 179n., 191n.
Attributes, 63, 320, 345 Beard, Robert W., ix
of attributes, 150-155, 345 Beaulieu, Marc, ix
of different orders, 346-347 Beaulieu, Richard, ix
of formal deductive systems, 164-166 Belfiglio, Anthony, ix
predicable and impredicable, 152-154 Beman, W. W., 191n.
of relations, 134-136, 143 Benacerraf, Paul, viiin., 195n.
and sets, 177-178 Bernays, Paul, 81n.
Aussonderung Axiom, 179-180 Bernstein, B. A., 283n.
Austin, J. L., 192n. Beta case, 223
Auswahl, Axiom der, see Choice axiom Biconditional, 29, 335-336
Aut, 11 Binary
Axiom, see Chapter 6 fractions, 201-202
of F.S., 237 functions, 222
of H.A., 261 relations, 117, 134
of L.S., 237 Binding of free variables, 90, 111; see also
of N, 283 EG, UG
of PQ, 237 Blackwood, R. T., ix
of Pn, 237 Blum, Alex, ix, 8n.
of R.S., 227 Bohnert, Herbert G., ix
of RSlf 294 Boole, George, 7, 341
of ZF, 178, 180, 181, 182, 184, 185, 193, Boolean
198, 199, 207, 211 algebra, 175-176
Abstraction, 180 expansion, 172n., 341-343
of Boolean algebra, 175 Booth, John Wilkes, 141, 146
Choice, 198, 199 Both, 13
Class algebra, 175 Bound
Comprehension, 180 least upper, 349
Empty Set, 181 occurrence of a variable, 84, 293
Extensionality 178 Brackets, 276-279; see also Parentheses
of Identity, 328 Bradley, M. C., 258n.
Infinite list, 227 * Branching theory of types, 345
of Infinity, 193 Browning, Lorin, ix
Multiplicative, 198 Brutus, 140
Pairing, 182 Burali-Forti, C., 209n.
Peano’s, 191 Burali-Forti Paradox, 209, 21 In.
Power Set, 185 Bybee, Michael, x
Reducibility, 349
Regularity, 211 c
Replacement, 207
schema, 180 C D., 34
Separation, 180 C.'N.F., 274
Specification, 180 C.P., 51, 57-58
of Subsets, 180 .cwff, 314, 322
Union, 184 Caesar, 140
Calculus of the method of deduction, 54, 257-258
extended function, 331, 350 of the natural deduction’ apparatus, 328
first-order function, see Chapter 10 Compound statement, 8-9, 63, 218
function, 259 Comprehension, 180, 190
predicate, 290n. Commutation, principles of, 39, 42, 102,
propositional, 164, 169, 176, 259; see also 171, 249, 250-252, 273, 338
Chapters 8 and 9 Conclusion, 3-5, 55, 56
sentential, 169 Condition
Campbell, Paul J., 333n. necessary, 18
Cantor, Georg, viii, 176, 177, 179, 189, 190, sufficient, 18
195, 201, 210, 349 Conditional
paradox, 190 corresponding, 29-30, 50, 54-55, 343
Canty, John Thomas, 258n. proof, 49-52, 54-55, 56-60, 250, 298, 307
Card, 195 statement, 16-18, 29-80, 49-51, 54-55,
Cardinal number, 190, 195-202, 210 56-60, 67, 218, 343
Carnap, Rudolf, 153n. Conj., 34
Carroll, Lewis, 141 Conjunct, 8
Cartesian Product, 186, 197 Conjunction, 8-9, 69, 116, 176
Castor and Pollux, 8 principle of, 34, 102, 242, 245, 247
Categorical Conjunctive normal form, 274-276, 340
proposition, 170, 173 Boolean, 341-343
syllogism, 174 Connective, 9, 13
Characteristic function, 201 Consequent, 16
Choice Axiom, 198, 199-200 Consistency, 61-62, 79n., 164-165, 231,
Church, Alonzo, 227n., 237, 266n., 279n., 295-296, 323-328
317n., 350n. Constant
Circle, vicious, 158, 160 individual, 63, 64, 85, 191, 320, 324, 326
Class, see Chapter 7 predicate, 291
addition, 175 propositional, 290
Algebra, 163, 170-176 Constructive dilemma, 34, 249, 335
complement, 171 Contingent, 28, 61-62
empty, 172, 175 Continuum hypothesis, 210
expression, 172-173 Contradiction, 28, 53, 55, 59-60, 61-62, 153,
inclusion, 170, 174 164, 176, 179-180, 214-215, 342; see
intersection, 170 also Appendix C
multiplication, 175 Contradictory, 10, 28, 61-62, 66, 68, 173
product, 170 Contraries, 66, 68
simple-term, 172 Convention
union, 170 of association to the left, 50, 219-220, 293
universal, 172 governing 12, 219, 292
Clemens, Samuel, 141 governing ‘ 219, 292
Closed, 314, 322 governing and ‘^p’, 91-92
Closure, 321 governing juxtaposition, 292
Cohen, Paul J., 180n., 200, 210, 212 governing order of precedence, 219-220,
Com., 39 292-293
Comey, D. D.t 260n. governing order of statement variables, 20,
Complement, 171, 184-185 27
absolute, 182 governing punctuation, 277
relative, 184-185 governing scopes of quantifiers, 67n.
Completeness Converse nonimplication, 227
deductive, 165-166, 231, 250, 253, 257- Conversion, 173-174
258, 260, 264, 274-276, 318-328 Coordinates, 186
expressive, 164, 218 Copi, I. M., 60n., 146n., 153n., 209n., 258n.,
functional, 218. 220-225, 281-282 331n., 333n., 344n., 350n.
388 Index

Corresponding conditional, 29-30, 50, 54-55, alternative,' 281

343 joint, 281
Corsini, Raymond, x Denumerable, 189, *332
Cortez, 27 Denying the antecedent, fallacy of, 24
Countable, 189 Derived rule of inference, 229, 238-239
Counting, 190, 194 Descartes, 64
Curl symbol, 10, 12 Description, definite, 140, 146-149
Curry, H. B., 279n. Destructive dilemma, 34, 249, 335
Deuteronomy, 15
D Dhar, Suchitra, ix
Dilemma, 34, 249, 335
D.D., 34 Diogenes, 177
D.N., 39, 40 Discharged, 58
D.S., 33, 34 Disjoint, 166
D.T., 245-247, 298-300, 322 Disjunct, 10
Dagger function, 281-282 Disjunction, 10-14, 176, 218, 227
Dangling variable, 87 Disjunctive normal form, 258, 340
Decision problem, 257 Boolean, 341-’343
Dedekind, Richard, 179n., 190, 191 Disjunctive syllogism, 11-12, 16, 20-21, 33,
Deduction, 157-158 34, 102, 249
method of, 257-258; see also Chapter 3 Dist., 39
natural, 71-75, 89-100, 113-114, 307-311, Distribution, principles of, 39, 42, 102, 171,
328 172, 247, 250, 252-253, 337, 338
theorem, 242-243, 245-247, 298-300, 322 Dodgson, Charles Lutwidge, 141
Deductive Dom, 187
argument, 3-4 Domain, 187
completeness, 165-166, 183, 250, 253, Donnellan, Keith, 146n.
257-258, 260, 264, 274-276, 318-328 Dot symbol, 8-9, 219, 276-279
system, see Chapter 6 Double negation, principle of, 29, 39, 40,
Defined terms, 159-160, 260, 291-292 102, 171, 241, 242, 247
Definite description, 140, 146-149 Doubleton, 178
Definition, 157-158, 159-160 Dual, 302
analytical, 334 Duality, 302-305
circular, 158 Dyadic, see Binary
contextual, 147
explicit, 146-147, 191 E
formal, 152
of identity, 152, 178, 187, 349 E proposition, 66-69, 170, 173, 174, 304
implicit, 190-191 EG, 73, 75, 90, 94-95, 114, 130-131, 307
inductive, 216 El, 73-75, 84, 89, 90, 96-99, 1J4, 130-131,
of logic, 1 ’ 308-310
of material implication, 17, 29 Ecclesiastes, 15, 71
recursive, 216-217, 261, 283, 293, 331 Edwards, James, ix
in a system, 332 Effective, 22, 41, 217, 227, 230, 257, 293
in use, 147 Einstein, 161
De M., 39 Either, 12, 13
Demonstration, 108-114, 229, 230, 245, 262, Elementary valid argument form, 33-34
294-295 Elimination, 42
in H.A., 262 Eliot, George, 142
in R.S., 229 Else, 143-144
in RSt, 294-295 Eminhizer, Earl Eugene, ix
De Morgan, Augustus, 7, 29, 116, 304 Empty
theorems, 29, 39, 102, 172, 250, 304, 337, . ‘class, 172, 175
338 set, 180-181, 192
Denial, 10, 337; see also Negation Enderton, H. B., 191n., 212
Enthymeme, 137-139 function, 346-347, 349
Enumerable, 189 function calculus, see Chapter 10
Enumeration, 189 proposition, 347-348
Epistemological paradoxes, 344 Fid, 187
Equinumerous, 188, 192 Form
Equiv., 40 argument, 19-25, 30, 343
Equivalence Boolean normal, 341-343
formal, 349 elementary valid argument, 33-34
logical, 29, 39, 40, 78-81, 96, 105-107, normal, 311-318, 340-343
121-122 prenex normal, 311
material, 28-29, 40, 176, 219, 250 Skolem normal, 315
Equivalent, 188 specific—of an argument, 21, 25, 34
systems, 260 specific—of a statement, 27
Esau, 15 statement, 27-30
Euclid, 52, 158-159, 161, 162 of valid categorical syllogism, 174
Euclidean geometry, 158-162, 166, 168, Formal
259-260 criterion, 168
Evans, Mary Ann, 142 deductive system, 162-166, 175, 295, 297
Exceptive statements, 143 definition, 152
Excluded middle, 241 equivalence, 349
Exclusive nature of validity, 20
classes, 173 proof of validity, 32-34, 39-42, 50, 59,
disjunction, 11, 13-14, 227 89-90, 228
Exhaustive classes, 173 truth, 27
Existential Formula, 16,3-164, 165, 166-169, 215, 293
generalization, 73, 75, 90, 94-95; see also associated propositional, 295
EG prenex normal form, 311
instantiation, 73-75, 84, 89, 90, 96-99, see Skolem normal form, 315
also El syntactical, 227
quantification, 65, 86 of type R, 315
quantifier, 65 well formed, 167-169, 215-226, 261, 283,
Exp., 40 293, 331
Expansion (Boolean), 172n., 341-343 Foulkes, Paul, 60n.
Exponentiation, 200-202, 207, 210 Fraenkel, Abraham, 153n., 179, 180n., 199n.,
Exportation, principle of, 40, 42, 49-50, 245, 210n., 212
250 Free occurrence of a variable, 84, 85, 111—
Expressive completeness, 164, 218 112, 180, 294, 320
Extended function calculus, 331, 350 Freeing of bound variables, 90; see also El,
Extensionality, principle of, 39; see also UI
Replacement, rule of Frege, G., 7, 146n., 149n., 182, 192, 194,
Ezekiel, 130 195, 204, 237, 331
Function, 187-188, 208
binary, 222

F calculus, 259; see also Chapter 10

characteristic, 201
F.S., 237 dyadic, 222
Factors, 251, 255 monadic, 220
Fallacy order of, 346
of affirming the consequent, 24 propositional, 64, 67, 84-85, 89-92, 118,
of denying the antecedent, 24 180
Feigl, H., 353n. singulary, 220
Field, 187 ternary, 222
Finite, 41, 184, 188, 190, 215n., 322 triadic, 222
First-order truth, 220-223
attribute, 349 unary, 220
Functional completeness, 218, 220-225, 260, I
281-282
I proposition, 66-^60, 170, 173-174, 304
G I.P., 53
Id., 141-143
Galileo, 157, 188 Idempotence, 171
Gamma, 322 Identity, 135, 140-149, 328-331
Gardner, Martin, 341n. axiom for, 328
Gcach, P. T., ix, 283n. definition of, 152, 178, 187, 349
General proposition, 64-69, 78, 83-88 of indiscernibles, 141, 178
Generality principle of, 141-143, 178, 330
advantage of, 163 of sets, 177-178
limited, 124 If, see Implication
unlimited, 123 Immediate inference, 173-174
Generalization, 64 Impl., 40
existential, 73, 94-95; see also EG Implicans, 16
universal, 72-73, 99-100; see also UG Implicate, 16
Genesis, 15, 130 Implication, 16-18, 22, 28, 68
Gentzen, Gerhard, 71n. material, 17, 21-24, 28, 40, 250
Geometry, 72, 158-164, 166, 259 Implicative statement, 16
Godel, Kurt, 200, 210, 319n., 331n. Implicit
Gotlind, 237 definition, 190
Gould, J. A., ix, I46n., 331n. relations, 122-123
Grant, 116 Implies, 16
Grelling, K., 344, 347, 350, 352 Impredicable, 152-153, 154, 344
Grounded set, 211 Inclusion, 170, 174, 178
Inclusive disjunction, 11
H Incompleteness
deductive, 264, 330-331
H.A., 261 functional, 226
H.S., 33, 34 of a set of rules, 333-336
Halmos, Paul R., ix, 180, 212 Inconsistent, 61-62, 79n., 164-165, 318, 323;
Hausdorff, Felix, 212 see also Contradiction
Heath, Thomas L., 161n., 162n. Independence
Hebrews, 71 of axioms, 161, 165, 231-237, 262-264
Henkin, Leon, 319n. of a tautology, 264-266
Henry, O., 141 Independent variable, 220
Hereditary, 231-237, 265, 333-336 Indirect proof, 52-54, 55, 59-60, 61-62, 250
Heterological, see Appendix C Indiscernibles, identity of, 141, 178
Hierarchy , Individual, 63, 182, 291, 320, 326, 345
of logical types, 153, 345-346 arbitrarily selected, 72, 84
of orders, 346 constant, 63, 64, 85, 291, 320, 324, 326
of propositions, 347-348 symbol, 291
Hilbert, David, viii, 210 variable, 64, 84-85, 291
Hilbert-Ackermann system, 260-276, 284 Induction
History mathematical, 185, 204, 223-226
of logic, 7 strong, 204, 225-226
of mathematics, 158-164 transfinite, 204
Hiz, Henry, 266n. weak, 223-225
Horseshoe symbol, 17 Inductive
Huntington, E. V., 175n. argument, 3
Hypothetical .definition, 216
statement, 16 • set, 193
syllogism, 22-24, 33, 34, 42, 243-244 Inference, 2, 6-7
 Index 391
derived rule of, 229 K
immediate, 173-174
involving propositional functions, 89-92 Kepler, 157
mediate, 174 Kulathungam, Lyman C. D. K., 60n.
rules of, 33-34, 39-42, 71, 102, 228, 261, Kung, G., 260n.
283, 294 Kuratowski, C\, L86n.
Inferring, 1-2
Infinite, 184, 188, 190, 205, 322
list of axioms, 227 L
Infinity, Axiom of, 193
Initial segment, 204 L.S., 237
Initially placed, 311 Lambda, 322
Instance, substitution, 20, 27, 33-34, 64, 67, Lambros, Charles H., 60n.
71, 81-85 Landau, Edmund, 19ln. *
Instantiation, 64 Langford, C. H., 279n.
existential, 73-75, 96-99; see also El Language, 5-6, 168-169, 213-215
universal, 71, 93-94, 148-149; see also UI levels of, 213-215, 350-353
Interchange of bound variables, 307 natural, 5
Interpretation, 163-164, 165, 213-214, 215, object, 213-215, 350-653
218, 220, 221, 320, 326, 334 ordinary, 6
alternative, 163-164 Syntax, 213, 214, 227
normal, 218, 220, 221, 228, 230, 238, 254, Larger than, 189
261, 320, 321, 326 Leach, Gary, ix
proof of consistency, 165 Least upper bound, 349
proof of independence, 165 Lee, J. M., 60n.
of a system, 320 Leibniz, 141, 178
Intersection, 170, 183 Lemma, 193
Intransitive relations, 135 Levels of language, 213-215, 350-353
Invalid, 3, 21 Lewis, C. I., 279n.
Invalidity Liar paradox, 214; see also Appendix C
proof of, 20, 24-25, 48-49, 78-81, 105- Limited
107, 121-122, 136-137, 232-236 generality, 124
and relational arguments, 136-137 scope of an assumption, 57-59, 73, 75, 90,
Iota, 146 96-98
Irreflective relations, 136 Lincoln, 116, 126, 141, 146
Irrelevant, 53 Lobachevsky, 161, 259
Isaac, 116 Logic
Isaiah, 15 definition, 1
Isomorphic, 204 science of, 157
study of, 1
symbolic, 5-7
J task of deductive, 3
Logical
Jacob, 15 analogy, 19-21
Jaskowski, Stanislaw, 71n. equivalence, 29, 39, 40, 78-81, 96, 165-
Jech, Thomas J., 212 107, 121-122
Job,, 15 proof, 167
Johnson, Dr. Samuel, 289 sum, 170, 252
Joint denial, 281 terms, 167
Jonathan, 15 truth, 79n., 108-115, 120, 319, 321
J0rgensen, J., 282n. types, 153-154; see also Appendix C
Jourdain, P. E. B., 176n. Logistic system, 166-169, 175; see also
Justification, 33, 41, 239 . Chapters 8, 9, 10, passim
Juxtaposition, 292 Lowell, Percival, 9
392 Index
Lower function calculus, 290n. Multiply general propositions, 83-88
Lukasiewicz, J., 214n., 237, 259, 279, 230, Mutually exclusive, 166
283n.
Luke, 71 N

M N, 282-289
Nagel, Ernest, 331n.
M.P., 33, 34, 228 Napoleon, 155
M.T., 33, 34, 249 Natural deduction, 71n., 114, 307-311, 328
Many-one, 187 Natural numbers, 188, 190-195
Many-valued logics, 259 Negation, 10, 13, 65, 176, 337
Mapping, 190 double, 29, 40, 102, 171, 241-242, 247,
Mars, 9 249, 250, 337
Marsh, Robert Charles, 153n. quantifier, 109-110
Martin, James A., 8n. Negative proposition
Material particular, 66
equivalence, 28, 40, 176, 219, 250 singular, 64
implication, 17, 21-24, 28, 40, 102, 250 universal, 66
nonimplication, 227 nEgO, 66
Mathematical induction, 185, 204, 223-226, Neither, 13
349 Nelson, L., 344n.
Matrix, 313 Newman, James R., 33 In.
Matthew, 130 Newton, Isaac, 157
Max, 334 Nicod, J. G. P., 282-289
Maximal consistent set, 325 Nondenumerable, 189, 210, 332
Mediate inference, 174 Non-Euclidean geometry, 161-162, 166
Membership, 177 Nonimplication
Metalanguage, 213-215, 350-353 converse, 227
semantical, 213-215 material, 227
syntactical, 213-214 Nonreflexive relations, 136
Metatheorem, 223 Nonsymmetrical relations, 135
Meyers, C. Mason, 8n. Nontransitive relations, 135
Micah, 71 Normal form, 311-318, see also Appendix B
Middle, excluded, 241 Boolean, 341-343
Min, 334 conjunctive, 274-276, 340
Minogue, Brendan P., ix disjunctive, 258, 340
Mirimanoff, D., 211 prenex, 311
Model, 78-81, 117, 105-107, 121-122, 136- Skolem, 315
137, 232-236, 320-322, 326 of type R, 315
three-element, 232-237, 262-263, 334 ^ Normal interpretation, 218, 220, 221, 228,
four-element, 234, 264 230, 238, 254, 261, 320, 321, 326
six-element, 265 Nu, 71, 91, 141n.
system of logic, 260, 276, 284 Null-class, 181
Modus Number, 190, 191, 192
Ponens, 22-24, 33-34, 40-41, 89, 224, 225, cardinal, 190, 195-202, 210
228, 249, 283-284 finite, 78, 136-137, 184, 185, 188, 190,
Tollens, 22-24, 32-34, 40-41, 249 195, 196-197, 205, 322
Monadic function, 220 infinite, 184, 188, 190, 196-197, 205, 227,
Montague, R., 21 In. 322
Mu, 91, 130, 141n. , natural, 189, 190-195
Multiplication ordinal, 190, 202-210
cardinal, 197-199 . transfinite, 205, 208, 209
ordinal, 206 • Numerals, Arabic and Roman, 6-7
Multiplicative Axiom, see Choice Axiom Numerical propositions, 145
 Index 393
0 Phi, 65, 68, 90-92
Pi, 337-3.38
O proposition, 66-69, 170, 173-174, 304 Place marker, 64
Object, 177 Plato, 116, 117, 118, 151
language, 213-215, 227, 350-353 Plus symbols, 192, 227
Obverse, 173 Poincar£, Henri, 157
One-to-one, 188 Polish notation, 279-280
Operationalism, 332 Pollux, 8
Operators, 215, 223, 279-280 Ponewiy see Modus Ponens
stroke and dagger, 281-282 Porte, Jean, 266n.
Opposition, Square of, 66, 68, 304 Porter, William Sidney, 141
Or, 10-14, 16 Possible nonempty universe, 78-81, 319-320,
Order, 185-186, 190 321
of functions, 346 Post, E. L., 161, 259
partial, 202-203 criterion for consistency, 164, 231, 296
of precedence, 219, 277, 292-293 Postulate, sec Axiom
of propositions, 347-348 Power set, 185, 189-190
of quantifiers, 121 Axiom, 185
serial, 196 Precedence, order of, 219, 277, 292-293 ■
simple, 196, 203, 201 Predicable, 152
total, 196, 203 Predicate
type, 204 calculus, 290m.
of words, 13 constant, 150, 291
Ordinal number, 190, 202-210 symbol, 291, 320, 326
term, 63, 170
P variable, 1.50, 291
Prefix, 313
P0, 237 Premiss, 3
PN, 237 suppressed, 137
Pair set, 178, 192 understood, 229
ordered, 185 Prenex normal form, 311
unordered, 182 Primary occurrence, 149
Paradox Primitive
Cantor, 190 operators, 223, 281
epistemological, 344 symbols, 215, 260, 290
Crclliug, 344, 347, 3.50, 352 terms, 166, 260
heterological, 344 Principal Matheuuitica, 7, J53».. 260, 331 n.,
liar, 214. 348 347n., 348ri., 349-350, 352
logical, 344 Prior, A. N., 283n.
of material implication, 28 Proclus, 161
Mirimanoff, 211 Product of classes, 170
Russell, 152-153, 179, 180 Proof
semantical, 344 of analyticity, 230-231, 284, 296, 321
Skolem, 332 circular, 158
Parallel postulate, 160-161, 166 conditional, 49-52, 54-55, 56-60, 250, 298,
Parentheses, 12, 64n., 67, 219, 276-279, 307
292-293 of consistency, 164-165, 231, 295-296
Parenthesis-free notation, 279-280 of deductive completeness, 253-257, 264,
Part, well formed, 293, 311 274-276, 318-328
Particular propositions, 66-68 of deductive incompleteness, 264-265
Peano, Guiseppe, 7, 191, 192, 193, 194 formal 32-34, 39-42, 50, 59, 89-90, 228
Peirce, Charles, 1, 7, 281 n. of functional completeness, 220-225, 281-
law, 62 282
Pfeiffer, John E., 341 n. of functional incompleteness, 226
394 Index
Proof (cont.) universal, 65, 85-86
of incompleteness of rules, 333-336 words, 126
of independence of axioms, 165, 231-237, Quantifier
262-266 existential, 65
indirect, 52-54, 55, 59-60, 61-62, 250 initially placed, 311
of invalidity, 20, 24-25, 48-49, 78-81, negation, 109-110
105-107 order of, 121
in R.S., 240-241 rank, 315
reductio ad absurdum, 52-54, 55, 60n., scope, 67, 84, 86-87, 111, 311
61-62, 161 universal, 65
shorter, 52, 61-62, 102 Quaternary relations, 117
of tautology, 54-55, 61-62 Quine, W. V. O., 194n., 21 In., 212, 281n.,
of validity, 32-34, 39-42, 61-62 283n., 328n.
versus demonstration, 240-241
Proposition, 2, 5; see also Statement
categorical, 170, 173-174 R
general, 64-69, 78
multiply general, 83-88 R.R., 301
negative, 64, 66 R.S., 218, 264
numerical, 145 RSlf 290
orders of, 347-348 RS/, 324
particular, 66-69 Ramified theory of types, see Appendix C
relational, 116-127 Ramsey, F. P., 344, 345, 348n., 351n.
singly general, 83 Ran, 187
singular, 63-64, 84 Range, 187
subject-predicate; 66-69, 170 Rank, 315
universal, 66 Rankin, Terry L., ix
Propositional Rasiowa, 237
calculus, 164, 169, 176, 259; see also Real variable, 84n.
Chapters 8, 9 Reasoning, 1-2
constant, 290 Recursive
function, 64, 67, 84-55, 89-92, 118 definition, 216-217, 261, 283, 293, 331
symbol, 215, 290, 320, 326 rule, 261, 283, 293
variable, 216, 290 Reductibility axiom, 349-350, 353
Protasis, 16 Reductio ad absurdum, 52-54, 55, 60n., 61-
Proverbs, 15, 71, 130 62, 161
Psalmsy 15, 71, 130 Redundance, 40-41, 165
Psiy 68, 90-92 in punctuation, 277
Psychology, 1 Referring, 145, 146, 148
Punctuation, 12-13, 67, 219, 276-279, 292- Reflexive relations, 135-136, 202
293 Refutation by logical analogy, 19-21
Putnam, Hilary, viii, 146n. Regression, 158, 160
Pythagoras, 158-159 Regularity Axiom, 211
Relations, see Chapter 5
many-one, 187
Q in set theory, 185-187
Relative complement, 184-185
QN, 109-110 Replacement
Quantification, 64-69, 85-86, 90, 150-155, Axiom Schema of, 207
259; see also Chapter 10 mle of, 39-40, 248, 271, 301
existential, 65, 85-86 Representation, 253-254
order of, 121 Representative, 192, 195-196, 209-210
rules, 71-76, 89-100, 113-114; see also Restricted function calculus, 290n.
EG, EX, UG, UI Riemann, 161, 259
 Index 395
Rigor, 162-163, 166-168 function, 346-347
Ring, Elizabeth Muir, ix proposition, 347-348
Robinson, R. M., 205n. Self-evidence, 160-161
Rockefeller, 5, 20 Sellars, W., 353n.
Romans, 130 Semantical, 168-169, 213, 218n., 220
Rosser, J. Barkley, 218, 259 paradoxes, 214-215; see also Appendix C
Rule terms, 349
of conditional proof, 49-52, 54-55, 56-60, Sentence, 2, 5
250, 298, 307 Sentential calculus, see Calculus,
derived, 229, 238-239 propositional
of indirect proof, 52-54, 55, 59-60, 61-62, Separation, 180, 190
250 Serial order, 196
of inference, 33-34, 39-42, 72-75, 89-100, Set, 176-212
102, 113-114, 228, 261-262, 283, 294, empty, 180-182
333-336 grounded, 211
of logic, 167 inductive, 193
quantification, 71-76, 89-100, 113-114; see infinite, 209
also EG, El, UG, UI selection, 198
recursive, 216, 283, 293 successor, 191, 192, 205
of replacement, 39-40, 248, 271, 301 transitive, 194
of thumb, 42, 159 well ordered, 203
Russell, Bertrand, 78, 146, 147, 149, 153, Set Theory, see Chapter 7
179, 181, 182, 192, 194, 195, 198, 199n., as a first-order theory, 331-332
204, 211, 260, 348, 350n., 353 Shannon, Claude E., 341n.
Russell paradox, 152-153, 179, 180 Sheffer, II. M., 281n., 282n.
Shoes and socks illustration, 198-199
s Shotwell, Jennifer M., ix
Shukla, Anjan, 266n.
Saccheri, Gerolami, 161 Sigma, 337-338
Samuel, 15 Similar, 203
Sanford, David H., 9n. Simons, Leo, 333n.
Satisfiable, 321 Simp., 34
simultaneously, 324 ' Simple
Satisfy, 179-180, 321, 332 class terms, 172
Saul, 15 order, 196, 203
Scharle, Thomas W., 266n., 283n. propositional functions, 79n.
Schema, 180, 207, 223, 225 statement, 8-9, 78
Scherer, Donald, 60n. theory of types, 153-154; see also
Schdnfinkel, Moses, 81n. Appendix C
Science Simplification, principle of, 34, 41, 42, 90,
of geometry, 158-164, 166, 259-260 102, 249
as organized knowledge, 157 Simultaneously satisfiable, 324
of physics, 157, 163 Singleton, 182
of reasoning, 1 Singly general proposition, 83
Scope Singulary proposition, 63-64, 84, 112
of an assumption, 57-59, 73, 75, 96, 99- Singular function, 220
100 Size, 195, 196
indicator, 149 Skolem, Thoralf A., 179n., 332
of operator symbols, 219, 292 normal form, 315
of a quantifier, 67, 84, 86-87, 111, 311 paradox, 332
Scott, Kenneth J., x Smaller than, 189
Scott, Sir Walter, 140, 146 Smith, Ronald C., ix
Secondary occurrence, 149 Socrates, 3, 63, 64, 71, 75, 95, 116, 117, 118,
Second-order 151
396 Index
Solomon, Song of, 15 199n„ 212
Sound, 5 Syllogism * •
Specific form categorical, 174
of an argument, 21, 25, 34, 48 disjunctive, 11-12, 16, 20-21, 33, 34, 102,
of a statement, 27 249
Spinoza, 159 hypothetical, 22-24, 33, 34, 243-244, 249,
Square of opposition, 66, 68, 304 335
Statement, 2, 5; see also Proposition Symbols, 6
arithmetical, 145 advantages of, 6-7
component, 8 operator, 215
conditional, 16-J8, 29-30 primitive, 215
contingent, 28, 62 propositional, 215
contradictory, 28, 62 quantifier, see Chapter 4
exceptive, 143 relational, see Chapter 5
form, 27-30 truth-functional, see Chapter 2
hypothetical, 16 uninterpreted, 163, 215
implicative, 16 Symmetrical relations, 134
simple and compound, 8-9, 78, 218 Syntactical, 168-169, 213
specific form of a, 27 formula, 227
tautologous, 27, 62 variable, 244
truth-functional component, 9 Syntax language, 213-214, 227
truth-fiinctionally compound, 9 System
variable, 20 deductive, see Chapter 6
Stoll, Robert R., ix, 176n., 182, 195n., 212 logistic, 166-169, 223; see also Chapters 8,
Strengthened rule of conditional proof, 9, 10, passim
56-60
Stroke function, 281-282
Strong T
disjunction, 11, 218
induction, 225-226 Table
Subclass, 173 three-valued, 232-236, 263, 334
Subcontraries, 66, 68 four-valued, 264
Subject term, 63, 84, 170 six-valued, 265
Subject-predicate propositions, 22, 66-69, truth, 10, 17, 19-24, 27-30, 32, 41, 48-49,
170 61-62, 108, 220-225, 253-254, 257
Subset, 178, 180, 185, 322, 325 Tapscott, Bangs 1.., ix
Axiom, 180 Tarski, A., 353n.
proper, 178 Taut., 40
Substitution, 42, 85, 176 Tautologous, 27-28, 55, 61-62
different from replacement, 42 Tautology, 27-28, 55, 232, 275-276, 295,
instance, 20, 27, 33-34, 04, 67, 71, 84-86, 296, 319, 342
90 principles of, 40, 171, 250, 338
rule for functional variables, 317n. proof of, 54-55, 61-62
Successor, 191, 192, 205 and validity, 29-30, 49-50, 54-55, 343
such that, 179 Temporal words, 126
Sum Terms, 63-64, 157-158, 166, 170, 172
cardinal, 196 defined, 154-155
logical, 170, 252 logical, 167
ordinal, 205-206 predicate, 63, 170
set-theoretic, 170, 183, 192, 196 primitive, 166
Summand, 252, 255 relative, 214
Superlative, 148 semantical, 349-350
Suppes, Patrick, 179n., 180n., 182n., 191n., simple class, 172
 Index 397
subject, 63, 170 u
undefined, 159-160, 164, 166, 293
Ternary UG, 72-73, 75, 90, 99-100, 114, 130-131,
functions, 222 307-308
relations, 117 UI, 71, 75, 90, 93-94, 113, 130-131, 307
Tetradic relations, 117 Unary function, 220
The, 132, 146-149 Undefined terms, 159-160, 164, 166, 293
Theorem, 229, see also Chapter 6 Union, 170, 183
Thing, 63 Axiom, 184
Thomas, Norman L., ix Unit set, 182
Three-valued Universal
logic, 259 affirmative, 66-68
tables, 232-236, 263, 334 class, 172, 175
Tilde, 10 generalization, 72-73, 75, 90, 99-100; see
Timothy, 71 also UG
Todhunter, 159n. instantiation, 71, 75, 90, 93-94, 146, 148;
Tollens, see Modus Tollens see also UI
Total order, 196, 203 negative, 66-68
Totally reflexive relation, 135 proposition, 66-67
Townsend, Burke, ix quantification, 65, 85-86, 345n.
Trans., 40 quantifier, 65, 293
Transfinite set, 180
induction, 204 Universe
numbers, 205, 208, 209 possible nonempty, 78-81, 319-321
Transitive proof of its nonexistence, 180
relations, 135, 202 Unless, 13
set, 194 Unordered, 182
Translation, 69, 124-127
Transposition, principle of, 40, 174, 241,
242, 250 V
Triadic, see Ternary
Truth, 4-5, 320, 326-327 Valid, 3, 4n., 20, 21, 24-25, .33, 79, 80-81,
extrasystematic, 162 89-90, 169, 174, 321
formal, 27 categorical syllogisms, 321
function, 220-223 and tautologous, 29-30, 49-50, 55, 343
-functional component, 9 Validity, 45, 63, 321
-functional connective, 9, 13 conditional proof of, 49-52, 54-55, 56-60,
functionally compound statement, 9, 14, 250, 298, 307
39 formal nature of, 20
hereditary, 231-232, 333 formal proof of, 32-34, 39-42, 50, 59, 89-
logical, 79n., 108-115, 120, 319, 321 90, 228
necessary, 27 indirect proof of, 52-54, 55, 59-60, 61-62,
self-evident, 160-161 250
table, 10, 17, 19-24, 27-30, 32, 41, 48-49, semantical and syntactical, 169, 213
61-62, 108, 220-225, 253-254, 257 strengthened conditional proof of, 56-60
unconditional, 55 and tautology, 55, 343
value, 9, 48-49, 61-62, 320, 326-327 Value, 188, £34
Turing, A. M., 279n. Van Heijenoort, Jean, 179n., 190n., 191 n.,
Turquette, A. R., 259 192n„ 198n., 209n., 331n.
Twain, Mark, 141 Variable
Two-valued logic, 259 apparent, S4n.
Type R, 315 attribute, 150
Types, logical, 153-154; see also Appendix C binding, 90, 111; see also EG, UG
Variable (cont.) Weak
bound occurrence, 84, 293 disjunction, 11, 176, 218
dangling, 87 induction, 223-^25'
free occurrence, 84, 85, 111-112, 180, 294, Wedge symbol, 11-12, 218, 219, 261, 292
320 Well formed
freeing, 90; see also El, U1 formula, 167-169, 215-226, 261, 283, 293,
functional, 317n. 331
independent, 220 part, 293, 311
individual, 64, 84-85, 291 Wetterau, Elaine, x
interchange of bound, 307 Whitehead, Alfred North, 7, 153n., 182n.,
predicate, 150, 291 260, 348
propositional, 216, 290 Wiener, N., 186n.
real, 84n. Wilder, Raymond L., 190n., 204n.
relation, 150 Williams, Frank, ix
statement, 20, 290 Wittgenstein, L., 350n.
syntactical, 292
Vee symbol, 11-12, 218, 219, 261, 292
Vel, 11
Vicious circle, 158, 160
Y
principle, 348
Yields symbol, 229, 242, 262, 283, 294-295
Von Neumann, John, 192, 194, 205n.
Yuting, S., 21 In.

z
w
ZF, 176-212
wff, 168-169, 215-226, 234, 261, 283, 293, Zermelo, Ernst, 153n., 179, 180, 181, 192,
331 198, 209
closed, 314, 322 Zerinelo-Fraenkel Set Theory, vii, 176-212
Wajsberg, 283n. Zero, 191, 192, 193
Wang, Hao, 191n., 328n. Zinov’ev, A. A., 260n.