UNIVERSITY OF N.S.

'i, LIBRARIES
S164/123

11\\1111111\1\111\\111\\1\111\\1\1111\1\1\11\1\1\1\\111\111\11\\1

>001542931
 IN R u N
TO Lie

James D. Carney
A rizona State University

Prentice-Hall, Inc.
Englewood Oliffs, New Jersey
"
 PREFACE

This book is an effort at providing an introductory text to symbolic

PRENTICE-HALL INTERNATIONAL, INC., London
logic on the university level. No prior knowledge of either philosophy or
PRENTICE-HALL OF AUSTRALIA, PTY. LTD., Sydney
mathematics is assumed. It is hoped that sections have been so set up
PREN'fICE-HALL OF CANADA, LTD., Toronto D lh' with their accompanying exercises that a diligent and intelligent reader
PRENTICE-HALL OF INDIA PRIVATE LIMITED, New e ~
will find the book to a large extent self-teaching. To aid in this end,
PRENTICE-HALL OF JAPAN, INC.,Tokyo
partial answers to the exercises are supplied with most sets of exercises.
The aim of this book is to provide a beginning student with a
working knowledge of the foundations of modern logic-the lower
predicate calculus with identity. Accordingly, Part I emphasizes the
© 1970 by formal language, transformations, proof construction in a natural
PRENTICE-HALL, INC. deduction system, and translation. I have tried to develop the natural
Englewood Cliffs, N.J. deduction system for predicate logic as formally as is in keeping with
the general aims of the book. The natural deduction system for the
sentential calculus is an adaptation of that found in E. J. Lemmon's
Beginning Logic, which is, in turn, an adaptation of Gentzen's 1934-35
All rights reserved. paper. The quantification rules of the predicate calculus are the stan-
No part of this book may be reproduced dard, non-independent four. To simplify these rules somewhat, I have
in any form or by any means . stated them in terms of limited and non-limited individual constants.
without permission in writing from the publIsher. Neither system has independent rules, but sacrificing simplicity leads to
easier development of the system. Of special interest is Beth's semantic
tableaux method for constructing proofs in logic. This is introduced in
three exercise sections. The first section (at the end of Chapter Four)
Library of Congress Catalog Card No.: 7l-97926 deals with sentential logic, the second (at the end of Chapter Six) deals
with predicate logic, and the last (at the end of Chapter Seven) deals
with predicate logic with identity. The rules and procedures for con-
13-498709-8 structing trees are taken from Richard C. Jeffrey's Formal Logic: Its
Scope and Limits.
Current printing (last digit) Part II emphasizes logical axiom systems and metalogic, stopping
short of the completeness proof for the predicate calculus. The com-
10 9 8 7 6 5 4 3 2 1
pleteness proof of the predicate calculus is regarded as the demarcation
point between beginning logic and advanced logic. Three axiom systems
are presented in Part II: one for syllogistics, taken from .}:Jukasiewicz;
Printed in the United States of America the sentential calculus, found in Principia; and an infinite axiom system
vi Preface

for the predicate calculus, taken from Ohurch. Very few theorems are
derived, though the exercises call on the reader to derive a legion of
theorems. In the development of the sentential calculus little more is
done than deriving the machinery to prove the replacement theorem.
The development of the axiom system for the predicate calculus is
aimed at helping the student become acquainted with axiom schemes
and what it is like to construct scheme proofs.
For some semester courses Part I is all that can be covered,
especially if most of the exercises are done. But courses vary; many
TABLE OF
will find time to do some or all of Part II. Not only is it fruitless to work
each and every exercise, but it is also advisable to omit some sets of
CONTENTS
exercises within sections. From time to time, usually at the end of an
exercise section, exercises are introduced which go beyond the analysis
in the section. These exercises can be omitted without the loss of
continuity.
The manner of presentation in the book owes a great deal to Margaris, Part I: NATURAL
Ohurch, Lemmon, Mates, and Hughes and Londey (see Further DEDUCTION SYSTEMS
Readings). For ideas for notation and organization, I am especially
indebted to Angelo Margaris' First Orde'r Mathematical Logic.
Chapter 1: Validity
I wish to express my appreciation to those who helped in the
preparation of the manuscript: Alan Lesure, the understanding editor,
1.1 Arguments
Professor Phillip Von Bretzel and Lt. Walter Jones, who did some of the 3
1.2 Necessarily true statements
proofreading, Mrs. Ruth Bardrick, for typing assistance, and my wife, 6
1.3 Valid logical formulas
who has been able to raise a family, teach French, and do expert typing 9
1.4 Formal language
on a difficult manuscript, all at the same time. 13
1.5 The use/mention distinction
14
Chapter 2: Sentential Language
2.1 Conjunction and denial
2.2 16
Other statement connectives
2.3 19
Determining the truth-value of compound statements 21
2.4 Stat~ment connectives and the natural language
2.5 24
Puttmg the natural language into symbols 28
2.6 Formation rules for the sentential language
31
Chapter 3: Decision Procedures
3.1 Truth-tables
3.2 36
Valid, inconsistent, and contingent formulas
3.3 38
Arguments and truth-tables
3.4 40
A short-cut truth-table method
3.5 45
Logical equivalence and transformations
3.6 47
Other two-place operators
3.7 54
Normal forms and testing for validity
57
Supplement: Logical networks
61
vii
viii Table of Contents Table of Contents
ix
Chapter 4: Natural Deduction System SC

4.1 Rules of inference and proofs 68

4.2 Four additional rules of inference 72
Part II: LOGICAL
4.3 Assumption discharging rules 76 AXIOMATIC SYSTEMS
4.4 Useful strategies 83
4.5 System SO 86
4.6 Some theorems of SO Chapter 8: Formal Axiom Systems
89
4.7 Two derived rules 91
4.8 Soundness, consistency, and completeness of SO 8.1 The development of geometry
97 186
8.2 Formal axiomatic theories
187
8.3 System CS
190
8.4 Metalogical properties of CS
Chapter 5: Predicate Language 194
8.5 Axiom systems and logic
199
5.1 Predicates 102 Chapter 9: Axioms jor the Sentential Calculus
5.2 Quantifiers 106
5.3 Interpretation of quantifiers 110 9.1 Preliminary discussion of PM
5.4 Valid predicate formulas 113 204
9.2 System PM
5.5 Proving the invalidity of predicate formulas 116 206
9.3 Development of PM
5.6 Proving the validity of predicate formulas 118 207
9.4 Derived rules
5.7 Polyadic predicates 120 210
9.5 Oonsistency and soundness of PM
5.8 Formation rules for the predicate language 128 214
9.6 Independence of the axioms of PM
217
9.7 Completeness of PM
220
9.8 Completeness of SC
224
Chapter 6: Natural Deduction System PC
Chapter 10: Axioms jor the Predicate Calculus with Identity
6.1 Individual constants and tautologies 134
6.2 Universal elimination and existential introduction 136 10.1 Preliminary discussion of LPC
10.2 227
6.3 Existential elimination and universal introduction 140 System LPO
10.3 230
6.4 Strategies in using quantification rules 147 Development of LPC
10.4 232
6.5 System PO 150 LPC with identity
10.5 236
6.6 Theorems of PC 152 ConSistency of LPC with identity
10.6 238
6.7 Soundness, consistency, and completeness of PC 155 Soundness of LPC with identity
10.7 239
Oompleteness of predicate logic theories
242

Chapter 7: Predicate Logic with Identity Further readings

245

7.1 Identity 160 Index

248
7.2 Some properties of two-place predicates 162
7.3 Predicate logic with identity: System POI 166
7.4 Proofs for arguments 169
7.5 Symbolizing using the identity sign 171
''1'.6 Definite descriptions 175
 I
NATURAL
DEDUCTION
SYSTEMS

,.
 All of us reason. That is, we draw
validity conclusions from what we take to be
true. Sometimes we reason cor-
rectly and sometimes we do not.
That is, even when what we take as
true is true, sometimes our con-
clusions follow and sometimes they
do not. Gradually most of us become
aware that there are certain norms
for reasoning correctly, and if we
follow them and reason from satis-
factory data our conclusions can be
relied on. The study of these norms
or principles of reason is the aim of
this book. And we will use sym-
bolic or mathematical methods to
systematically set out these norms.
It would be futile at this time to
attempt to define 'symbolic logic'.
Rather we will try in this chapter
to give a more or less intuitive
account of the central idea and
purpose of logic.

1.1 Arguments

If we assume that the statements

Sophocles wrote Oedipus Rex.

If Sophocles wrote Oedipus Rex, then he was a Greek.

are true, then our logical intuition tells us that we should be able to
conclude that Sophocles was a Greek. In this type of reasoning we
take the two statements above as assumptions or premises and infer the
conclusion, Sophocles was a Greek. The word argument will be used to
designate a set of premises and an accompanying con/clusion. We
commonly indicate the presence of an argument by using words such as
'therefore' and 'so' between the premises and the conclusion. The

3
4 Validity .1 Arguments

argument corresponding to the above inference may thus be explicitly Furthermore we may have a valid argument made up of one or more f aI se
set down as follows: .
premIses and a true conclusion. Thus in the argument:
Sophocles wrote Oedipus Rex. All U.S. Senators are charismatic.
If Sophocles wrote Oedipus Rex, then he was a Greek. McCarthy is a U.s. Senator.
Therefore Sophocles was a Greek. Therefore McCarthy is charismatic.
we .fin~ a true conclusion that validly follows from premises, the first of
Noone has difficulty seeing that the conclusion of the above argument whICh IS false and the second true.
follows from the premises. In turn, few have difficulty seeing that the What is ruled out if an. argument is valid is the comb'Inat'IOn 0 f true
.
conclusion of the next argument does not follow from the premises. prmll1ses and false conclusIOn. The only combinatl'on th t
. " . a cannot OCCur
if an argument
. . IS valId IS that of true premises and a J.a£' I
se coneI '
USlOn
All Communists are atheists. The validIty of an argument only assures us that I'f th . .
Bertrand Russell is an atheist. e premIses are
true, then the conclusion is also true In other word .
Therefore Bertrand Russell is a Communist. ... . . .. .' s an argument IS
vahd Iff. (If and only If) It IS ImpossIble for its premises to b e t rue an d ItS
.
conclusIOn false.
We may thus have arguments in which the conclusion follows from the
premises and arguments in which the conclusion does not follow. We Exercises
need not be reminded that there are arguments in which it is not so 1. Which of the following are valid arguments?
easy to see whether the conclusion follows. In fact it is such short- (a) Every conservative believes in stability.
comings in our logical intuition that, in part, motivate our interest in Every fascist believes in stability.
logic. Therefore every conservative is a fascist.
(b) For any number x, x + 0 = x.
In logic we are interested in the relation between the conclusion of an
Therefore there is a number x such that 0 + x = 0
argument and its premises. We are concerned with whether a conclusion
(c) With Wilson we had a Democrat and a war' with FD' R d T
does or does not follow from the given premises. If the conclusion does h dD ,an ruman we
a. emocrats and wars. With Kennedy and LBJ we have had th
follow from the premises the argument is said to be a valid or sound thmg. Thus Democratic Presidents cause wars. e same
argument. If the conclusion does not follow from the premises, then the (d) If I ~ad my way, national primaries would take the place of pa t
ventlOns. r y con-
argument is said to be invalid or unsound.
It is well to note at the outset that there can be valid arguments that 2. Give examples of valid arguments in which:
have one or more false premises, and valid arguments with a false con- (a) All the components are false.
clusion, and there can even be valid arguments with both false premises (b) One or m~re of the premises are false and the conclusion is true.
(c) The premIses are true and the conclusion is true.
and a false conclusion. For example each premise and the conclusion of
the next argument are false; nevertheless this argument is valid: 3. Test your logical intuition by seeing if you can tell which of the foll .
statements are correct. Let P Q Rand S st d £ owmg
£'
(a ) If P lollows ' " an or any statement
All creatures that fly have wings. from Q, then not Q follows from not P .
All wallabies fly. (b) If P follows from Q and R follows from P, then R follows from Q
Therefore all wallabies have wings. (c) If Q follows from P and also not Q follows from P, then R follows fr~m P
(d) If Q follows from P and if not P obtains, then not Q. .
(e) If Q follows from P, if R follows from not P and if not R £ II f
In turn, the correct appraisal of an argument as invalid does not not Q, then Q. , 0 ows rom

imply that any of its components are false. An invalid argument may be
made up exclusively of true components, as the next example illustrates:
(f) ifromR £pollows
.
from P, not S from Q, and Q from R, then not S follows

Answers
All birds have wings.
All roadrunners have wings. 1. (a) is invalid, (b) is valid, and (c) is invalid; (d) is not an argument.
Therefore all roadrunners are birds. 3. Only (d) is incorrect.
 6 Validity
1.2 Necessarily True Statements
7
1.2 Necessarily true statements
would be true.
. ;All
. nonnecessary statements may b e sal'd t 0 b e con-
Logic makes precise the conditions for valid argumentation. Before tingent. Contmgent statements are thus those for which there can be
introducing the apparatus needed for this task we need to see the close circumstances in which they could be true or false ..iFor example:
connection between a valid argument and a necessarily true statement.
If children are to obey, force must be used on th
What, first of all, is a statement? Sentences are usually classified as The United States is a republic. em.
declarative, exclamatory, interrogative, or imperative. With respect to
some types of declarative sentences it is meaningful to assign either the are both contingent. The first happens to be false and th dh
truth-value true or the truth-value false. By a statement will be meant a b e secon appens
to be true, . ut the world could be such that their truth-values would
declarativ/e/ sentence that is either true or false. Some examples of have been dIfferent.
~ '-- - -

statements are: To make the connection between validity and necessal'l'1 y true
statements, we must next introduce the conditional t . A
." sen ence. con-
(1) 2 +2= 4. ihhonal sentence IS one that has the form
(2) All triangles are isosceles.
(3) There is life on other planets. If ,then---

(1) is true, (2) is false, and (3) is a conjecture whose truth or falsity is It will be found useful to have a way of talking ab ou t w h a t·IS £ound In.
unknown at this time. In our discussion of logic we shall confine our- the first blank and what is found in the second bl nk f t t
. . -. a 0 a s a ement that
selves to statements. The components of arguments, for example, are has a condItIOnal form. What is found in the fir s t bl ank WI'11 b e called
statements. the antecedent of the conditional while what is fou d' th d
. ' n m e secon blank
We will call truth and falsity truth-values. The truth-value of a wIll be called the consequent of the conditional.
statement is thus its truth or falsity. If a statement is true, its truth-value For any argu men t we ·can construct a corresponding
- condI't'IOnaI
t . . . h
is true and if it is false its truth-value is false. In some cases, as the sen ence b y conJommg t e premises using 'and's a d t t' th
. n rea mg e result
third example above illustrates, we may not know the truth-value of a as the antecedent m the conditional form and b k' h .
. ' y ma mg t e conclusIOn
statement. But if anything is a statement it must be either true or false. the consequent m .the conditional form . To 1'11us t rat e, b eIow are two
r
I
A statement may be either necessarily true, necessarily false, or arguments and then corresponding conditionals:
co-ntingent. Every statement, in fact, is either necessarily true, neces-
ARGUMENT:
sarily false, or contingent. A statement is neceSSMY if no state of affairs
Some men are mortal.
could possibly exist that could change its truth-value. As G. W.
Therefore some mortals are men.
Leibniz, the seventeenth-century German mathematician, logician, and
philosopher, put it, aJlecessary statement is one which is true (or false) CORRESPONDING CONDITIONAL:
, in all possible worlds. IThus, to illustrate,
If some men are mortal, then some mortals are men.

ARGUMENT:
If Frances is wise and witty, then Frances is wise.
If Frances is wise then someone is wise. 6 < 7 (reads: "6 is less than 7").
7 < 10.
Therefore 6 < 10.
are necessarily true, whereas
CORRESPONDING CONDITIONAL:

Some square objects are round. If 6 < 7 and 7 < 10, then 6 < 10.
We
. can now make tl1e connec t'IOn t h at IS
. Important
. for thO d'
is n~gessarily false-we could imagine no set of circumstances in which it CUSSIOn the t' b . IS IS-
, connec IOn etween a valid argument and a necessarily true
8 Validity 1.3 Valid Logical Formulas
9

statement. It is simply this: An argument is valid if and only if its Answers

corresponding conditional is necessarily true. The conditionals corre- 1. Only (b).
sponding to the last two arguments are necessarily true; thus the argu- 4. Only (al and (d).
ments are valid.

Exercises
1. Which of the following are statements~ 1.3 Valid logical formulas
(a) Pass the tanis root.
(b) Reading Kant isn't easy. When a statement is necessarily true, it is usually the case that state-
(c) Why don't you think up a nice sentence~ ments with the same form are likewise necessarily true. But here we
(d) Where is the master key~
must make it clear what we are talking about when we talk about a form
(e) Put the boxes in here.
that a statement has.
(f) Back to stacking books.
To begin to make this clear, consider these two necessarily true
2. Try to classify the following as necessarily true, necessarily false, con- statements:
tingently true, contingently false, or none of these.
(a) Every brother is a sibling. (1) If 6 < 7 and 7 < 10, then 6 < 7.
(b) For every number x and y there is a z such that x + y = z. (2) If the sun is shining and the temperature is lloa, then the sun is
(c) Somebody loves everybody. shining.
(d) Everybody is loved by somebody.
(e) There is an honest and intelligent U.S. politician. Let us try to analyze a common form that these two examples have.
(f) The first man to land on the moon is alive today.
Let us begin by replacing 6 < 7 in (1) with P and 7 < 10 with Q. The
(g) If G. Frege read this book, he would express his disapproval.
(h) If x is next to y, then it is false that y is next to x. result of such a replacement for (1) is
(i) There is intelligent life on a planet other than Earth.
(j) There is no intelligent life on Earth. (3) If P and Q, then P.

3. Write the corresponding conditionals for the following arguments. Next let us regard P and Q as variables for statements. Thus as x stands'
{a) Only idiots would buy calls or puts.
Alfred bought some calls.
for numbers in, say, x +
5 = 10, so P and Q stand for statements. As we
have seen, we obtain (1) from (3) by replacing P with 6 < 7 and Q with
Therefore Alfred is an idiot.
(b) If Alfred sells short, the market goes up. 7 < 10. In turn, we can obtain (2) from (3) by replacing P with 'The sun
If Alfred sells long, the market goes down. is shining' and Q with 'The temperature is lloa,.
Since Alfred is broke, he cannot sell short or long. So either the market We shall call (3) a conditionaljol'mula, since variables are employed.
won't go up or it won't go down. Both (1) and (2) will be spoken of as substitution instances or just
(c) For any number x there is a y such that x is less than y. Therefore there instances of formula (3). A substitution instance of a formula of the sort
is a y such that y is greater than any number. .
(d) Since someone is admired by everyone, it follows that everyone admIres we are considering is the statement that results from the given formula
someone or other. when the variables occurring in the formula are replaced throughout by
statements, it being understood that the replacing is done uniformly.
4. Which of the conditionals corresponding to the above arguments are
necessarily true statements~ Thus if we replace both P and Q with 'This text is dull', we obtain an
instance of (3), namely
5. Why are the following arguments valid, given the account of 'valid arguments'
found in Section 1, and why are their corresponding conditionals necessarily (4) If this text is dull and this text is dull, then this text is dull.
true statements1
(a) Hotel Utah is in Salt Lake City. Here we have a statement that results from replacing each occurrence of
Therefore 2 + 2 = 4. P and Q with 'This text is dull', and this is done uniformly-the replace-
(b) Hotel Utah is in Salt Lake City and Hotel Utah is not in Salt Lake City.
'I,Therefore 2 + 2 = 3. ment we do for a variable in one of its occurrences we do for it in all of its
 Validity Valid Logical Formulas 11
10
(8) is necessarily true. However if we seek the sentential formula of
occurrences. However tIle .collowing
.L'
statement is not an instance of (3):
which (8) is an instance we obtain, for example,
If this text is dull and the sun is shining, then the sun is shining.
If P, then Q. \~
since the substitution is not done uniformly-the same statement is not If P and Q, then R.
put in for P in each of its occurrences. P.
With (3) we have a common form of (1), (2), and (4). Other common Q.
forms that (1), (2), and (4) have are
none of which are valid.
(5) If P, then Q. To describe the logical form of (8) in virtue of which it is a neces-
P. sarily true statement we need to introduce other kinds of variables. Let
Q. us use the letter x as a variable for names of individuals such as 'Phil'.
An important feature of formula (3), which the other t.hree formulas And let us use F and G as variables for expressions such as 'men',
I k is that every instance of formula (3) is a necessanly true state- 'mortal', 'bachelor', and 'lonely being'. With this new notation we may
~:n~. In other words, any statement of the form 'If P and Q, then P' is state the pattern of (8) relevant to its truth, as follows:
necessarily true. Formula (3) is not the only formula with this feature.
(9) If every F is G and x is F, then x is G.
Other examples of formulas that only have necessarily true statement
instances would be (9) is a valid formula, for whatever individual is picked out for x, and no
(6) If P or Q and not P, then Q. matter what property is chosen for F and G, the result will be a neces-
P or not P. sarily true statement. For example if we allow x to be 6 and F and G to
If P and if P then Q, then Q. be the properties respectively of being an even number and being
divisible by 2, the instance of (9) below is obtained:
In contrast, the next formulas are such that not all instances are
necessarily true: (10) If every even number is divisible by 2 and 6 is an even number,
then 6 is divisible by 2.
(7) If P or Q, then P.
P orQ.
If Q and if P then Q, then P. (10), like (8), is a necessarily true statement. (8) and (10) are thus
necessarily true because they are instances of the same valid logical
Let us now introduce a second use for the term 'valid'. Earlier we formula: (9). Note that (9) is not a sentential formula since sentential
introduced 'valid' as an appraisal term for arguments. We. will ~0;V formulas are formulas made up only of variables such as P and Q and
speak of some formulas in logic as valid. To say a formula III .10gIC IS expressions such as 'or', 'if, then', 'not', and 'and'.
l'd will mean that all instances of the formula are necessanly true We may now bring together some of the elements of this discussion.
~:a:ements. The formulas in groups (6) and (3) are all examples of valid Earlier it was said that an argument is valid iff its corresponding con-
formulas, whereas those in groups (5) and (7) are all invalid. ditional is necessarily true. In this section the point is made that if a
The formulas above are made up of variables such as P and Q and conditional is an instance of a valid logical formula, then it is necessarily
.
of expreSSIOns suc h as 'or' " 'l'f then' , 'not' , and 'and'. Let us call these true. Thus to show an argument to be valid it suffices to show that its
sentential formulas. Statements (1) and (2) may. now be .said to be corresponding conditional is an instance of a valid logical formula. If an
necessarily true in virtue of being instances of a valId sententIal formula. argument corresponds in this way to a valid logical formula, we will say
It is easy to think of statements that are necessarily true but that are that the premises of the argument logically imply the conclusion. The
not instances of valid sentential formulas. Consider, for example, immediate task before us is twofold. We wish to set down the structure
of arguments relevant to their validity, thus obtaining our logical
(8) If all bachelors are lonely and Phil is a bachelor, then Phil is
formulas. And we wish to set down exact techniques by which we can
lonely.
 Validity 1.4 Formal Language 13
12
determine whether a formula is valid or not. (We sh~uld not~ that it is Answers
not until we reach Section 7.2 that definitions can be gIVen for Important 1. Only (d) and (f) are not sentential formulas. Only (a), (b) and (e) are not
notions such as 'formula in logic' and 'logically imply'.) valid.
4. Only (a) and (f) are invalid.
Exercises
1. Which of the following are sent en t'IaI f ormuI a s1. Which formulas are valid1 5. If there are rules, then there must be human agreement.
(a) P and not P.
(b) If POI' Q, then P.
(c) If if P then Q, then if not Q, .then not P.
(d) If every x is F, then some x IS F. 1.4 Formallanguage
(e) If P or not P, then P.
(f) If x is F, then something is F. In order to study the exact conditions under which a logical formula is
(g) If P and Q, then Q and P. valid, and in order to display the logical form of arguments, a formal or
2. Give instances of (c) and (d) that express necessarily true statements. Why artificial language is necessary. This means, in part, that we need to set
can you not do this for (a)1 down the expressions to be used in the language, give an exact inter-
3. Provide an instance of each of the argument patterns below. The premises pretation of the expressions, and set down rules for when the expressions
appear above the bar and the conclusion is below the bar. are correctly combined together. The formal language used to express
(a) If P then Q sentential formulas-the sentential language-is a comparatively simple
Q language, and it is introduced in the next chapter. The expressions used
P in this language are variables for statements P, Q, R, ... and are ex-
(b) POI' Q pressions corresponding to devices in the natural language we use to
Q or P connect together statements, for example, 'or' and 'and'. However, exact
(c) If P, then Q interpretations are given for these latter expressions so that no am-
If Q, then R biguity or vagueness can result in their use. Later additions will be
If P, then R made to this language. For example we will introduce variables for
(d) P and not P individuals x, y, z, ... and variables for properties F, G, H, .... The
Q resulting language is called the predicate language and is taken up in
(e) Every x is F Chapter 5.
x is F Having introduced this formula language, we can introduce tech-
(f) Some things are F niques by which we can determine whether a formula expressed in this
Some things are G language is valid or not. We will introduce validation techniques first
Some things are F and G for sentential formulas and later for predicate formulas.
(g) If P, then Q and not Q If an argument is expressed using the formal languages, then we can
Not P say without hesitation that is is valid if its corresponding conditional
(h) For any x if x is F then x is G exemplifies a valid logical formula. If, on the other hand, an argument is
If every x is F, then every x is G not expressed using the formal language, then before using the tech-
4. An argument pattern is valid iff its corresponding conditional is valid. Which niques oflogic one must first paraphrase the argument so that it employs
of the above arguments are valid 1 the expressions of the formal language. This means that we must
If a statement is necessary in virtue of being an instance ?f a valid logical engage in three separate steps. First, layout the formal language.
5. formula, then it is now generally referred to as an analyt~c statem~nt. On Second, develop the systematic methods to determine the validity of a
the basis of the results of 3 and 4, provide some ex~mples of ~nalytlC state- formula expressed in the language. Third, develop the art of proper
ments. Can you give some examples of nonanalytlC necessarIly true state- paraphrase of arguments into the language if they are not already
ments1
--,~ .,
14 Validity
The Use/Mention Distinction
15
expressed using the language. It should be stressed at this point that
bool{s,
. we will ignore . the. use/mention
. distinction unl ess confus "
IOn IS
today the preferred way to express statements in mathematics and the
hkely to result from Ignormg It. We have up to this time b . £
other sciences is to use the formal language presented in this book. . thO t· een, In act
fol I owmg IS prac ICe. The reason for this is to av 01'd th e unSIghtly
. '
In the next chapter the formal language needed for sentential logic . . .
prolIferatIOn
. of smgle quotation marks . It is hoped th a t th b
e a sence of
is introduced. But before turning to it, one more matter needs to at
quotatIOn marks when terms are mentioned has not and will not undul
least be touched on in this introductory chapter. upset the reader. y

1.5 The use/mention distinction

It is a truism to say that an assertion about a thing contains the name of

the thing and not the thing itself. When we speak about Lompoc we use
the name 'Lompoc'; the Oalifornia city does not appear in our assertion.
In turn we may have occasion to speak not of the city W. O. Fields made
famous, but of the name of the city. For example we might wish to say
that the name of the city is made up of six letters, or that it is disyllabic.
The customary way of doing this is to use single quotation marks. Thus
to say the name of the city has six letters, following this convention, we
say
(1) 'Lompoc' has six letters.

In turn to say something about the city rather than the name, we do not
use any quotation marks at all, as the next sentence illustrates.

(2) Lompoc is near the coast.

A kind of nonsense is obtained if we replace thing with name of thing and

vice versa, as the next two sentences illustrate.

(3) Lompoc has six letters.

(4) 'Lompoc' is near the coast.

(1) is about a word; (2) is about a city; (3) is about a city but makes the
queer statement that it has six letters, whatever that could mean; and
(4) is about a word that is near the coast-imagine millions of pieces of
paper with 'Lompoc' on them. In (1) the name is commonly said to be
mentioned, while in (2) it is said to be used.
Oonfusions can result from carelessness with the use/mention dis-
tinction. This is especially likely when we operate on a double single
quqte level. But following the widely practiced custom in introductory
 2.1 Conjunction and Denial
17
The first two columns list all possible assignments of truth-value to the
pair P and Q, and the last column gives the corresponding truth-values
for P A Q. This table indicates that when P is true (abbreviated T) and
Q is T-first line-then P A Q is T; and when P is T and Q is false
(abbreviated F)-second line-then P A Q is F; and so on. The table
The main interest in this chapter is above is an example of a truth-table. It will be spoken of as a basic
sentential with part of the formal language truth-table since it provides the interpretation for A. An example of a
needed to display the structure of conjunction is
language arguments relevant to questioning
whether the premises logically imply 2 +3= 5 A 2 +3= 4.
the conclusion. The part of this
formal language to be considered which is false since, even though the first conjunct is true, the second
in this chapter is the part that conjunct is false.
relates whole statements. This lan- The most familiar way to deny a statement is to prefix the words
guage will be made up of two kinds 'it is not the case that' to it. For example, we may deny 'Richard is a pale
of ingredients: variables for state- person' in this way:
ments and symbols to relate state-
ments. We will call this fragment of It is not the case that Richard is a pale person.
the formal language the sentential
language. Other ways to deny this statement would be

Richard is not a pale person.

2.1 Conjunction and denial It is not true that Richard is a pale person.
It is false that Richard is a pale person.
We are all familiar with the process of joining two statements together
by using 'and'. For example 'Richard is a pale person' and 'Richard is a The next symbol, ""', will be understood to correspond to the words
poet' may be so combined as follows: 'it is not the case that' in the natural language. We thus write the
Richard is a pale person and Richard is a poet. negation of P as ""P. "",P is called the negation of P. In the natural
language the negation of P is true when P is false and is false when P is
or, simply,
true. The same will be true with "",P, as the following basic truth-table
Richard is a pale person and a poet. indicates:
According to the usual meaning of 'and', P and Q is true iff P along with P "",P
Q is true. We will now introduce the symbol A and so understand it
T F
that it will correspond to this use of 'and'. P A Q is called a conjunction, F T
and the components P and Q are called conjuncts.
As indicated above, P A Q will be so understood that it is true iff P If Pis T, then ,......,p is F, and if P is F, then "",P is T. Thus
along with Q is true. This interpretation for the symbol A can be given
using the table below: "",Mao is a poet.

P Q PAQ is false, whereas

T T T
T F F "'" ""Mao is a poet.
F T F
F F F is true since Mao is a poet.
--,~ .,

16
18 Sentential Language
19
It is not difficult to build compound statements using A and '--'. Let Exercises
us do this. Suppose
1. F~rmulate the statements corresponding to each of the form
mme the truth-value of the statement. ulas and deter-
R: Richard is a pale person. P: 5 + 2 = 7.
P: Richard is a poet. Q:2+4=7.
then (a) P A Q
(1) R A ,-..,p (b) ,--,(P A Q)
(c) ,-...;P A ,......,Q
is the compound statement: Both Richard is a pale person and it is not (d) ~'"'-'P
(e) ,-......, ,--,(,-...,p A ,-....,Q)
the case that Richard is a poet. Or by using parentheses as we use (f) P A Q A P A Q
punctuation in our natural language, we can obtain
Answers
(2) ,,-,(R A P)
1. Only (b) and (d) are T.

"it is not the case that both Rand P". Other examples of compound
statements would be:
2.2 Other statement connectives
(3) r-..>R A ,,-,P and (4) r-v(,--,R A ,....,P)
It will prove useful to introduce more stat
. emen t connectIves
. namel
connectIves corresponding to 'or' 'if then' d "f d I" Y
"it is not the case that R and it is not the case that P" and "it is not the . . '" an 1 an on y If'. One wa
case that both it is not the case that R and it is not the case that P." to ~~ thIS IS by ~efining our new connectives using A and "-' Th~
addItIOnal connectIves we will use can be . t, d d' . .
We may also inquire whether such compound statements are true follows: m 10 uce m thIS way as
or false. Given, say, that R is T and Pis F, is (1) T or F~ It is T. To
figure this out, we first replace the capital letters with the indication of (1) P V Q =dr r_+",p A "-'Q).
the truth-value of the statements they express. Thus with (1) we (2) P-+Q=dr"'-'PVQ.
(3) P<ri- Q =df (P -+ Q) A (Q -+ P).
obtain
T A,.....,F
In (1) V corresponds to 'or', in (2) -+ corresponds to 'if then' and' (3)
<~> corresponds to 'if and only if'. ' , m
Each of the definitions indicates that h t . .
since ,,-,F is T (see the table for ,-..,) and since TAT is T (see the table ,_ , . " w a IS on one SIde of the
-df SIgn IS Just another way of writing what·s th tl .
for A), (1) is T. We may work out the truth-value of (2), (3), and (4) as 'I'h P Q' . 1 on e 0 leI' SIde
us V IS Just a shorter way of writing r-..;(,-.....,P A ""Q) d'
follows, given that R is T and P is F. r-..J(,......,p A r-..-Q) h tl h ' an
. as exac y t e same meaning as P V Q Wh t ' th
mcamng of "-,(",,P A ,-.....,QP ( . . a IS e
(2) ,-,.,(T A F) (3) ,.....,T A ,-,.,F (4) ",-,(,-,.,T A ,--,F) is T and' F I h .,....., ",-,p A ,.....,Q) IS T whenever P is T or Q
,.....,( F ) FAT ,-,.,(F AT) ]S on y w en both P is F and Q' F Tl' .
the truth-table for V is IS . lIS bemg the case,
T F ,-.v ( F )
T P Q PVQ
T T T
The symbols ,....., and A are both called stcttement connectives, even T F T
though '--' is a modifier. We might well think of statement connectives F T T
as operating on statements. ,-.v can be looked at as a one-place con- F F F
nective because it operates on one statement, and A is a two-place PVQ'
. IS called a disjtlnction. The comlJonents of P V Q,
cOl!,~ective because it operates on two statements. called disjuncts. P and Q, are
20 Sentential Language
2.3 Determining the Truth-Value of Compound Statements
21
Proceeding along the same lines, the truth-table below gives the 2. Formulate the statement corresponding to each of the formulas and determ'
intended interpretation of ---->- and +->. the truth-value. me
P: A triangle has three sides.
P Q P---->-Q
Q: A triangle has four sides.
T. T. T. T.. (a) P ->-Q
T. F F· F (b) Q ---->- P
F T. T. F (c),--,P---->-Q
F F T. T. (d) ,--,Q --+ P
(e) ,--,P V,-...;Q
This table shows that P ---->- Q is false only when P is T. and Q is F, other-
(f) ,--,(P V Q)
wise it is T; and P +-> Q is T iff P and Q have matching truth-values.
(g) P+-> Q
P --+ Q is called a conditionctl, with P the antecedent and Q the conse- (h) ,--,P+-> Q
quent. P +-> Q is called a biconditional.
As indicated above, ---->- corresponds to 'if, then'. The truth-table 3. Which. of the
. following statements are simple and whl'ch are compoun1
d
interpretation of --+ fails to capture the full meaning of 'if, then' as it is (a) WIlson IS not on the job. .
often used in the natural language. If this is not obvious, then the dis- (b) Wilson is on the job.
(c) If Wilson is on the job, then Bob is on the job.
cussion in Section 2.4 will make this clear. ,Vhat is well to emphasize at
this point is that, first, P ---->- Q is equivalent to ,-....;p V Q, which is
4. The statement connectives introduced in this chapter are tmth-f1tnctional
equivalent to r--.;(P /\ ,-...;Q). This means that to assert P ---->- Q is to assert because· the compound statements formed by using tllem are t rue or £aIse
that it is not the case that P.is true and Q is false. Whenever we assert depend Illg on the truth-value of the simple component t t t N
It K 'k ( ) tl ' T ' s a emen s. ow
'if P, then Q', we mean at least that it is not the case that the antecedent e mean now slat. 0 express that 'a knows that', we write Ka. A
is true and the consequent false. Second, it is this feature of the use of compound sentence can now be formed using this s b I F
Itt' P b 'M t' t I h ym o. 'or example
'if, then' that is sufficient for mathematical and scientific use of 'if, then'. c Illg e ar III s 0 e t e cat', we can write 'Dick knows th t M ~t' '
stole the cat' as a al III
All that we need to mean by 'if P, then Q' in standard mathematical (1) KaP.
usage is ,-....;p V Q or ,-...;(P /\ ,-...;Q). The above truth-table interpretation If we allow => to read ' - - - is less probable than ____
compound sentences such as " we can form
for these statement connectives also corresponds to the standard uses in
(2) P=> Q.
the sciences.
Show that neither K nor => is truth-functional. Show this also for 'because'.
We 'will call a statement that uses no statement connectives a
simple or (ttomic statement, and a statement formed by using connectives,
a compound or molecular statement. Thus a more accurate name for the Answers
five symbols just introduced would be compound statement forme1's rather
I. (a) F, (b) T, (c) T, (d) F, (e) T, (f) F, (g) T, (h) F, (i) T, and (j) T.
than statement connectives. Nevertheless we will continue to speak of
these five symbols as statement connectives. 2. (a) F, (b) T, (c) T, (d) T, (e) T, (f) F, (g) F, and (h) T.
Exercises 3. Only (b) is simple.
1. Give the truth-value of each of the following statements.
(a) 2 > 1 /\ 1 > 2.
(b) ,,-,(2) 1 /\ 1 > 2).
(c) 2 > 1 V 1 > 2.
(d) ,-...;(2 > 1 V 1 > 2). i 2.3 Determ"mmg th e truth-value of compound statements
(e) '--'(2) 1 ---->- 1 > 2).
(f) 2 > 1 --+ 1 > 2.
(g) 1 > 2 ---->- 2 > 1.
!:1 f~;'ming compound formulas or statements we can encounter am-
(h) 1 > 2 +-> 2 > 1. 19l1l y. For example how are we to understand this formula:
(i) r-v(1 > 2 +-> 2> 1).
m (1 > 2 /\ 1 > 2) --+ 1 > 2. (1) P /\ Q V R
 Sentential Language 2.3 Determining the Truth-Value of Compound Statements 23
22
Capital letters are here used as sentential constants; they are abbrevi-
Is it a conjunction with one conjunct a disjunction, or is it a disjunction
ations, stand-ins, for actual statements. A little knowledge of U.S.
with one disjunct a conjunction? In other words, is (1) to be read
political history at the beginning of the twentieth century will reveal that
(2) Both P and Q or R. P, and Q or R. Rand B are true, whereas Wand L are false. The question can be
raised, is the following compound statement true or false?
or
(3) Either P and Q or R. P and Q, or R.
R A (W V r->L)
To remove this ambiguity we continue to use parentheses. To indicate
(2) we will write: To solve this, we first indicate the truth-values of the simple statements
as follows:
P A (Q V R)
T A (F V,....",F)
and to indicate (3):
Next we consider the connective with the least scope. Each occurrence
(P A Q) V R
of a connective connects or operates on one or two parts of a statement,
Intuitively called the scope of that occurrence of the connective. Parentheses
clearly mark out the scope of connectives when more than one con-
""PVQ nective is used. In the above case:
is understood as not P or Q. But if we wish to negate P V Q we will write:
r->connects to L.
V connects Wand ,...., L.
,,-,(P V Q) A connects Rand (W V ""L).

(Note the difference between r-..;(P V Q) and ,...."P V ,-....,Q and ,.....,(P A Q)
Thus, in the above statement ,-..." has the least scope and A has the greatest
and ,-....JP A ,......,Q.) ·When we have strings of conjunctions
scope since,......" connects the least number of parts, and A connects the
greatest number. Starting with the operator with the least scope, we ask
if the compound it forms is T or F. ,-....,F is T, so we have
or disjunctions
T A (F V T)
PvQVR
The V connective is now the operator with the least scope, and since
we need not use parentheses (though we can if we like) since (P A Q) A R F V Tis T, we obtain
is equivalent to P A (Q A R) and (P V Q) V R is equivalent to
P V (Q V R). TAT
For the purpose of becoming acquainted with using the five con-
nectives, let us use the following capital letters for the indicated and since TAT is T, the answer to whether the statement we started
statements. with is true or false is that it is true. This problem could be solved as
follows:
R = Theodore Roosevelt was the only presidential nominee of the
Bull Moose Party. T A (F V ,,-,F)
BWilliam Jennings Bryan at one time lent his name to real
=
T A (F V T )
estate selling in Florida.
T A (T )
W = ·Woodrow ·Wilson had a lovable personality.
l'
L = Senator Robert La Follette was a gracious loser.
 2.4 Statement Connectives and the Natural Language 25
Sentential Language
24
it follows that the next statement is false:
To consider a second example, what is the truth-value of
(2) John got well and had an operation.
r--'(E -4- ((r--'W V B)H r--'E))
Part of what is implied, normally, by (1) is that John got well after he
.
Working this in succeSSIVe st eps, we find, as follows, that the answer had an operation, which, if true, makes (2) false if we understand 'and'
in (2) to be used like 'and' in (1) to express a temporal sequence. Now
is T:
if 'and' as used in (1) and (2) were synonymous with /\, then (1) would be
r--'(T -4- ((r--'F V T)H r--'T)) true iff (2) were true. This thus shows one use of English 'and' that
,-.J(T -4- ((T V T)H F )) differs from /\. To take another familiar example, from the truth ofthese
r--'(T -4- ( T H F ))
statements:
c-->(T -4- ( F))
r--'( F ) Tom came to the party.
T Jane came to the party.

Exercises it does not follow that

1. Determine the truth-value 0 f each 0 f tl1e f 0 11ow ing formulas, given the (3) Tom and Jane came to the party.
indicated truth-values of its components.
P:T if (3) is understood in one of its most popular senses-to express that
Q:F Tom and Jane came to the party together. It is easy to see that (3) would
E:T follow if 'and' in (3) were replaced by /\. It may be that as more and
(a) ((P -4- Q) /\ Q) -4- P
more people take logic and mathematics courses, the use of 'and' and the
(b) r--'(PH (Q V E))
(c) ((P-4-Q) /\ (Q-4-E))--+ (P-4-E) other connectives in the natural language not in accordance with truth-
(d) (P --+ Q)H (r--'Q -4- r--'P) tables might someday come to be regarded as deviant or incorrect, but
(e) ((P -4- Q) H (r--'Q -4- r--'P)) /\ E this is certainly not true today.
(f) ((P /\ Q) -4- E) -4- (P -4- (Q -4- E)) 'Or' in the natural language can have what has become known as an
(g) r--'(r--'(Q --+ r--'P) --+ (P -4- E)) --+ r--'(P /\ Q)
exclusive use. Often we so use 'or' that the compound statement it forms
is true iff one or the other disjunct is true, but not true when both are
Answers true. For example, if one said
1. Only (b) is F.
Hale Boggs is either a Senator or a Congressman.

one would be saying that Hale Boggs is either a Senator or a Congressman

2.4 Statement connectives and the natura/language but not both. This use of 'or' is commonly called the exclusive use. The
use of 'or' which would correspond to V is thus called the nonexclusive or
That the connectives of the natural language- ,not,"or, " an d' ,1,.
'·f then' '
inclusive use. A capital v, V, could be introduced and interpreted so as
and 'if and only if' -are used in a variety of ways and often are not ll1ter- to correspond to the exclusive use of 'or'. Accordingly we would supply
changeable with r--', V, /\, --+, and H , respectiv~ly, is dis~uted b.y v.~ry the truth-table below for V:
£ For our purposes it counts for little to reVIew the wIdely dIffenng
ues:~ of English connectives. However, it will help to il:umin~t~ the P Q PVQ
nature and importance of the formal language for sententIalloglC If we T T F
take notice of some of these uses. T F T
As 'and' is sometimes used, from the truth of this statement: F T T
F F F
(1) John had an operation and got well.
 26 Sentential Language
2.4 Statement Connectives and the Natural Language 27
Or we could introduce V by this definition: antecedent is true and the consequent false. The symbol for the con-
ditional, --+, thus captures a common feature of all uses of 'if, then' ,
P V Q =df (P V Q) A ",-,(P A Q) since P --+ Q is equivalent to ,-....,(P A r-->Q). And it is the maintaining
of this feature, as indicated earlier, that is sufficient for the use of 'if,
It is interesting to note also that 'or' in the natural language is often then' in mathematics and scientific contexts.
used to express an exhaustion of alternatives. This is best explained by The obvious gap that exists between the statement connectives and
an example. In one familiar context in which one would say that Rio their counterparts in the natural language need be no cause for alarm',
Mundi is either a socialist country or a capitalist country, the statement rather it makes the activity of this chapter all the more necessary.
would imply that these are the only alternatives. These two alterna- First, it is our desire or goal to form systems that will generate all the
tives-being socialistic or capitalistic-exhaust the possibilities. When valid and only the valid sentential formulas. If the class of formulas were
such exhaustion is implied, 'or' is not used truth-functionally since the in part formed by connectives in the natural language, their multiple and
truth-value of the compound statement is not an exclusive function of changing uses would make such systems impossible. Second, the tech-
the truth-value of its components. niques that we will begin to develop in the next chapter for determining
Turning to 'if, then', it is now commonplace to say that there are the validity of arguments are to be applied to arguments employing the
many uses for this important connective in the natural langugage. statement connectives. The truth-table interpretation for the con-
Consider, for example, this list: nectives conforms to standard mathematical and scientific usage. When,
however, we come upon an argument in a nonscientific context, methods
(4) If the sun shines, then I will swim. of paraphrase will be at hand and standards for correct paraphrasing will
(5) If it snows, Saul won't show up. be given so that arguments in ordinary contexts can often be appraised
(6) If iron is heated sufficiently, it will glow red.
by the forthcoming standard logical techniques. More will be said about
(7) If x is a body, then x does not travel faster than light.
(8) If x is red, then x is not blue. this last point after the reader becomes acquainted to some extent with
(9) If there are rules, then there is agreement in a society. the terms of the formal language and with the structures of statements.
(10) If Stevens is a doctor and all doctors are well paid, then
Stevens is well paid. Exercises
(11) If 8> 7 and 7 > 6, then 8> 6.
(12) If McCartliy had run, Nixon would have lost. 1. Show that each of the italicized connectives in the following statements, if
replaced by the indicatedsta tement connective, may not preserve truth-value.
(a) Ruth got married and had twins. (A)
These are not statements using --+, for if they were, then for each one,
(b) I won't go unless she apologizes. (V)
the fact that the antecedent is false or the consequent is true would (c) I love you but I love your sister as well. (A)
alone be sufficient to establish the truth of each of the conditionals. In (d) LBJ would have won in '68 if he had run. (--+)
fact, no combination of true/false antecedents and consequents by itself (e) Either you drop that stick or I'll tell Mother. (V)
would be sufficient to declare any of the above true. We must dis- (f) He read the logic manuscript and got apoplexy. (A)
tinguish, then, the uses of 'if, then' in the above from --+, remembering (g) There is coffee on the stove if you want some. (--+)
that P --+ Q is equivalent to ,-....,(P A "'-'Q). In (10) and (11), at least, we 2. Provide a reading for P V Q.
can say that the antecedent entails or implies the consequent, these being
analytic statements. With (9) we have a conceptual connection, and 3. Given P: ']' and Q: F, determine the truth-value of each of the following
formulas:
with (6) a casual one. Statement (12) provides us with an example of a (a) ,-...,p V Q
counterfactual conditional, a statement that may be true or false, even (b) P +--+ (P V P)
though both the antecedent and the consequent are false. Statement (4) (c) (P V Q) V (P V Q)
is used to inform someone of the speaker's intentions and (5) is used to
predict the beh~vior of another person. However, it should not be 4. I~valuate the following claim: The statement 'If P then Q' asserts no more
overlooked that whenever we use, 'if, then', what we say is false if the than r--..;(P A ,-....,Q), though one may say it because one knows or believes that
-,- some non-truth-functional connection exists between P and Q.
28 Sentential Language 2.5 Putting the Natural Language into Symbols
29
Answers P if and only if Q
P if Q, and Q if P
1. (a) 'Ruth got married A Ruth had twins' is equivalent to 'Ruth had twins A
P exactly if Q
Ruth got married'.
P is (a) necessary and sufficient (condition) for Q
(g) If there is no coffee on the stove, then (g) is false.
PiffQ
3. (a) F, (b) F, and (c) T.
If P then Q, and conversely
If P, then Q
2.5 Putting the natural language into symbols When P, then Q
In case P, Q
In translating the natural language into the sentential language some Q provided that P
confusion can arise because of the various ways in which we express P is (a) sufficient (condition) for Q
biconditionals, conditionals, conjunctions, and disjunctions, and the
Q is (a) necessary (condition) for P
P implies Q
various ways we have for denying statements. Q if P
In paraphrasing the natural language into our notation it will prove Q when P
useful to have a list that covers the most frequently encountered idioms Q in case P
that can be expressed using statement connectives. (In the chart P only if Q
below we assume that the use of the natural language conforms to P only when Q
P only in case Q
standard mathematical and scientific usage.) Below is such a list. 1 In
each section each statement on the right can be expressed using the PAQ PandQ
statement on the left. For example, in the fourth section we find that PbutQ
P unless Q can be expressed as P V Q (which is equivalent to r-vP -+ Q). P although Q
Both P and Q
In the second section we see that P only if Q is equivalent to P -+ Q,
Not only P but Q
whereas P if Q comes to Q -+ P. P despite Q
Sometimes 'or' is used in the inclusive sense and sometimes in the PyetQ
exclusive sense. When 'or' is used in the first way we have P V Q, P while Q
and when it is used the other way we have (P V Q) A r-v(P t\ Q). These PvQ P or Q or both
two ways are indicated in the chart by placing '[sometimes]' after P or Q. P or Q [s]
Since the use of 'unless' and 'except when' can also express the exclusive P unless Q [s]
sense of 'or', [s]'s are placed after them in each section. P and/or Q [in legal documents]
The words 'necessary' and 'sufficient' are frequently used in various Either P or Q [s]
P except when Q [s]
contexts. For example we may say that having a right angle is necessary
for a figure to be a square, and we may say that a figure's being a square (P V Q) A ",-,(P A Q) P or Q but not both
is sufficient for a figure to have a right angle. Vlfe can express these P or Q [s]
relations using -+ as follows: Either P or Q [s]
P unless Q [s]
P except when Q [s]
x is a square -+ x has a right angle.
P or else Q [s]

P Q, in other words, can be read "Q is a necessary condition for P" or

-+
",(P V Q) Neither P nor Q
"P is a sufficient condition of Q." P ~ Q, in turn, can be read 'P is a
Not P [or the result of transforming P by putting 'not' just after
the verb or an auxiliary verb]
1 This list is a modification of a list in Stephen Cole Kleene, ~Mathematical Logj:c
P does not hold
(New York: John ~Wiley & Sons, Inc., 1967), pp. 63-64. Adapted by permission of It is not the case that P
the Pl1,blisher. P isn't so
so Sentential Language 2.6 Formation Rules for the Sentential Language 31

necessary and sufficient condition for Q'. (These relationships are all set Answers
down in the chart above.) To say P only if Q is to say Q is necessary for
1. (a) P -+ E, (c) A -+ R, and (e) (E A D A I) -+ P.
P, thus P -+ Q, whereas to say P if Q is to say Q is sufficient for P. To
say P provided that Q is to say Q is a sufficient condition for P. 2. (a) P-+E (e) (P-+Q)-+R (f) M-+P
~E (R AS) -+ (,""T V ,,-,U) P -+ (L V H)
~P S H~A
Exercises Q-+~(P V U) "-' L V "-' P
,....,P-+ (S A A)
1. Paraphrase the following statements into the sentential language using the
M-+H
suggested letters for the simple statements.
(a) Only those who do exercises will pass logic. (E, P) 3. Only (a), (c), and (f) have analytic conditionals. The argument is valid,
(b) This litmus paper turns red if it is placed in acid. (R, A) the premises logically imply the conclusion.
(c) This litmus paper has been placed in acid only if it turns red. (A, R)
(d) You won't pass the course unless you do the exercises. (P, E)
(e) If you do the exercises you will pass the course provided that you are
diligent and intelligent. (E, P, D, I) 2.6 Formation rules for the sentential language

2. Paraphrase the following arguments into the sentential language. Allow It is time that we made quite clear exactly what is being called the
letters to stand only for simple statements.
sentential language. Up to this time the reader has noticed that the
(a) A student passes logic only if he does the exercises. But students never
do exercises. Therefore a student never passes logic. sententiallallguage uses symbols for statement connectives and P, Q, ...
(b) He will not come unless it rains. It is raining, so he will come after all. as variables for statements. Parentheses are also employed to remove any
(c) If we do not draw and do not lose, then we win. Therefore if we do not ambiguity. The reader has also, more or less, been able to see when these
draw then we win, or if we do not lose then we win. elements are properly put together. But it is now time to make all this
(d) P provided that neither Q nor R. oxaot.
T, if Rand S. What is being called the sentential language essentially consists of
If U, then S but not T.
three elements: First, a collection of symbols; these are:
P and Q being true is sufficient for U being false.
(e) If P implies Q, then R. (a) Statement variables: P, Q, R, ...
Either not T or not U is necessary for both Rand S. (b) Statement connectives: r-v, A, V, -+, and~
S (c) Parentheses: (, )
If Q, then neither P nor U.
(f) P, if M; but if P, then either L or H. Now H if and only if A. On the (The three dots, ' ... ' indicate that we suppose we have an infinite
other hand, not L unless not P. Not P only if S and A. So a sufficient number of statement variables available. When we reach the end of the
condition for H is M. alphabet we can continue the series by using subsoripts PI' Qv ... and
then P 2, Q2' ... , and so on.) Second, a truth-table interpretation for the
3. Try to determine whether the conditionals corresponding to each argument statement connectives. Third, rules for when we have a proper string
are analytic. If the conditionals are analytic, what does this imply about
of symbols.
the argument?
The reader has been able to rely on his intuition to determine when
4. For each pair below state all necessary and sufficient relationships that exist. one has a proper string of symbols. For example, we intuitively reject
(a) x is a normal man; x is a biped. the following:
(b) x is a nickel; x is worth five cents.
PQ vP Pv (Q-;.-+Q) ,...,yp ,,-,(PlY Q)
(c) x is divisible by 2; x is an even number.
(d) x is greater than 6; x is greater than 34 as proper strings of symbols. And on reflection we see that the criterion
(e) x is a flaming match; x is struck against a match book.
that controls what we regard as a proper string is this: Proper strings of
(f) x is a philosophy teacher; x is a radical.
(g) x's action is freely done; x's action has a cause. symbols can have statements as instances, but improper strings in
(h) x thinks; x is a per~on. general do not.
'"
 32 Sentential Language
Formation Rules for the Sentential Language
33
In this section we wish to set down rules by which we can determine
To see how these rules provide us with a procedure to determine
in all cases whether or not a string of symbols, a formula, is a proper whether any formula is a wff we show that
string, whether it is what is called a well-formed formula. We wish Our
notion of a well-formed formula to be an effective notion. We have an (( (P -->- Q) A C"-'P) -->- Q)
effective notion of a well-formed formula when we can settle in a routine
fashion whether any possible formula is well-formed or not. To be more is a wff in the following way:
specific, the notion of a well-formed formula is effective when a digital
P is a wff by rule 1.
computer can be programmed to check whether any formula is well- Q is a wff by rule 1.
formed or not. If no intelligence beyond the ability to follow clerical (P -->- Q) is a wff by rule 5.
instructions is necessary to check whether a formula is well-formed r-;p is a wff by rule 2.
or not, we have an effective notion of a well-formed formula. The rules ((P -->- Q) A r-;P) is a wff by rule 4.
we set down will insure that the notion of a well-formed formula is ( (( P -->- Q) A r - ; P) -->- Q) is a wff by rule 5.
effective.
The rules that provide the procedure for determining whether a On the other hand, since we cannot show, for example, that
formula is well-formed are called the formation Tules. The formation (Y P -->- Q)
rules for the sentential language may be given as follows (abbreviating
'well-formed formula' as 'wff'): is a wff by using rules 1-6, it is not a wff.
It is important to note that by these rules
1. Any statement variable is a wff.
2. Any wff preceded by r - ; is a wff.
3. Any wff followed by Y followed by a wff, the whole enclosed in
parentheses, is a wff.
4. (3), replacing Y with A. for example, is not a wff. The reason for this is that given rule 5, A -->- B
5. (3), replacing Y with -->-. must be surrounded by parentheses if it is to be a wff. Thus (1) must be
6. (3), replacing Y with~. surrounded by parentheses to be a wff. This is necessary to remove the
possible ambiguity that was discussed in Section 2.3. We do not wish
To complete the definition of a wff we need add that a formula is wf iff to admit, for example,
its being so follows from the formation rules. (This condition will hence-
forth be supposed in formation rules for wffs.) (2) P A Q -->- P
A useful device in stating these formation rules is to use a term that
can range over wffs. We will use the capital letters A, B, and 0 for this as a wff.s~nce we cannot tell if instances of this formula are conjunctions
purpose. A, B, and 0 are thus variables for wffs and are called metalogical or condItlOnals. The requirement to insert parentheses when using the
variables. The prefix 'meta' ('about') indicates that they are not part two-place connectives eliminates the possibility of such ambiguous
of the language in question-the sentential language-but are used to formulas. However, no ambiguity results when the outer parentheses
talk about the language in question. P, Q, ... have statements as are dropped, so to reduce the number of parentheses it is convenient to
instances; A, B, and 0 have well-formed formulas in the sentential adopt the convention of dropping the outer parentheses. Thus (1) will
language as values. Using these metaterms we may restate the formation be regarded as a wff though (2), of course, is still not a wff.
rules in this manner: . There is another way to reduce the number of parentheses that appear
m form~las. We can Tank the connectives in relation to scope. The
1. Any statement variable is a wff. connectIves may be ranked in this order:
2. If Ais a wff, then r-;A is a wff.
3. If Aand Bare wffs, then (A Y B) is a wff. First ~
4. If Aand Bare wffs, then (A A B) is a wff. Second -->-
5. If Aand Bare wffs, then (A -->- B) is a wff. Third A, Y
6. If Aand B i1re wffs, then (A ~ B) is a wff. Fourth
 34 Sentential Language

f-* has the highest rank, with A, V having the same rank. To illustrate

this convention, formulas with and without parentheses appear below:

(P A Q) ---+ P PAQ---+P
P f-* (Q ---+ P) Pf-*Q---+P
r-.;P ---+ (r-.-P V Q) ,,-,p ---+,....,P V Q
(P A Q) ---+ (P ---+ Q) P A Q ---+ (P ---+ Q) " . Earlier, in Ohapter I, the point was
P ---+ (Q f-* P) no change eCISIOn made that some well-formed for-
P V (Q A R) no change mu!as in the sentential language are
ro edu valId (have only necessarily true
It should be stressed that these conventions are practical moves to statements as instances). Examples
make it easier to write formulas and to see the structure of formulas. of such formulas would be the
Formally speaking, what counts as a wff in the sentential language is following:
what satisfies the formation rules.
With the sentential language in hand, we are now in position to PIIQ->-P P->-P
introduce the effective procedures by which it can be determined whether (P->-Q) A P->-Q
a well-formed formula is valid or not. Such procedures are introduced Examples of invalid formulas
in the next chapter. would be those below:
P PA,.....,P P->-Q
Exercises
(P->-Q) AQ->-P
1. Is V a term of the sentential language?
An important question we may
2. Select a formula from this chapter, rewrite it so that it is a wff, and then now ask is whether there is some
demonstrate that it is a wff.
effective or mechanical decision pro-
3. State precisely why each of the following formulas is well-formed or why it is cedure by which we can tell whether
not well-formed. ~ form?la in the sentential language
(a) r--.lP(---+ (Q V R)) IS valId or not. A decision pro-
(b) ,......,p V Q A r-.JR cedure is mechanical for determining
(c) (P ---+ A) t"'-'Q whether a formula is valid or not
(d) (Q ---+ R) ---+ ((P V Q) ->- (P V R)) when we could programme a com-
4. Use conventions to eliminate as many parentheses as possible. p~ting machine in such a way that
(a) (((P ---+ Q) A r--IQ) ---+ ,......,P) WIth respect to any formula it would
(b) ((,-....;p ---+ P) ---+ P) at some point give a 'yes' answer if
(c) (P ---+ (Q ->- (P A Q))) the formula Were valid or a 'no'
(d) ,.....,(P ---+ (P ->- Q)) anSWer if the formula were not
(e) (P ---+ (Q V (R A Q))) valid. If there is a mechanical pro-
(f) ((P V Q) ->- (P V Q)) f-* ((P V Q) ---+ (P A Q)) cedure for determining whether well-
formed formulas in the sentential
Answers language are valid, then the notion
1. No, V is neither a statement variable nor one of the statement connectives of a valid sentential formula is an
appearing at the beginning of Section 2.6. effective one. It turns out that there
are several mechanical decision pro-
3. (d) is not a wff since A ->- B is not a wff. Outer parentheses are needed.
cedures for determining whether a
4. (a) (P ->- Q) A r-->Q ->- ,......,P, (b) (,......,P ->- P) ---+ P, (c) P ---+ (Q ->- P A Q), sentential well-formed formula is
(d} no change, (e) P ---+ Q V (R A Q), (f) P V Q ---+ P A Q +--+ P V Q ->- P II Q. valid or not. Two of these methods
"

35
36 Decision Procedures 3.1 Truth-Tables
37
are discussed in this chapter: truth-table methods and transformation
in what follows is a variable for the positive integers 1,2, 3, ... ). One
into what is called conjunctive normal form.
term m~ans ~1 or t.wo lines, two terms means 22 or four lines, three terms
mean 2 or eIght hnes, four terms mean 24 or 16 lines, and so forth.
3.1 Truth-tables The truth-table for (P -+ Q) A P -+ Q is constructed as follows:

Our first procedure for determining whether a well-formed formula in P Q

the sentential language is valid or not is the truth-table method. To T T T T T
introduce this method we must first consider how one can construct a T F F F T
truth-table for any well-formed formula of the sentential language. The F T T F T
F F T F T
procedure is quite simple. Suppose we wish to construct a truth-table
for this formula:
Each line is obtained by following procedures discussed in the last
chapter. On the first line we are given Pas T and Q as T. Starting with
A truth-table for a formula is a table in which the possible combinations the connective with the narrowest scope, the first -+, we enter a T
of truth-value for the component variables of the formula are listed and under it since T -+ T is T. We next enter a value for the connective
the resulting truth-value of the formula for each possible combination is which now has the narrowest scope, namely A. Since TAT is T we
indicated. To make this clea~ we observe that formula (1) contains two place a T under A. The last -+ is the connective with the widest sc ope.
different statement variables. This means that there are four possible The value of the antecedent is given under the A , namely T . S'Ince
ways in which truth-value may be assigned to these statement variables, T -+ T is T we place a T under -+. The second -+ is the main connective
which may be set down as follows: (the connective with the greatest scope). The first line thus shows that if
Pis T and Q F, then the formula is T.
P Q A truth-table can be constructed for any sentential formula. How-
T T ever, given the rule that for a formula with n variables we need 2" lines
T F to provide for all its truth possibilities, truth-tables become cumbersome
F T when we have more than three variables.
F F
Exercises
If the formula under consideration had three variables, rather than two
there would be eight possible ways in which truth-value may be assigned 1. Construct a truth-table for the following formulas.
(a) P V ""P
to these statement variables, namely:
(b) P A ""P
P R (c) (P -+ Q) -+ (Q -+ P)
Q
(d) ""Q -+ "-'P +--t P -+ Q
T T T (e) P V (Q V R) +--t (P V Q) V R
T T F (f) ",,(P A P +--t P)
T F T (g) (P -+ Q) A (Q -+ R) -+ (P -+ R)
T F F (h) ",,(P A Q) +--t '"'-'P A '""-'Q
F T T (i) ""(""(Q -+ ",-,P) -+ (P -+ R)) -+ ",,(P A Q)
F T F
F F T 2. Suggest some ways to reduce the number of T's and F's in a truth-table.
F F F
Answers
In general if a formula contains n number of variables, then the possible
ways in which truth-value may be assigned to the components can be 1. You ~hould get all ~'s under the main connective of (a), (d), (e) and (g),
calculated in this way: 2" = number of lines on truth-table (n here and all!, s under ~he mam connectives of (b) and (f), and T's and F's under the
roam connectIve of (c), (h) and (i).
 38 Decision Procedures
Valid, Inconsistent, and Contingent Formulas
39
3.2 Valid, inconsistent, and contingent formulas
or contingent. If a formula always comes out F no matter what truth-
value is assigned to its component variables, then it will always come out
In Chapter 1 (Section 1.2) we noted that statements can be divi~ed ~nto
false no matter what the instance. Thus if a formula always has the
those which are necessarily true, e.g., 'It is false that Amenca IS a
value F under its main connective of its truth-table, it must be incon-
republic and is not a republic'; those which are necessa~'ily false,. e.g.,
sistent. By the same reasoning it follows that if a formula has a mixture
'America is a republic and is not a republic'; and those winch are neIther
of T's and F's under its main connective of its truth-table, if it has
necessarily true nor false but are true or false depending on e~isting
neither all F's nor all T's, then it is a contingent formula.
states of affairs, e.g., 'America is a republic'. The first group of state-
ments we may speak of as necessarily true, the second as inconsistent, In other words, if we wish to know whether A is valid, inconsistent, or
and the third as contingent. contingent, we prepare a truth-table for A. If all T's are found under its
Somewhat analogous to this last classification, well-formed formulas main connective, it is valid. If all F's are found under its main connective ,
it is inconsistent. If neither is the case, the formula is contingent.
of the sentential language can be classified as valid, inconsistent, or con-
tingent. This classification may be explained as follows: This procedure may be applied to the first examples above of valid,
inconsistent, and contingent formulas. In the tables we will place the
possible truth-values for P and Q directly under the variables in the
A formula is valid if and only if all instances of the formula are true formula:
statements.
A formula is inconsistent if and only if all instances are false (I) P y r-JP (2) P A r-Jp (3) P y Q
statements.
T T F T F F T T T
A formula is contingent if and only if by suitable choice of sub- F T T F F T T T F
stitution the formula can have true instances or false instances. F T T
F F F
Examples of valid formulas are found in the first group below, incon-
sistent formulas in the second group, and contingent formulas in the last We observe all T's under the main connective of (I), all F's under the
group. main connective of (2), and a mixture of T's and F's under the main
connective of (3).
py,.......,p P A,.....,P PYQ As noted above, contingent formulas can have true or false statements
P A Q-+ P ,....",r--.J(PA,......,P) P as instances. To illustrate this, note that the following necessarily true
(P-+Q)AP-+Q ,,-,(P A Q -->- P) P Y C. . ..JP A P) statements,

If a formula in sentential logic is such that it always has the value T If this object is red, then it is colored.
10 matter what truth-value we assign to its component sentential
2 + 2 = 4.
If all men are mortal and Socrates is a man, then Socrates is mortal.
variables then it must be a valid formula. We need not conS1'der
I

instances' of a sentential formula in determining whether or not it is are instances, respectively, of the following contingent formulas:
valid. For if a sentential formula always comes out true no matter what
truth-value we assign to its component variables, then all instances of P
P
the formula must be true statements since the truth-value of the state-
P
ment instance is solely a function of the truth-value of the component
statements. Truth-tables provide a way of determining whether or not
In addition, the following inconsistent statement is an instance of P
a formula comes out T no matter what truth-value is assigned to its and P A Q:
component statement variables. Consequently we can use truth-tables
to determine whether a formula in sentential logic is valid or not. In Some men are wise, and no men are wise.
turn we can use the tables to determine whether a formula is inconsistent A formula is a ta1ltology iff it takes the value T for every assignment
40 Decision Procedures
3.3 Arguments and Truth-Tables 41
of truth-values to its components. All valid sentential formulas are
The argument pattern of which (1) is an instance is
tautologies. However, when the sentential language is expanded to
include other kinds of symbols we will find logical formulas that are (2) P --+ Q
valid but that are not tautologies. Consequently it is useful to have ""Q
classification for formulas that take the value T for every assignment of ""p
truth-value, and these are designated to be tautologies. (The explanation
for exactly what is a component in any kind of logical formula-later (1) is obtained from (2) by substituting simple statements for P and Q,
to be called a basic component-is given in Section 6.1.) The notion of a namely
tautology in the sentential language is effective since there is an effective,
mechanical procedure for deciding whether or not a formula is a P: George needs a vacation.
Q: Classes should end.
tautology-construct the truth-table for the formula.

Exercises In Chapter 1 (Section 1.3) the point is made that an argument is

valid iff its corresponding conditional (conjunction of premises--+
1. Construct a truth-table for each formula and state whether the formula is
conclusion) is necessarily true. And a conditional is necessarily true if it
tautologous, inconsistent, or contingent.
(a) P --+ "" ""p is an instance of a valid logical formula. This being the case, the
(b) ",,(P V Q) --+ ""p A ""Q validity of (1) can be established by showing that
(c) (P --+ Q) --+,....,P V Q
(d) ",,(P --+ P) (3) (P --+ Q) A ~Q --+ ""P
(e) (P --+ Q) A (R --+ Q) --+ (P --+ R)
(f) P A (Q V R) ~ (P A Q) V (P A R) is a tautology. By using truth-tables we can quickly verify that it is a
(g) (P --+ ""P) --+ ""p tautology, thus establishing the validity of (1).
(h) (P A ""Q --+ R A ""R) --+ (P --+ Q) It is useful to put this in slightly different terms. Let us now speak
(i) P V F~ P
of valid argument patterns. An argument pattern is valid iff it cannot
2. In exercise 3.1.1 sort the formulas out in terms of being tautologous, incon- have an instance such that the premises are true and the conclusion is
sistent or contingent. (3.1.1 means Chapter 3, Section 1, exercise 1.) false. If (3) is a tautology, then it immediately follows that (2) is a valid
argument pattern. And if an argument pattern is valid, then any
Answers
instance of the pattern is valid. Showing (3) is a tautology thus es-
1. (d) is inconsistent, (e) is contingent, and the rest are tautologous. tablishes that (1) is valid.
2. See answers for exercise 3.1.1. To take a second example, suppose we wish to establish the validity
of the following argument:

If Tom is guilty, then if Robert is guilty, Jones is guilty.

3.3 Arguments and truth-tables
If Tom and Robert are guilty, then Jones is guilty.
Let
Truth-tables can be used to establish the validity of certain kinds of
P: Tom is guilty.
arguments, namely, those arguments whose validity is a matter merely Q: Robert is guilty.
of how statements are related by statement connectives. R: Jones is guilty.
Suppose we were asked to consider the validity of the following
(simple) argument: The argument pattern is

(1) If George needs a vacation, then classes should be ended. Classes P --+ (Q --+ R)
should not be ended. Therefore George does not need a vacation.
PAQ--+R
42 Decision Procedures 43

Using truth-tables one may confirm that its corresponding conditional is And since it is valid, the argument is valid. But in symbolizing the
valid; thus the argument is valid. argument in this way we did more than is needed if we follow the principle
We should note that if an argument pattern is invalid, it does not of correct paraphrase. No mistake is made in symbolizing the argument
follow that an instance of the pattern is invalid. For example in the manner above. But following the principle, we could just as well
have made these assignments:
P
Q P: Miller's statement is meaningful.
Q: Miller's statement is either true or false.
is clearly an invalid argument pattern. However, the instance of this
pattern below is a valid argument: and in so doing we would have expressed the logical structure relevant to
the validity of the argument. For the argument is valid if and only if
Some men are lovable.
Some lovable things are men.

In the application of sentential logic to arguments (and to what is to

come, predicate logic), the principle for correct paraphrase or correct
translation must be observed. This principle may be stated as follows: is valid.
It is not too difficult to so imagine the context that the following
Principle for correct paraphrase: The correct paraphrasing of a would be a valid argument,
statement into the formal language is relative to a given argument
context. The statement is correctly paraphrased when the structure (4) Either we will disarm or there will be nuclear war.
of the argument relevant to validity is exposed. So since we will disarm, there will be no nuclear war.

The force of this principle will only be appreciated as we build the formal If the first premise were symbolized P V Q, this would violate the prin-
language so that we are in position to expose the logical structure of a ciple for correct paraphrase. Insofar as (4) is valid, the first sentence
simple statement. At this time the language is rather limited since all we must be understood as follows: Either we will disarm or there will be
have are sentential variables and statement connectives. However, a nuclear war and it is false that we will disarm and there will be nuclear
few simple examples will help to illustrate the application of this principle. war. Thus to expose the structure of (4) relative to the validity of (4),
Suppose we wish to establish the validity of the following argument: the first sentence may be paraphrased as;
If Miller's statement is meaningful, then it is either true or false.
(P V Q) A r-.J(P A Q)
His statement is neither true nor false.
His statement is meaningless. or, simply,

If we follow the procedures used up to this time, we will let sentential r--..;(P AQ)
variables replace simple statements, for example
Earlier we noted that
P: Miller's statement is meaningful.
Q: Miller's statement is true. Some men are lovable.
R: Miller's statement is false. Some lovable things are men.
The argument pattern obtained would thus be
is a valid argument and it is an instance of
P-+QVR
,--,(Q V R) P
r-.,/p Q
-"
 44 Decision Procedures 3.4 A Short-Cut Truth-Table Method 45
If, for purposes of applying validating techniques of logic to this argu- 3.4 A short-cut truth-table method
ment, it were symbolized in this way, then the principle for correct
paraphrase would be violated. The validity of this argument does not Truth-tables enable us to decide in a mechanical way whether a formula
rest merely on how statements are related by statement connectives. is a tautology. But with formulas having more than three different
Rather it rests on the meaning of 'some' and on how the expressions sentential variables the tables pass beyond what is reasonable for a
'men' and 'lovable things' are related. We are not at this time in position person to do. For example, to determine the validity of this moderately
to correctly paraphrase this argument. We must wait until additions simple (valid) argument pattern
are made to the sentential language that will allow us to expose the
inner logical structure of simple sentences. (These additions are made (1) P -+ C--'Q -+ R)
in Chapter 5.) Q-+SAT
,....,T
Exercises
1. Use truth-tables to test the validity of the arguments in Section 2.5.2.
2. Let us call a sentential argument an argument whose validity is merely a we must have an unwieldy table with 32 rows. A short-cut method of
matter of how statements are related by statement connectives. Symbolize determining validity is much in order.
and test the validity of those arguments below that are sentential arguments. There are several short-cut truth-table methods. The one selected
If the argument is sentential, be sure that letters are used only for simple here may be outlined as follows: Let A be a well-formed formula. We
statements.
(a) If either Nixon or Rockefeller wins the nomination, then McCarthy will suppose that A is not a tautology and proceed to fill in truth-values for
be nominated and a Democratic victory will be assured. Therefore either the components of A until either a contradiction results or we can
Nixon will not win the nomination or McCarthy will be nominated. consistently assign a truth-value to each of its components. If a contra-
(b) Only servants of the Lord are missionaries. Not all missionaries are diction results, then our initial assumption that A is not a tautology is
Catholics. Therefore at least one servant of the Lord is not a Catholic. false. Hence A is a tautology. If we can consistently assign a truth-value
(c) Either D or both Rand F. If D, then F. Therefore F.
to each of the components, then A is not a tautology. This method is
(d) P provided that Q if P. If P, then Q. Therefore P.
(e) Not Q. Therefore not P unless if Q then R. called the reductio ad absurdum truth-table test. Often in mathematics and
(f) If P then Q and if ,....,p then R, so it is false that both Q and R. philosophy one proves P true by assuming not P, and from this assump-
(g) P and not-Q imply Rand not-R, so P implies Q. tion deriving a contradiction. If not P does imply a contradiction, then
(h) If you love her and give her gifts, then she will fall in love; but if you P must be true. The procedure in employing this useful technique to
love her and do not bring her gifts, then she will not love you. Hence if test the validity of sentential formulas consists of two steps:
you love her, she will love you if and only if you bring her gifts.
(i) Anderson did not marry a beautiful woman because he is happy. For
1. Assume the formula is not a tautology by placing an F under the
if he is happy, then he is not jealous, and we know that if someone
main connective.
marries either a beautiful woman or a very popular woman, then he will
2. Follow out the consequences of this. If we arrive at a contra-
be jealous and disappointed.
diction, the formula is a tautology; if we do not, it is not a
3. Prove the following: tautology.
A
(a) If A -+ B is a tautology, then 13 is a valid argument pattern.
]'01' example, we take this formula and assume it is not a tautology
A A (step 1):
(b) If 13 is a valid argument pattern, then any instance of 13 is a valid
argument. (P -+ Q) -+ (,--...,Q -+ ,....,P)
F
Answers
2. Only (b) and (i) are not sentential arguments. All the sentential arguments An F can appear under the -+ in the above formula only if the ante-
are valid except (f). cedent is true and the consequent is false. P -+ Q must be T and
,.
46 Decision Procedures Equivalence and Transformations 47

~Q -+ ,-....;p must be F, which we will now indicate below: F for P is consistent. So the formula is invalid. Note that had putting
F for Q -+ R led to a contradiction, we could not have concluded from
(P -+ Q) -+ C--Q -+ r--.;P) that cdonc that the formula is valid. In such a case we must also try
'1' F F T for Q -+ R, which might yield a consistent assignment to the whole
formula and hence show it to be invalid. So whenever a choice is open,
If "'-'Q -+ r--.>P is F then its antecedent must be T and its consequent F, (tU alternatives must lead to contradictions if the formula is valid.
and this can only occur if we place F for Q and T for P as follows: The validity of the argument pattern (1) that appeared at the
beginning of this section may be demonstrated by using the rcductio test
(P -+ Q) -+ C""'Q -+ r-...;P) on its corresponding conditional as follows:
'1' F '1'F F F'J'
(P -+ (,-....,Q -+ R)) A (Q -+ SA '1') A "",,'1' -+ (P -+ R)
All the occurrences of Q must have the value F and those of P, T, so: '1''1' '1'FFF F'1' FF '1'FF'1'FF

(P -+ Q) -+ C.....,Q -+ ,-....,;P)
'1' '1' F F ,...., F F ,....,'1' Exercises
1. Use the red1lCtio method and state whether the following formulas are
But now we have a contradiction-which may be indicated by a line- tautologous.
for P -+ Q cannot be T if P is T and Q is F. Thus our original assump- (a) P V P-+ P
tion that this formula is F must be false. It is T and is valid. All this (b) (Q -+ R) -+ (P V Q -+ P V R)
(0) (P <-+ Q) -+ (P -+ Q) A (Q -+ P)
can easily be wri.tten on one line. For example, the proof of the in- (d) (,-....;p -+ R) A ",-,R -+ P
validity by the rcductio method of (e) ,....,(P A Q) -+ ",-,P
(f) (P A "-'Q -+ Q) -+ (P -+ Q)
(g) (P V Q) A (P -+ R) A (Q -+ R) -+ R
(h) (P<-+Q)<-+ (P A Q) V (,,-,P A ",-,Q)
(i) (P -+ Q) V (P -+ R) -+ (P -+ Q A R)
would proceed as follows:

(P-+Q) A (R-+S) A (r-..;R VS)-+ (Q-+S) Answers

'1' '1' '1' F '1' F '1' '1' F '1' F F '1' F F
1. All but (e) and (i) are tautologous.
We can consistently supply either Tor F for P in completing the assign-
ment of truth-values. Thus following out the consequence of an F under
the main connective has not resulted in a contradiction; the formula is
not valid. 3.5 Logical equivalence and transformations
In using the rcdnctio test one will occasionally encounter choice-
points. For example, the J'cdnctio test proceeds mechanically on the Two formulas in logic, A and B, are said to be logically cqnivalent if
formula below up to the point indicated: and only if A <-+ B is a valid formula. With respect to sentential
formulas, A and B are logically equivalent if and only if A <-+ B is
(P V S) V (Q -+ R) -+ (P V Q -+ (S -+ R)) tautologous. For example, the following pairs of formulas are logically
'1' '1' '1' F F '1' F '1' F F equivalent:

No further assignment of values is determined. We can, for example, P AQ and Q A P (P V Q) V Rand P V (Q V R)

put Tor F for Q -+ R. Let us put F, then Q is T and putting either Tor "'-' "-' P and P P -+ Q and ",-,P V Q

"
48 3.5 J,ogical Equivalence and Transformations 49

since connecting them with a biconditional produces a formula that can be transformed into
is tautologous. For example, that P -+ Q is logically equivalent to
r-..Jp V Q is established in the table below. r-..J(P A Q) V R

since law 13 states that a conditional P -+ Q is equivalent to r-..Jp V Q.

P -+ Q ~r-..JP V Q
Similarly, the last formula can itself be transformed into
T T T T F T T
T F F T F F F
F T T T T T T
F T F T T T F
by using law 5, DM. And erasing double denial using law 2, DN, we
can obtain:
Truth-table methods can be used to verify that each of the bi-
conditional formulas below is valid. These are well-known and useful
equivalences that have been given names (and abbreviations), as has been
indicated. We will make frequent use of these equivalences in what Such transformations can continue indefinitely. In transforming A into
follows. B we use biconditional formulas that are valid. B obtained in this way
is equivalent to A since we come upon B by successively replacing
Law 1 (P -+ Q) ~ (,-..,Q -+ r-vP) Contraposition (Con) formulas with formulas that are their equivalents.
Law2 P~r-..Jr-..JP Double Negation (DN) If we made all the steps involved in these transformations explicit,
Law 3 P Q~ Q V
V P} Commutation (Comm) we would see that we make use of two rules, a rule for substitution and a
Law 4 P A Q~Q A P rule for replacing formulas with their logical equivalents. These rules
Law 5 P V Q ~ "-'(r-..JP A ,-...,Q)} de Morgan's Theorems may be set down as follows:
Law 6 P A Q~r-..J(r-..JP v,-...,Q) (DM)l
Law 7 (P V Q) V R ~ P V (Q V R)}
Association (Assoc) Rl. Rule of substitution: A simple or compound sentential formula
Law 8 (P A Q) A R ~ P A (Q A R)
may be substituted uniformly for a single sentential variable.
Law 9 P V (P A Q) ~ P}
Absorption (Abs) R2. Rule of replacement: If A is equivalent to B then A may be
Law 10 P A (P V Q) ~ P
replaced by B whether A is alone or part of a formula.
Law 11 P A (Q V R) }
~ (P A Q) V (P A R)
Distribution (Dist) Returning to our first transformation, we are justified in moving from
Law 12 P V (Q A R)
~ (P V Q) A (P V R)
Law 13 P -+ Q ~ r-..Jp V Q Implication (Imp)
Law 14 (P ~ Q) ~ (P -+ Q) A (Q -+ P) Equivalence (Equiv) to
Law 15 P V P~ P}
Idempotent (Idem)
Law 16 P A P~P

given the rule of replacement and the logical equivalence below:

The first use of these laws will be to transform formulas. If A is
logically equivalent to B we may transform A into B. For example, the P A Q -+ R ~ r-..J(P A Q) V R
conditional
This equivalence is obtained by substituting P A Q for P and R for Q in
law 13, implication. The next transformation involves substituting
,,-,(~ A Q) for P, and R for Q in law 5, and using the rule of replacement.
Erasmg double denial uses law 2, the substitution of P A Q for P and
1 After the English logician Augustus de Morgan (1806-78). again, the rule of replacement accounts for the final transformation:
 Decision Procedures 3.5 Logical Equivalence and Transformations 51
50
The justification for the above transformations may be explicitly set RI nor R2 nor substitution will any longer be cited in making trans-
out as follows-allowing AI B to signify that A is substituted for B. formations.
Let us now employ this short-cut in making some transformations
1. PAQ--+R given and reduce further the amount of writing by omitting line numbers.
2. P --+ Qf-+ r-IP Y Q Imp
3. P A Q --+ Rf-+ r-I(P A Q) Y R P AQIP,
(P --+ Q) --+,.......,R
R/Q, RI
r-IP Y Q --+ ,.......,R Imp
4. r-I(P A Q) Y R 1,3, R2
r-I(r-IP Y Q) Y r-IR Imp
5. P Y Qf-+ r-I(r-IP A ,.......,Q) DM DN
r-I('---P Y r-I r-IQ) Y r-IR
6. ",,(P A Q) Y R f-+ ,.......,("" ",,(P A Q) A r-.JR) ",,(P A Q)/P,
(P A ,-.."Q) Y r-IR DM (law 6)
R/Q, RI
,......,R Y (P A r-IQ) Camm (law 3)
7. r-I("""'" r-I(P A Q) A ,.......,R) 4,6, R2
(r-IR Y P) A (('-...IR Y "",Q) Dist (law 12)
8. P f-+ r-I "",p DN
9. P A Qf-+ r-I ,.......,(P A Q) P A Q/P, RI
10. t".I((P A Q) A t".IR) 7,9, R2 Note that to use DM, to change a negated disjunction into a con-
junction, we must have a formula that is an instance of ,.......,(r-.JP Y ,.......,Q).
On the right we have indicated the justification for each line. A justifi- ,-...;( ""p V Q), the first disjunct of the third line, cannot be transformed
cation consists of the given formula, a law, or a use of RIoI' R2. In other using DM. However, using DN on ,.......,(""P V Q) we obtain '--'C--'P V
words, a transformation is a sequence of lines, each line either being "",......,Q), whichis a substitution instance of ""(",,P V ,.......,Q)(""QIQ) and
given, being a law, a substitution for a law (RI) or a line obtained by thus we transform ",,(,.......,P V Q) into P A ""Q using first DN and then DM.
using R2. It should be noted that the first two times R2 is used, B These transformation techniques can be usefully employed to prove
replaces A where A is the whole formula, while the last time R2 is used, that one formula is logically equivalent to another. Suppose we wish
A is part of a formula. to show that A is logically equivalent to B. If we can start with
In transforming formulas using our equivalence formulas, it will
prove convenient if we take certain short-cuts. To move from

(I)PAQ--+R and successively transform the A on the right into B we will have shown
to that A f-+ B is a tautology. For example, we can show that (P --+ Q) --+
(2) r-I(P A Q) Y R (Q --+ P) is logically equivalent to (P A Q) Y ",,(Q A ""P) as follows:

involves, as we have seen, using law 13, substitution, and the replace- (P --+ Q) --+ (Q --+ P)
ment of an equivalent by its equivalent (R2). We may, for the sake of f-+ (P --+ Q) --+ (Q --+ P)
brevity, now omit setting out the lines that involve law 13 and making f-+",,(P--+Q) Y (Q--+ P) Imp
f-+ ,.......,(""P V Q) Y (""Q Y P) Imp (2)
substitutions, and move directly from (1) to (2), giving as our justi- f-+ ""-,(",,P Y "" ""Q) Y (""Q V P) DN
fication law 13, or Imp. In such a context we can regard Imp as indicating f-+ (P A ""Q) Y ,....",("",....",Q A ""P) DM (laws 6 and 5)
the collapsing of four lines into two. Similarly, to move from, say, f-+ (P A ""Q) Y ",,(Q A ,....",P) DN

RYR--+Q Every tautology is logically equivalent and every inconsistent

to formula is logically equivalent. Let us represent a tautology by T and
an inconsistent formula by F. Now what can be said about a dis-
junction that has an inconsistent formula as a disjunct? That is, what
involves law 15, RI (RIP), and replacing equals with equals (R2). But Can be said about:
we will collapse these steps and move directly from the first to the
second, giving as our justification law 15, Idem. Consequently, neither PYF
,.
52 Decision Procedures
3.5 Logical Equivalence and Transformations
53
The disjunction is true if P is true; otherwise it is false. It therefore is
Exercises
logically equivalent to just P. Similarly, if one of the conjuncts of a
conjunction is a tautological formula, if we have 1. Use truth-tables to show the 20 laws are tautologous.

PAT 2. State the ways the following uses of DM, Con or Imp are erroneous, though
each pair is equivalent.
then the conjunction is true if P is true and false if P is false. Therefore (a) ,.....,(P V Q)
PAT is logically equivalent to just P. These truths are expressed in ,-....; ,--.,(,.....,P A ,.....,Q) DM
the next two laws below. (b) P V Q
r--.JP-+Q Imp
Law 17 P A T+-+ P Dropping tautologies (DT) (c) r--.J(P V r--.J ,-.....Q)
Law 18 P V F +-+ P Dropping inconsistencies (DI) ,-.....,--.,(,--.,p A ,-...,,.....,,.....,Q) DM
(d) "-'P V f'-'Q
One of these laws-DT-is used in the following proof of equivalence by ,.....,(P AQ) DM
transformation. We wish to prove that (e) P -+,.....,Q
,.....,Q-+ P Con
PVQ-+PAQ (f) ,.....,(P V Q) V P
,.....,,.....,(,.....,P A ,.....,Q) V P DM
is logically equivalent to P +-+ Q.
3. By transformation prove the following pairs of formulas are equivalent. For
P V Q -+ P A Q+-+ P V Q -+ P A Q at least one transformation take no short-cuts.
+-+ ,.....,(P V Q) V (P A Q) Imp (a) P+-+,.....,Q
+-+ ,.....,(,.....,,.....,P V ,....., ,.....,Q) V (P A Q) DN (2) Q+-+ ,.....,P
+-+ (,.....,P A ,.....,Q) V (P A Q) DM
(b) ,.....,(R -+ ,.....,S)
+-+ «,.....,P A ,.....,Q) V P)
A «,.....,P A,.....,Q)VQ) Dist RAS
+-+ (P V (,.....,P A ,-..."Q)) (c) P A (P-+Q)
A (Q V (,.....,P A ,.....,Q» Comm (2) PAQ
+-+ «P V ,.....,P) A (P v,.....,Q» (d) ,.....,(P -+ ,.....,Q)
A «Q V ",-,P) A (Q v,-...,Q)) Dist (2) (P A Q) V (P A Q)
+-+ (P V "'-'Q) A (Q v,.....,P) Comm,DT(2) (e) P-+Q V,.....,Q
+-+ (,....",.....,p V ,.....,Q) A (,.....,,.....,Q V ,.....,P) DN (2) T
+-+ (,.....,P -+ ,.....,Q) A (,-....;Q -+ ,.....,P) Imp (2) (f) (P -+ Q) V (P -+ R)
+-+ (Q -+ P) A (P -+ Q) Con (2) P-+QVR
+-+ (P-+Q) A (Q-+ P) Comm (g) P A (P V Q)
+-+P+-+Q Equiv P
(h) (P -+ Q V ,.....,Q) A W
Be sure to see why the uses of DN are necessary. W
If any disjunct of a disjunction is a tautology, the whole disjunction (i) (P A (Q A S» V (,.....,P A (Q AS»
is a tautology, and if any conjunct of a conjunction is inconsistent, so is QAS
the entire formula. Again, letting T stand for a formula that is always (j) ,.....,(P A ,.....,(P A ,.....,(P A Q»)
true, a tautology, and F stand for a formula that is always false, an ,.....,p V "-'Q
inconsistent formula, these truths may be expressed in the following (k) (P -+ Q+-+ "-'Q -+ ,.....,P) A R
laws: R
Law 19 P V T+-+ T (1) (,.....,p V (Q V "-'Q» A R
Law 20 P A F +-+ F R
54 Decision Procedures 3.6 Other Two-Place Operators
55
Answers exception of --+, we will allow the column numbers to be symbols for the
possible two-place operators.
3. (j) r-J(P /\ r-;(P /\ r-J(P /\ Q)))
r-JP V r-J r-J(P /\ r-;(P /\ Q)) DN,DM
r-JP V (P /\ r-J(P /\ Q)) DN A (1) B =dfA--+A A (9) B =df A --+,,-,B
r-JP V (P /\ (r-JP V r-JQ)) DN,DM A (2) B =df r-JA --+ B A (10) B =df (A --+ B) --+ ",-,(B --+ A)
(r-JP V P) /\ (r-JP V (r-JP V r-JQ)) Dist A (3) B =df B --+ A A (11) B =df r-JB
r-JP V (r-->P V,--.,Q) Comm, DT A (4) B =dfA A (12) B =df ,-....,(A --+ B)
(r--'P V r-JP) V r-JQ Assoc A (6) B =df B A (13) B =df r-JA
r-->P V r-JQ Idem A (7) B =df r-J( (A --+ B) A (14) B =df r-J(B --+ A)
--+ r-J(B --+ A)) A (15) B =df r--'(,-....,A --+ B)
A (8) B =df r-J(A --+ r-JB) A (16) B =df ,-....,(A --+ A)

3.6 Other two-place operators Of special interest is the ability to select one of the columns in the
above table and define denial and the remaining connectives in terms of
The four symbols V, /\, --+, and +--+ are constants that act on pairs of this selected column. This is possible by selecting column 9 or column
things. We may speak of them as operators and say that they act on 15. Using the sym~ol I-called the Sheffer stroke-for the two-place
pairs of truth-values. For example, T V F acts on the pair T and F operator correspondmg to column 9, let us see how we can define,-...., V
and /\ in terms of /. ' ,
such that the result is T. Furthermore we may speak of these symbols
as two-place truth-value operators since they relate pai1'8 of truth- If we again look at the table for possible combinations for a two-
values. So far only four two-place truth-value operators have been con- place operator, we see that P / Q has this matrix:
sidered. But there are other possible two-place truth-value operators,
and the table below sets out all the possible combinations for a two-place P Q
truth-value operator: T F T
T T F
F T T
P Q v ->- +--+ A V -1--+ t F T F
T T T T T T T T T T F' F' F' F' F' F' F' F'
T F' T T T T F' F' F' F' T T T T F' F' F' F'
F' T T T F' F' T T F' F' T T F' F' T T F' F'
F' F' T F' T F' T F' T F' T F' T F' T F' T F' If a. formula can be found using only this connective and it is logically
(I) (2) (3) (4) (5) (6) (7) (8) (9) (IO) (Il) (I2) (I3) (I4) (I5) (I6)
eqUIvalent to a formula using only r-J, then we could define r-J using /.
P I Q naturally reads 'not both P and Q'. This indicates that P / P is
Each of these columns would allow us to introduce a two-place con- equivalent to r-J(P /\ P). ,-....,(P /\ P), by Idem, reduces to ",-,P. Thus
nective. For example, we might introduce V and define it by column we may use this definition for "'-' in terms of /:
10. This symbol so defined would roughly correspond to English 'or'
when we so use it that P or Q is true iff either one or the other disjunct
is true, but not true when both are true. Or we could introduce the
symbol-t--+ for column 12. P -t--+ Q could be read "P does not imply Q"
since P -t--+ Q is equivalent to P /\ ~Q or r-J(P --+ Q). Further~ore we can obtain such definitions for the four familiar two-place
Using denial with --+, V, or /\, one can define the remaining 15 two- connectr;es an~ .indeed for the 12 other connectives. One way to
place operators indicated in the above table. If, for example, a formula proceed III obtammg these is as follows: Take V for example. P / Q, as
using only r-J and --+ can be found equivalent to a formula using only V, w~ have seen, can be written as ",-,(P /\ Q). This can be transformed by
then we could define V using r-J and --+. Let us use r-J and --+ and define usmg DM into a formula only using V and r-J: ",-,P V "'-'Q. From
the r.j:lmaining 15 two-place operators. For convenience, but with the "",p V ,-....,Q we can obtain just P V Q by negating each disjunct; thus
 Decision Procedures
"""Normal Forms and Testing for Validity 57
56
if we do this to P I Q-negate each constituent-we should obtain this Answers
equivalence: 1. P! Q may be read "neither P nor Q."
2. (b) F, (f) T, (g) T, and (h) T. (f),(g), (h), and (i) are tautologies.
4. There will be 2 2n truth functions for n variables. A three-place truth-
Now we can remove the ,-..., using the above definition, thus obtaining a functional operator will thus have columns for 256 operators.
definition for V using only I:

A V B =df (A I A) I (B I B) 3.7 Normal forms and testing for validity

Since PIP is equivalent to ,-...,P, (P I Q) I (P I Q) is equivalent to There is a second procedure by which one can decide if a sentential
,-...,(P I Q). And since PI Q means 'at least one of P and Q is false', formula is valid. We can use the equivalence laws to reduce any formula
,-...,(P I Q) means 'it is false that at least one of P and Q is false' or 'both to a normal form. There are several normal forms customarily treated
P and Q are true'. Thus in beginning logic books. In this section our interest is in the normal
form called the conjunctive normal form (abbreviated CNF). Using
A II B =df (A I B) I (A I B) CNF to determine whether a formula is valid is tedious business,
especially when a method such as the reductio test is available. However,
Exercises reducing formulas to normal forms provides good practice in using the
1. Give a reading for P ! Q and for the other unfamiliar operators indicated in equivalence laws and, as we will see in Part 2, CNF's can be highly
the above table. important for theoretical purposes in logic.
To define the notion of a conjunctive normal form it is well to start
2. Determine the truth-value of the following formulas given P: T and Q: F.
with what we will take to be a single formula. By a single formula we will
Which of the formulas are tautologies?
understand either a sentential variable or a negated sentential variable.
(a) PIQ
(b) (P ~ P) I ,-...,Q Double negated formulas will not count as single formulas. If we let
(c) P!,-...,Q AI> ... , An be a list of n single formulas, then we will call an elementary
(d) (P I Q) I (P I Q) disjunction a formula of the form
(e) ,-...,(P! Q) I (,-...,p
! ,-...,Q)
(f) (,-....;p<,---} P) I P
(g) (P ~ Q) ~ ,-...,(P-t-+ Q)
(h) (P (16) Q)<,---} P II'-""P
Examples of elementary disjunctions would be
(i) (P (14) Q) ~ Q II ,-...,P
3. For two- place truth-value operators there are 16 possible operators or con- PvQ
nectors, as the table above shows. Construct a table for possible one-place P vQ v,-...,p
truth-value operators. P
,-...,p
4. How many possible truth functions are there for n variables?
5. Use only'-"" and V and then use only'-"" and II and define the remaining 15 'l'he last two examples above are the limiting cases where n = 1. Thus
connectives. a single formula alone will count as an elementary disjunction.
6. Define the remaining 15 connectives by only using I and then by only We can now say a formula is in conjunctive normal form if and only
using !. if it has the form

7. Show why one cannot define the remaining two-place connectives by using
either II and V, ~ and II, 01'"-* and v.
58 Decision Procedures Normal Forms and Testing for Validity 59

where AI> ... , An are elementary disjunctions. Where n = 1, all of the transformations:
above examples will count as being in conjunctive normal form. Other
examples of CNF are (P -+ P V Q) A (P V Q -+ P) Equiv
(,-...;p V (P V Q)) A (,-...;(P V Q) V P) Imp
(,-...;p V (P V Q)) A (,-...;(,-...;,-...;P V ,-...; ,-...;Q) V P) DN
(P V r-.;Q) A (P V ,-...;Q)
(,-...;p V (P V Q)) A «",-,P A r-;Q) V P) DM
,-...;P A (Q V R)
(",-,P V (P V Q)) A (P V C---P A ",-,Q)) Comm
(R V ,-...;Q) A (P V ,-...;R V S)
(",-,P V P V Q) A (P V ",-,P) A (P V ,-...,Q) Dist

Ignoring the cases where n = 1, a formula is in CNF if it is a conjunction The last line is the CNF of the formula we began with. We should note
of disjunctions and all negation signs negate single variables. It would be that some formulas have more than one CNF.
well to note that the following are not in CNF: If a formula is valid, then its CNF is valid. In order for a CNF to be
valid, each of the conjuncts must be valid. Each conjunct of a CNF will
(,-...;p A Q) v,-...;R (P A Q) V (P A R) be valid if and only if a variable and its negation appear in the con-
V Q)
,-...;(,-...;p r-.;(R V Q) A (P V ,-...;P) junct. (The conjuncts, we remember, are disjunctions, and a disjunction
is valid iff a variable and its negation appear as disjuncts.) The first
Any propositional formula can be put into CNF by following these transformed formula above is valid, but the second is not valid because
steps in this order: in its CNF, though a variable and its negation appear in the first two
conjuncts, such a combination does not appear in the last conjunct-
(1) Use Imp and Equiv to obtain a formula containing only the (P V "'-'Q). If a formula is in ONF, we can tell by simple inspection
connectives,-...;, V, and A. whether it is valid. To test P -+ (Q -+ P A Q) by reduction to ONF we
(2) Use DM and DN to remove all negations outside parentheses.
(3) Use DN to remove all double negations.
proceed as follows:
(4) Use distributive laws as many times as necessary to produce a
conjunction of disjunctions of single formulas. P -+ (Q -+ P A Q)
,-...;P V ("'-'Q V (P A Q)) Imp
,-...;P V «,-...,Q V P) A (,-...;Q V Q)) Dist
It should be remembered that since (P V P) +-+ P and (P A P) +-+ P, (",-,P V ,-...;Q V P) A (",-,P V ,-...;Q V Q) Dist
we may replace P V P by P at any time and PAP by P.
To put Q -+ (P -+ (Q -+ P)) into CNF we proceed as follows, first At a glance we see the ONF of P -+ (Q -+ P A Q) is a tautology; thus
removing -+. this formula is a tautology.

Imp Exercises
1. Determine the validity of each of the following by putting it into CNF.
Since we have a string of disjunctions, we can drop parentheses, which (a) P-+ (Q-+ P)
(b) Q -+ (P -+ (Q -+ P))
gives us (0) P A (P -+ Q) -+ Q
(d) (P V Q -+ ,....."R) -+ (R -+ P) V (R -+ Q)
(e) (P -+ Q) -+ (P A R -+ Q A R)
(f),-...,(P A Q -+ «P -+ R) -+ Q))
(g) (P +-+ Q) -+ ,,-,(P A ",-,Q)
This is an elementary disjunction. Since in the limiting case where
n = 1 an elementary disjunction is in CNF, we can stop, the formula is 2. One can obtain a disjunctive normal form (DNF) of a formula by expressing
in CNF. the disjunction of the formula that describes its true cases. To apply this
method to, say,
The next example below uses all four steps. The problem is to
transform P +-+ P V Q into CNF. It can be put into CNF by these (P -+ Q) +-+ (Q -+ P)
-./ ..
60 Decision Procedures Normal Forms and Testing for Validity 61

consider its truth-table and the cases in which the formula comes out true. (b) How could you show that any of (1) to (5) in (a) above could be used to
It comes out true only in two cases, namely express all formulas using one-place, two-place, three-place, or any
truth-functional connectives?
P Q
T T
F F Answers

Now express each case as a conjunction with the variable for true and the 1. All are valid except (d) and (f).
negation of the variable for false. Thus in this example we have
2. (a) (b) (P II Q) V (P II ~Q) V (~P II Q) V (~P II ~ Q)
PIIQ (b) A DNF is valid iff it has 2 n disjuncts where n equals the number of
~P II~Q component variables in the formula.

Finally, link the resulting formulas by using as many disjunctions as neces-

sary. Doing this with the example, we obtain 3. (b) Any formula employing truth-functional connectives has a truth-table.
Thus every formula has a DNF. And thus any formula using one-place,
(P II Q) V (~P II ~Q) two-place, three-place, and so forth, truth-functional connectives can be
expressed using ~, V, and II. Since V can be defined by using only f"oo.J
which is the DNF of the original formula. To take another example, the true and II, any formula can be expressed using the pair ~ and II. Since II
can be defined using ~ and V, any formula can be expressed using (2),
cases for
and so on, through (5).

are
P Q
T T Supplement: Logical networks
T F
F T Today a great deal of calculation is done by computing machines. A
F F modern digital computer is to a large extent a complicated logical
thus the DNF for the formula is machine. Certain of the components of computing machines can be
(P II Q) V (P II ~Q) V (~P II Q) V (~P II ~Q) regarded as performing logical operations. We shall not be interested in
the electronic details of these components. Rather we will see how
(When a formula has only one true case, the resulting conjunction is counted sentential logic can be applied to simplifying a circuit.
as the DNF, and when a formula has no true cases, it is equivalent to Electronic computing machines can be built of the following switches
P II ~P, which is counted as being in DNF.) among others:
(a) Put the formulas found in Exercise 3.7.1 into DNF.
(b) When and only when is a formula in DNF valid?
3. With the use of certain equivalence laws any formula can be put in CNF, (1) Inverter switch: This is a device with one input terminal and
and with the use of truth-tables any formula can be put in DNF. one output terminal. When the input is hot (voltage), the output
(a) How can this fact be used to show that any formula using two- will be cold (no voltage), and vice versa. This component acts
place statement connectives can be expressed using (1), (2), (3), (4), or like the connective ~.
(5) below: (2) Or-switch: The or-switch has two or more inputs and one output.
(1) ~, II The output is hot, unless all inputs are cold. This component
(2)~, V acts like the connective V.
(3) ~,-*
(3) And-switch: There are two or more inputs and one output. The
(4) I output is hot when all inputs are hot; otherwise it is cold. This
(5) t component acts like the connective II.
""'/-,
62 Decision Procedures 3.7 Normal Forms and Testing for Validity 63

These three components can be represented by the following figures: to the above network is

INVERTER SWITCH OR-SWITCH (P A '""-'Q) V R

p-G-"'p P--[J --p

Q--
v vQ Given the truth-value for the variables above, we can calculate that the
output is T (hot). To take another example, what is the output of the
following when P is F and Q is T?

•
P--+--- G
p-D
!O'
AND-SWITCH
A II!

AQ I ..
A --p

D
]IIi

Q- .
Q------ !III

If more than two inputs go into an box, then we have a string of con-
A
With A as the main connective, we see that the formula will be
juncts, and if more than two inputs go into a V box we have a string of
disjuncts. For example ,-....;P A ( V )

The V switch connects P and Q, so the formula is

A _ A A B A 0 ,.....,.,p A (P V Q)

0- which, given the above truth-values, shows that the output is T (hot).
Sometimes the formula corresponding to one network may be logically
Now if we allow hot to correspond to T and cold to F, we can see one equivalent to a formula that is simpler. ,-...,,(P V ,-....;Q) is equivalent to
application of logic to these components. For example, suppose we have ,-....;P A Q. Networks corresponding to these formulas would be
this arrangement:
(1) P - - - , " , , " 1 1 0 ' ' _

:---D-~D v

R------------------------------~·.-
Clearly, other things being equal, (2) is a simpler network than (1).
When networks have the same formula, they will have the same hot or
where Pis T, Q is F and R is F. Will the output be hot or cold? To cold output. Thus network (2) may be substituted for network (1) in
answer this we need first to figure out the formula analogous to this the circuit with a reduction in cost and complexity, if, again, other
things are equal.
network. This is easily done. The last box gives us the main connective
Let us measure simplicity of a circuit in terms of the number of
v. This V connects R and a conjunction, as follows:
statement connectives in circuit formula. Thus A is simpler than B iff
A ) V R A has fewer connectives than B. To illustrate, ,-....;(P A Q) is simpler than
",P A "-'Q since the former has two connectives, while the latter has
The conjunction is made up of P and ,-....;Q. Thus the formula analogous three connectives.
~-I ,
64 Decision Procedures Normal Forms and Testing for Validity 65

Generally, to find a simpler circuit formula one will find useful most
of the 20 laws set down in Section 3.5. For an illustration of how these
laws can be used in simplifying formulas, consider the circuit below:
(e) P -It-'-- B [J-
p--+---0 , Q ------------------------~..- v ___

Q........--LG~-
R ---------------------------------------..
-

_[J-[J-D- (d) P ------------.~

o -_..-E]--
It may be simplified in this way:

(P V (Q A ,.....,Q)) A (P V Q)
«P V Q) A (P V ,.....,Q)) A (P V Q)
«P V ,.....,Q) A (P V Q)) A (P V Q)
(P V ,.....,Q) A «P V Q) A (P V Q))
Dist
Comm
Assoc
Q-D -- O~
R-

(P V ,.....,Q) A (P V Q)
P V ("-'Q A Q)
Idem
Dist
(e)P G
""

DI
P
Q - -..- G II .----__lBIIIr_ G ----
The simplified circuit will thus be merely P -
R--ojOlIO-G
Exercises
1. Indicate the formula of each of the following circuits, simplify the formula,
and draw the simplified circuit.
(f) P

Q
oA
II
,
>0 [3- [J-
D
(a) P -->
...

Q
[0 ___ v I

.
II

(g)
s
P Q

(b)
FLO III"

v
-- ~-----'f
t
c]---G-
t
Q
'"
101
D 10'

-c]--G--8-G
66 Decision Procedures

(h) P R

L-------r-----'A - D- v
natural
deduction
Truth-table techniques provide us
with an effective procedure for
deciding whether a well-formed for-
mula of the sentential language is

EJ -- CJ ______
valid or not. In this chapter we wish
-1110'__ '-_---I
system sc to set down a system in which
proofs are constructed for theorems.
The theorems will all be valid
sentential formulas, and all valid
Answers sentential formulas will be theorems
1. (a) (P V ,-...,Q) A Q is equivalent to P A Q. of the system. Such a system, which
(b) ~P V (P A Q) is equivalent to ~P V Q. collects together all valid sentential
(e) ,,-,(~P A ~Q A ~R) is equivalent to P V Q V R. formulas, is called the sentential
(g) ~(P A ~(P A ~(P A Q))) is equivalent to ~(P A Q). calculus. Actually, the name 'sen-
(h) (~P A (Q A R)) V (P A (Q A R)) is equivalent to Q A R. tential calculus' is given to anyone
of various equivalent systems in
which theorems are tautologies and
all tautologies are theorems. The
system developed in this chapter
wiII be called system SO (after
sentential calculus). The system SO
is a natural deduction system since
it makes use of rules without any
axioms and since derivations roughly
correspond to how we "naturally"
construct them.
Why introduce a system that
has tautologies as theorems-that
constructs proofs for tautological
formulas-when we have truth-
table methods for determining
whether a sentential formula is or
is not a tautology? One reason for
doing this at this time is that we
wish to make additions to SO.
There are well-formed formulas
in logic other than sentential formul-
as, namely, predicate formulas.
And these also make up two
classes-valid and invalid formulas.
/.
67
 68 Natural Deduction System SO
69
However, there are no mechanical formulas, would go like this:
procedures like truth-tables that en-
able us to tell whether any pre-
dicate formula is or is not valid. To 1. If P ---+ Q and P, then Q.
establish the validity of such for- 2. If Q and if Q ---+ ,......,R /\ S, then ,......,R /\ S.
mulas we must use a system that is 3. If,.....,R /\ S, then ""R, so ,......,R must follow from the set.
constructed by making additions to
the sentential calculus. These ad- In figuring out that ,......,R follows from the set, what we do is to assume
ditions to system SC will be made in that each formula is true and we follow out the consequences of this
Chapter 6.
assumption. In doing this we see why ,......,R must also be true if we assume
the formulas in the set are true. In figuring out that ""R is a consequence
from the set we in effect show that an argument pattern is valid, namely
4.1 Rules of inference and proofs

Much of the system to establish formulas as tautologies, which will be P ---+ Q, Q ---+,......,R /\ S, P :. ,......,R
given later (Section 4.5), makes use of rules of inference and proofs using
rules of inference. What we wish to do now is to acquaint the reader where the premises are separated by commas before the therefore sign,
with these rules and the notion of a proof. The procedure we will follow and the conclusion follows this sign.
is, first, to informally discuss the rules of inference used in the forth- Whether or not one fully realizes it, rules of inference are employed
coming systems and proofs using these rules. Then, later, we will in figuring out that ,......,R follows from or is a tautological consequence of
incorporate the main body of this discussion into the formation of a the set of formulas. Returning to the numbered lines above, the rule
system that has tautologies as theorems. justifying the inference of Q given P ---+ Q and P is called modus ponens
Suppose the question is raised whether the formula ,......,R follows from (abbreviated MP) and may be formulated as follows:
the following set of formulas:
RUI,E OF Modus Ponens (MP):
Given A ---+ B and A, we may infer B.

On line 2 what warrants going from Q and Q ---+ ,......,R /\ S to "-'R /\ Sis
That is, suppose the question is raised whether, if this set of formulas is also the rule of inference, modus ponens. With the activity on line 3, a
true, ,......,R must be true. Supposing each of these formulas is true, must different rule of inference is supposed. It is the rule that allows the
,......,R be true~ Can we have this set true and yet have ,......,R false~ In the inference from ,......,R /\ S to ,......,R. This rule is often called the rule oj
last chapter two procedures were introduced by which one can determine simplification and is stated below:
whether ,......,R follows from the set of formulas above. If we are dealing
with sentential formulas, then B follows from A iff A ---+ B is a tautology. RULE OF SIMPLIFICATION (S):
B may be said to be a tautological consequence of A iff A ---+ B is a tau- Given A /\ B, we may infer A or we may infer B.
tology. Thus in asking if ,......,R follows from the above set of formulas we
are asking if ,......,R is a tautological consequence of the set of formulas. The capital letters A and B occur in the statement of these two rules.
So one could use one of the decision procedures introduced in the last 'fhese will again be taken as variables for well-formed formulas of the
chapter, say the reductio truth-table test, and calculate whether or not sentential language (see Section 2.6), as has been our practice from the
,......,R follows. beginning.
However, the steps we would naturally employ in figuring out Let us illustrate some simpler applications of these two rules of
whether ,......,R follows or is a tautological consequence of the above set of inference. Below, the last line of each sequence can be inferred from the
 Natural Deduction System se Rules of Inference and Proofs 71
70
The validation of this argument pattern will be written out as
previous lines by the indicated rule of inference:
folloWS, and what appears here is a proof:
(P V Q) /\ R P/\Q-+R
S p/\Q 1. P-+Q A
pvQ
R MP 2. Q -+,....,.,R /\ S A
w /\,.....,0 3. P A
,....,.,0 S ,....,.,P-+R V S 4. Q from lines 1 and 3 by MP
,....,.,P 5. ,-..,R /\ S from lines 4 and 2 by MP
(P /\ Q) /\ (Q V R) RvS MP 6. ,....,.,R from lines 5 by S
P/\Q S
This is an example of how proofs will be written. Each line is numbered
A rule of inference is something that allows us to go from some given and at the right we find a justification for each line. For the premises we
assumption to some other statement. A rule of inference can be valid or write A in the justification position. We justify lines that are premises
not valid. If a rule of inference is one that allows only true or acceptable by the rule of assumption, which we now state:
statements to follow from true or acceptable assumptions, it is a valid
rule. In other words, a valid rule can never lead us from a true assumption RULE OF ASSUMPTION (A):
to a false conclusion. In the matter at hand, a rule of inference is valid if An assumption may be introduced at any point in a proof.
and only if from assumed well-formed formulas of the sentential
language it allows the inference of formulas that are tautological ~he. deri:ed !ines in the ab~ve proof are the non-A lines and are given
consequences of the assumed formula. Both modus ponens and simplifi- JustIficatIOn m terms of theIr place of origin and rule of inference use d .
cation are such valid rules. Consider modus ponens. If A -+ B and A Generally before the numbered sequences of lines we will give the
are assumed to be true we have argument patter~ to be proved (as we do in the next proof). Finally we
note.that .each hne of the proof is either an assumption or comes from
A-+B prevIOus hnes by an application of one of the inference rules. Since each
TT of the rules is valid and is properly applied, the above sequence of lines
shows t~at ,....,.,R follows from the premise assumption, that ,....,.,R is a
in which case B can only be true. In turn, if modus ponens is applied, the tautologICal consequence of the premises. We thus have a proof that
conclusion must be a tautological consequence of the assumptions since ,-vR follows from the premises-in other words, we have established th
((A -+ B) /\ A) -+ B is a tautology. Since we are building the elements validity of the argument pattern above. e
for a system in which only tautological inferences will be allowed, we The rule of assumption no doubt seems to some readers to be
wish each of our rules to be valid. ~xcessively liberal. Can anything enter as an assumption ~ The answer
Having introduced the two rules of inference used in figuring out IS yes, for in logic t.he. concern is not with the nature of the assumptions
that ,....,.,R follows from the above set of formulas, we can noW more but wheth~r what IS mferred from assumptions logically follows or not.
formally write out the "figuring out" of ,....,.,R from the set of assumptions. In settmg out the rules for SC it will be convenient to dispense with
We will call what we do giving a proof. What we wish to do is to establish phrases such as 'given' and 'we may infer-'. Let us regard all such rule
the validity of the following argument pattern: talk as collapsed in a line, such as . The two rules so far considered
can thus be conveniently written:

MP A-+B S A /\ B A/\B
This is the pattern to be shown to be a valid one in the proof. It is made A ~' B
up of three premises separated by commas. The premises are separated -B-
from the conclusion ,....,.,R by the sign ~, called the turnstile sign. ~ can be
treated like: .. Av ... , An ~ B can be read "Av ... , An therefore B" or Thus the first will be read: Given A -+ B and A we may infer B.
"B follows from Av ... , An" or "There is a proof of B from Av ... , An. "
 72 Natural Deduction System SO
4.2 Four Additional Rules of Inference 73
Let us construct another proof for an argument pattern. What We
to introduce formulas using A into a proof. Using the convention adopted
wish to show is that, given as assumptions P -+ (Q -+ R), P -+ Q, and P,
at the end of the last section, this rule is:
we may validly conclude R. R is a tautological consequence of this set
of assumptions. The proof: RULE OF CONJUNCTION (OONJ):
Prove: P-+ (Q-+ R), P-+Q, P f- R
1. P-+(Q-+R) A
A,B
2. P-+Q A AAB
3. P A
4. Q -+ R 1,3 MP Its use is illustrated in the proof that follows:
5. Q 2,3 MP
6. R 4,5 MP Prove: P A Q f- Q A P
1. P A Q A
One final note of clarification. When a line is inferred from assumption 2. Q 1,8
lines by using one or more rules we will call this a deduction of the line in 3. P 1,8
4. QA P 2,3 Conj
question from the assumptions. The last proof thus establishes that
there is a deduction of R from P -+ (Q -+ R), P -+ Q, and P. The The proof that follows illustrates a special case.
characterizations given for the notion of a proof and the notion of a
deduction do not involve much precision. This is saved for later, when Pf-PAP
system SO is set down. loP A
2. PA P 1, Conj
Exercises
We will regard a rule that has metavariables separated by commas, as is
1. Construct a proof for each of the following argument patterns.
(a) P-+Q, PAR f-Q the case with Oonj, as applicable when the same line is an instance of
(b) P -+ Q, Q -+ R, P f- R both variables. This is the case with the proof above. Line 1 is taken as
(c) ,.....,P -+ Q A R, ,.....,P f- R an instance of A and B in the application of Oonj.
(d) (P-+Q)A (Q-+ R), P f- R The next rule to be introduced allows the introduction and removal
(Q -+ R), ,.....,P -+ Q, ,.....,p f- R
(e) ,.....,P -+
of negation signs. It is the rule of double negation and is set down
(f) P-+ (P-+Q), P f-Q
below:
2. Is the following principle true: If A -+ B is a tautology then 'Given A we may
infer B' is a valid rule. If it is true what are its consequences? RULE OF DOUBLE NEGATION (DN):
A ,-....;,....,A
Answers -A-
'"" ,.....,A '
1. (d) 1. (P-+Q) A (Q-+R) A
2. P-+Q 1,8 The rule of double negation allows us to go from any wff A to the formula
3. Q-+ R 1,8
with r-J ,....., prefixed, and vice versa. Thus DN justifies an inference of
4. P A
5. Q 2,4MP the second line below from the first, and vice versa.
6. R 3,5MP
2. Yes. It generates an infinite number of valid rules of inference.

The use of this rule is illustrated in the following proof:

4.2 Four additional rules of inference Prove: ,.,It""p -+ Q, P f- Q

1. '"" '"" P -+ Q A
The third rule of inference concerns the use of the conjunction. The rule 2. P A
says that given A and given B we may infer A A B. This rule allows us 3. f'"Oo..),.....,p 2,DN
4.Q 1,3 MP
74 Natural Deduction System SO Four Additional Rules of Inference 75

To use MP with ,-.., ,-..,p ~ Q and P we cannot take off the ,-.., '"'-' in the An argument pattern is valid if it has no instance that is made up of
antecedent of ,-.., ,-..,p ~ Q. To apply DN the I"-' ,-.., must have the scope true premises and a false conclusion. We can now construct proofs for
of the whole formula. For example we could use DN to remove the some argument patterns by using the rules just introduced and thus
"'" ,-.., from ,-.., ,-..,(P ~ Q). demonstrate that they are valid. In addition we can show that any in-
The next rule allows us to enter disjunctions into proofs and is valid argument pattern is invalid by supplying statements for its
components such that an argument with true premises and a false con-
RULE OF ADDITION (ADD): clusion results. To illustrate, consider these patterns:
A A
AvB' BvA P~Q,Qf-P
pvQf-P
From A this rule tells us we may derive A vB, and from A we may
derive B V A. A may be any wff of the propositional language and B Neither is valid. To demonstrate the invalidity of the first pattern we
may be any such formula-this being the case, of course, with all of the can make these substitutions:
rules. To illustrate Add's use, consider the next proof:
P: Rockefeller is a conservative (F)
Q: Rockefeller favors a growing economy (T)
P f- P A (P V Q)
l.P A
2. PVQ 2, Add obtaining a true premise and a false conclusion. The same statements
3. P A (P V Q) 1, 2 Conj substituted in the same way will demonstrate that the second pattern
is invalid.
The next rule again concerns the use of ~ and "-'. Given a con- If there is an instance of an argument pattern in which all the
ditional formula and the negation of its consequent, the rule permits premises are true but the conclusion is false, such a case is called a
us to infer the negation of the antecedent of the conditional. Here is the counterexample. Thus each of the above substitutions provides a counter-
rule: example for each of the above argument patterns. An argument
pattern is valid if it has no counterexamples. With respect to argument
RULE OF modus tollens (MT): patterns expressed in the sentential language, the quickest way to produce
A~B
,-..,B a counterexample is by using the reductio truth-table test. The pattern
.......,A
(P V Q) ~ R f- R V P
This is the other side of the coin to MP. Its use is illustrated below:
is invalid. Treating the turnstile like a conditional that is the main
Prove: ,-..,p ~ ""'Q, Q f- P connective, the reductio test would reveal that the premise is true and the
l. ,-..,P ~ "",Q A conclusion false when P is F, Q is F, and R is F. Thus we obtain a
2. Q A counterexample to this pattern by supplying false statements for P,
3. ,-..,.......,Q 2,DN Q, and R. We can, in fact, regard the reductio truth-table test as indi-
4. I"-' I"-'P 3,1 MT cating whether there is a possible counterexample. If we obtain a
5. P 4,DN
contradiction, this shows that there cannot be a counterexample.
Note that P cannot be inferred from 1 and 2 by MT. For MT or any of Exercises
the rules to apply, the same shape must obtain. With respect to MT we
1. Construct a proof for each of the following argument patterns.
must have a wff A connected by ~ to a wff B and have another line
(a) ,,-,p ~ '"'-'Q, '"'-' "-'Q f- P (e) I"-'P, ,,-,(Q A R) ~ P f- Q
made up of B with a before it. In the above proof where MT is
I"-'
(b) (P A Q) A R f- P A (Q A R) (f) P A Q f- P V Q
properly applied, I"-'P is an instance of A and """Q is an instance of B; (c) (P ~ Q) A (P ~ R), P f- Q A R (g) Q ~ (P ~ R), I"-'R, Q f- I"-' "-' "",p
thus "",Q is an instance of f"Ooo./B.
I"-' (d) P V Q ~ '"'-'R, P f- ,,-,R (h) Q, P ~ '"'-'(Q A Q) f- ,,-,p
76 Natural Deduction System SO 4.3 Assumption Discharging Rules 77
2. Show that the following patterns are invalid by supplying counter_ assumption in addition to the premise assumptions-and then attempt
examples. to derive B from A and the set of assumptions that make up the
(a) P -+ Q, r-IP I-,......,Q (c) P 1\ Q -+ R I- P -+ R
(b) P -+ Q, R -+ Q I- P -+ R (d) P-+QI-Q-+P
premises. If we succeed in deriving B from A and 6, then our new rule
declares a deduction of A -+ B from 6.
Answers In passing from 6, A I- B to 6 I- A -+ B we will say that A has been
discharged. Thus having assumed A, to obtain a conclusion of the form
1. (d) 1. P V Q -+ r-IR A A -+ B, when we enter A -+ B on the last line, A is said to be discharged.
2. P A The assumptions, if any, found in 6 are not affected; they remain.
3. PVQ 2, Add
4. r-IR 1,3 MP
Let us illustrate a simple application of this rule. Suppose we wish
(g) 1. Q -+ (P -+ R) A to establish the following argument pattern:
2. r-IR A
3. Q A P-+QI-PI\R-+Q
4. P-+ R 1,3 MP
5. r-.;p 2,4MT
6. r-.J'-""!"'-.Ip 5,DN Since the conclusion is a conditional, we may use the new rule. To use it
we not only take P -+ Q as an assumption, but we take the antecedent
2. (d) P: Reagan is an Arizonan (F), Q: Reagan is an American (T).
of the conclusion, P 1\ R, as an additional assumption. Then we attempt
to derive the consequent of the conclusion, Q. We thus proceed as
follows:
4.3 Assumption discharging rules
1. P-+Q A
In this section three rules will be introduced that involve the removal of 2. P 1\ R A
3. P 2,S
assumption lines in a proof. The rules in the previous section do not
4. Q 1,3 MP
affect the number of assumptions in a proof. The rules in this section 5. P 1\ R-+Q 2--4 by new ru}e
always result in the reduction of the number of assumptions in a proof.
To set out these rules we will find it useful to introduce the symbol
In using the new rule to justify line 5, we discharge the assumption on
6. Let 6 be a set of zero or more assumptions. Later we will encounter
line 2 and remove it from the assumptions of the proof. We wish to
situations in which 6 is a set of zero assumptions, an empty set.
indicate this in some way. We will do this by writing the last proof as
The first assumption discharging rule incorporates this valid
follows:
inference: If from a set of assumptions 6 and A we can derive B, then
A -+ B is derivable from 6. This is a valid rule. 6 1\ A -+ Bis
equivalent to 6 -+ (A -+ B). Therefore the formula A -+ B can be
. 1. P-+Q
PI\R
A
A
established as following from 6 by showing that B is deductible from
the assumptions 6 and A.
By using the turnstile sign this rule may be conveniently expressed
U 3. P
4. Q
5. P 1\ R-+Q
2, S
1,3 MP
2-4 by new rule
as follows:
The line from 2 to between 4 and 5 indicates that the assumption on line
Given 6 and A I- B, we may infer 6 I- A -+ B. 2 has been discharged on reaching line 5. Only the assumption on line 1
remains; thus 5 is derived from 1; the above argument pattern has been
This rule will be used in constructing proofs for argument patterns that established.
have conclusions in conditional form. What we may now do when the To set out this rule, which will be called rule of conditional proof, in
conclusion
-,. has a conditional form A -+ B is to assume A-enter A as an the form of the previous rules we may do the following:
 78 Natural Deduction System SO 4.3 Assumption Discharging Rules 79
RULE OF CONDITIONAL PROOF (Rep):
p -+ Q II R f- Q below:
D, Af-B
A-+B 1. P-+QIIR A
2. P A
In this rule A f- B indicates that B has been derived from A. 6,
A f- B thus reads that B has been derived from 6 and A. The angled [i Q
3. Q II R
4.
1,2MP
3,S
line indicates that A has been discharged. A -+ B is thus said to be 5. P-+Q RCP
derived from the set of assumptions, if any, that make up 6. 6. Q 2,5 MP (erroneous)
Rep can be used once, twice, three times, and so on in a proof, if
Line 6 is erroneous since it is obtained in part by using a rule on a line
necessary. For example, Rep is used twice in the following proof:
within the scope of a discharged assumption.
Prove: P II Q -+ R f- P -+ (Q -+ R) The next assumption discharging rule is relatively the most difficult
1. PIIQ-+R A to grasp. The rule states that given A V B if we assume A and derive 0

~
2' P A and if we assume B and derive 0, then 0 follows from A V B. If in
r::liR
3. Q
P II Q
A
2,3 Conj
1,4 MP
deriving 0 other assumptions 6 are employed, then 0 follows from
A V Band 6. This rule will be called rule of disjunction and may be
6. Q -+ R RCP indicated in this way:
7. P -+ (Q -+ R) RCP
RULE OF DISJUNCTION (Drs):

With assumption lines 2 and 3 discharged, 7 follows from 1 alone, thus 6,Av B
proving the validity of the argument pattern. Note that this argument 6lA f-C
6,Bf-C
pattern corresponds to the Rep rule. Argument patterns corresponding C
to each rule, with the exception of the rule of assumption, can easily be
constructed. The rule reads that given 6 and a disjunction A V B, if we assume A
Our drawing the lines to the left of the proof serves several functions. and obtain 0 (from A alone or in conjunction with 6) and if having
First, and foremost, it shows when an assumption has been discharged assumed the other disjunct B we obtain 0 (from B alone or in con-
and where it has been discharged. Second, it shows the scope of an junction with 6), then 0 follows from A V B and any other assumptions
assumption. In the last proof the scope of the line 3 assumption is 3 to used in the derivation. In this rule, as with all assumption discharging
5, while the scope of the line 2 assumption is 2 to 6. The scope of the rules, the proviso against using a line in the scope of an assumption after
assumption on line 3 is found within the scope of the assumption on line 2. the assumption has been discharged must be observed.
This is permissible. Scopes of different assumptions may also follow each The use of this rather complicated rule is illustrated in two proofs
other with no overlap. But what is not allowed is any overlap of the below:
scopes of two or more assumptions. Any such overlap will show up in Prove: P V Q f- Q V P
crossing lines. To avoid crossing lines we will lay down this simple 1. P V Q A
proviso for when Rep is used in a proof: One cannot use a line in the 2. P A
scope of an Rep assumption after the assumption has been discharged, 3. Q V P 2, Add
unless the line can be gotten from other assumptions. This proviso will
be followed and will apply to all the rules of inference. That is, in using
U 4. Q
5. Qv P
6. Qv P
A
4, Add
Dis
any of the rules of inference we must not use a previous line that is in
Prove: P V P f- P
the scope of an assumption already discharged unless the line can be
1. P V P A
gotten from other assumptions. Observing this restriction on the use of
rules of inference when assumption discharging rules are employed will r~' P A
~P A
prevent erroneous "proofs" such as the "proof" of the invalid pattern 4. P Dis
80 Natural Deduction System S() 4.3 Assumption Discharging Rules 81

This last example reveals a limiting case of the assumption discharging Since we count any line as following from itself, the shortest possible
rules. Having assumed A we will count A as itself following from A. proof can be constructed for P ~ P, namely
Also in both uses of Dis, 6. is empty. That is, there are no assumptions 1.P A
other than the disjunction A V B. Thus 0, Q V P in the first case and P
in the second, is derived from the disjunction alone in each proof. We illustrate a second use of RAA in proving the validity of an
The last assumption discharging rule is the rule of reductio ad argument pattern.

·,. . ",. . "

absurdum. The method of proof by reductio ad absurdum is already Prove: P ---+,....."p ~,....."p

familiar to us. It was used as a short-cut truth-table test to determine 1. P ---+ ,....."P A
whether or not a formula is tautologous (Section 3.4). This is the P A
3. P 2,DN
familiar pattern of argument in geometry, mathematics, and philosophy.
For example in deriving his theorems, Euclid often begins by assuming
the opposite of what he wants to prove. If that assumption leads to a
contradiction, or "reduces to an absurdity," then that assumption must
U 4. ,....."p
5. P /\,....."p
6. ,....." ,......,P
r--J

7. ,....."P
1,3 MP
3, 4 Conj
RAA
6, DN
be false, and so its negation-the theorem to be proved-must be true.
This proof can be shortened by assuming P rather than "" ",-,P. To
In using l'eductio for proofs of argument patterns we show that from
assume either,....." ,....."p or P is to assume the contradiction of the con-
6. and the denial of the conclusion, ",,0, a contradiction follows. A
clusion.
contradiction must be false; thus by MT it must be false that 6. and ",0
are true. If it cannot be the case that 6. and ",,0 then 6. ~ must be
valid, since we cannot produce a counterexample for the pattern
° To end this section, the ten rules for constructing proofs are sum-
marized below.
6. ~o. 1. ASSUMPTION (A):
To state this last assumption discharging rule we must first define a An assumption may be introduced at any point in a derivation.
contradiction. By a contradiction we will understand a conjunction in 2. Mod7ls ponens (MP):
which the second conjunct is the negation of the first conjunct. Thus A---+B
P /\ f"'.JP, f"'.Jp /\ ,-...., ,"",P, (P ---+ Q) /\ ,.....,,(P ---+ Q) are all contradictions. A
Now the last rule states that given a derivation of a contradiction B
B /\ ,....."B from an assumption A and any other assumptions 6., we may 3. Modus tollens (MT):
derive ,....."A from 6.. Here is the statement of the rule: A---+B
,....."B
RULE OF reductio ad absurdum (RAA): ,....."A
6.,rA~B/\,....."B
,......,A 4. SIMPLIFICATION (S) :
A/\B A/\B
In using RAA for constructing proofs for argument patterns the pro-
cedure is to enter the denial of the conclusion as an assumption in addition
---:::t B
to the premise assumption, 6., and to derive a contradiction. Having 5, DOUBLE NEGATION (DN):
derived the contradiction we can enter the negation of the additional A ,....." ,....."A
assumption, discharging the additional assumption. The use of RAA is ,....." ,....."A ' A
illustrated in this proof of the simplest possible argument pattern: 6. OONJUNCTION (CONJ):

Prove: P ~ P A,B
l.P A A/\B
I~' ,....."P A 7. ADDITION (ADD):
L..! P /\ ,,-,P 1,2 Conj
A A
4. ,.....",......,p RAA
5. P 4,DN AvB' BvA
82 Natural Deduction System SC Useful Strategies 83

8. DISJUNCTION (DIS):
6, A vB
6'bA f-O 4.4 Useful strategies
6, Bf-O
o A computer could be programmed that would produce a proof for all
valid argument patterns expressed in the sentential language, and would,
9. CONDITIONAL PROOF (RCP):
after a time, give up, if the pattern is invalid. But we do not have access
6,CA f-B
A-+B to such a machine, and even if we did, we ought to ignore it at this
stage. We wish to acquire the ability to construct proofs on our own.
10. Reductio ad absurdum (RAA): This is a creative act in most cases and thus requires insight, diligence,
6,c A f-BA,-.,B and, often, luck. However, there are certain strategies that will keep
"",A
one from blind alleys and that are generally successful. Here are some
of these strategies.
Exercises STRATEGY ONE: If the desired conclusion is of the form A ---+ B, add
1. Construct a proof for each of the following argument patterns. A as a further assumption and deduce B. Then use RCP to obtain
(a) P V (P A Q) f- P (f) P -+ (P -+ Q) f- P -+ Q A ---+ B. In turn, if the consequent of a conditional conclusion is itself a
(b) P -+ Q, Q -+ R f- P -+ R (g) P -+ (Q -+ R) f- P A Q ---+ R
conditional, then assume the antecedent of that conditional, and so on.
(c) P -+ Q f-,-.,Q ---+ ",-,p (h) P V P f- ,-.,(,-.,p A ,-.,P)
(d) P f- ,-.,(Q A ,-.,Q) (i) ,-.,P -+ P f- P '1'0 illustrate:
(e) (,-.,p -+ Q) A r-;Q f- P (j) P ---+ Q A R f- (P ---+ Q) A (P ---+ R)
P ---+ (Q ---+ R) f- Q ---+ (P ---+ R)
2. Why does the proviso: "One cannot use a line in the scope of a discharged 1. P ---+ (Q -+ R) A
assumption unless the line can be gotten from other assumptions" prevent A

~
'Q
crossing lines? A
r::liR
3. P
Q ---+ R 1,3 MP
Answers
p
1. (a) 1. P V (P A Q) A
A
6. P ---+ R
7. Q ---+ (P ---+ R)
2,4MP
ROP
ROP

a
3. P AQ
4. P
5. P
(i) 1. ,-.,P -+ P
A
3,8
Dis
A
We should note that if the conclusion had the form (A ---+ B) ---+ C, to
use this strategy one must assume the antecedent that is A ---+ B, not A.
STRATEGY TWO: If one of the premises or assumptions is of the form
a~p A A V B, then try Dis unless there is some other easier procedure. To
3. P 1,2 MP
illustrate:
4. P A,-.,P 2, 3 Oonj
5. ,-.,,-.,P RAA
6. P 5,DN (P ---+ R) A (Q ---+ R) f- P V Q ---+ R
(j) 1. P-+Q A R A 1. (P ---+ R) A (Q ---+ R) A
A 2. P V Q A
~p
3. Q A R 1,2MP 'P A
4. Q 3,8 4. P ---+ R 1,8
5. P ---+ Q RCP 5. R 3,4MP
A 6. Q A
~p
7. Q A R 1,6 MP [ 7.Q---+R 1,8
8. R 7,8 I~R 6,7 MP
9. P-+ R ROP UR Dis
10. (P ---+ Q) A (P ---+ R) 5,900nj 10. P V Q ---+ R RCP
"
 84 Natural Deduction System SO Useful Strategies 85

Line 2 was introduced using strategy one. Since line 2 is a disjunction, same procedure with Q. The proof is thus:
we engage in Dis strategy assuming P and Q and obtain R from each.
STRATEGY THREE: When the assumptions seem inadequate to get A
the desired conclusion, then use RAA. RAA will always work. To prove A
,",-,(P A '"'-'Q) f- P ~ Q makes an interesting proof, for we can assume P A
for an RCP proof and also engage in RAA by assuming "-'Q. To 3,S
illustrate: 2,4 Conj
RAA
A
A A
A 8,S
A 7,9 Conj
2,3 Conj RAA
1,4 Conj Dis
RAA
6,DN
RCP With line 1 being the only A line left we have derived 12 from 1, thus
establishing the above argument pattern.
Rather than write out lines 1 and 4 on line 5, we have indicated them by
line number.
STRATEGY FOUR: Conj, Add, RCP, and DN (one way) are rules for Exercises
introducing connectives, whereas MP, MT, S, DN (the other way), and I. Construct proofs for the following patterns.
Dis are rules for eliminating connectives. Generally speaking, in a proof (a) P --+ Q f- ,",-,(P A ,,-,Q)
use at first as many of the elimination rules as possible to get as many (b) P--+Qf-(Q--+R)~(P~R)
component variables separated on lines as possible. Then, if necessary, (c) 8 --+ (P ~ Q) f- (8 ~ P) ~ (8 ~ Q)
(d) P A Q ~ R f- P --+ (Q ~ R)
use the introduction rule corresponding to the main connective in the
(e) "",p f- P --+ Q
conclusion. Thus one uses elimination rules on the premises, intro- (f) P f- Q --+ P
duction rules on the conclusion, and one works towards the middle. Such (g) P V Q f- (P ~ Q) ~ Q
considerations often lead to adopting one of the strategies above and, (h) Q f- f',JR --+ f',J(Q ~ R)
with good fortune, can make clear a successful proof procedure. To (i) ",(P V Q) f- f',JP A f',JQ
illustrate, suppose we are called upon to validate the following argument (j) P V Q, '"'-'P f- Q
(k) "",p V '"'-'Q f- ,",-,(P A Q)
pattern: (1) """'(f',JP A f',JQ) f- P V Q
(m) P V (Q A R) f- (P V Q) A (P V R)
(n) P A (Q V R) f- (P A Q) V (P A R)

Working from the premise, we see we can use the elimination rule Dis.
Answers
The conclusion is not in conditional form and we are unable to get it
from P alone and then from Q alone in a Dis strategy; but we could use
RAA along with a Dis strategy. In using RAA within the Dis strategy
we need to obtain the conclusion twice, once from P and once from Q.
To obtain it from P using RAA we must assume its negation '"'-'P A '"'-'Q,
.r::
(Note: There are other ways to prove these patterns.)
I. (b) 1. P~Q
Q~R
3. P
A
A
A

and in obtaining it from Q we must again assume its negation. Having

assumed P and '"'-'P A "-'Q, we can use an elimination rule, S, on
"",P A '"'-'Q to obtain ,",-,P, which contradicts P. And we may follow the
hlLi Q
R
6. P~R
7. (Q~ R) ~ (P--+ R)
1,3 lVIP
2,4lVIP
RCP
RCP
,.
86 Natural Deduction System SO 87

(e) 1. ,......,P A follow the practice of assuming A, deriving B, and thus by Rep ob-

~
2' P A taining A -+ B. In using ROP in this way,/'., is empty. If B is derived
3. ,......,Q A
from A without any other assumptions by using valid sentential rules,
[i; P A t'o.'P 1,2 Conj
then B is a tautological consequence of A alone, and thus A -+ B is a
5. ,......,,......,Q RAA
6. Q 5,DN tautology-what we desire.
7. P-+Q RCP Earlier it was announced that the system will be called SO and that
(i) 1. ,......,(P V Q) A it is a system for the sentential calculus. The sentential calculus, as
2. P A indicated at that time, can be identified with an infinite number of
3. PVQ 2, Add
[}4. (3) A (1) 1,3 Conj
systems that have the feature of having theorems which are only

. 5. f"o-/P
Q
RAA
A
tautologies and having all tautologies as theorems.
The system for the sentential calculus, system SO, is made up of
[i 7. P vQ
8. (7) A (1)
9. "",Q
6, Add
1,7 Conj
RAA
four elements.
First, SO contains rules for a well-formed formula in SO. These
rules are the same as those for a well-formed formula in the sentential
10. ,...."p V ""Q 5,9 Conj
(j) 1. P V Q A language set down in Section 2.6.
2. t".IP A Second, SO contains ten basic or primitive rules. The ten basic rules
A for derivation in SO are those summarized at the end of Section 4.3.

~
'P
4. ,......,Q A Third, in SO we wish to have theorems that include the biconditional
[i PA""""P 2,3 Conj sign, +-0. As the inference rules are set up, though, ,......" A, v, and -+ can be
6. Q RM introduced in a proof, +-0 is not introducible. So we will provide a
7. Q A
8. Q Dis definition for this connective in terms of connectives that can be intro-
(k) 1. ",-,P V ,....""Q A duced in proofs. \iVe thus introduce the following definition for the
2. PAQ A biconditional:
3. P 2, S
. 4. ,..,.,,......,P 3,DN

[15.
6.
7.
,......,Q
Q
Q A,......,Q
1, 4 (by following proof of (j))
2, S
5,6 Conj
A +-0 B =df (A -+ B) A (B -+ A)

The effect of such a definition is to permit us to replace instances of what

is on the left side of the =df sign with instances of what is on the right
8. ,......,(P A Q) RAA
side, and vice versa. When this definition is applied we shall cite 'df'.
Some uses of this definition are illustrated below:
4.5 System SC
(P-+Q) A (Q-+ P)
In the next three sections we will ignore argument patterns and turn P+-0Q df
to our primary interest, which is a system that has only tautologies and
P +-0,......,R
all tautologies as theorems. Since a tautology is not an argument
(P -+ ,......,R) A (,......,R -+ P) df
pattern, proofs will not begin by taking premises as assumptions. Rather
proofs will begin by using either ROP or RAA. That is, the assumptions (P A Q -+ P) A (P -+ P A Q)
will not be premise assumptions but will be the assumptions obtained PAQ+-0P df
from using ROP or RAA that are later discharged.
As we will see, most of the theorems of the system will be con- Ji'inally, we need an effective procedure for determining when we have
ditional formulas. With a formula of the form A -+ B we will generally a proof and a theorem in SO. vVe need a defillition for a proof and a
 Natural Deduction System SO Some Theorems of SO 89
88
theorem in SC such that a computer could be programmed to check Answers
whether what is written down is a proof or a theorem. A proof in SC will
be such that every step can be checked to see if it is justified, with the 1. An infinite number. Why?
checking relying only on the forms of the steps, not on the interpretation 2. 1 is not an A line, 2 does not follow from 1 and is not an A line, and 4 is not
of the symbols. the last line in a sequence in which there are no undischarged assumptions.
To this end we first define a deduction in SC. A decl~lction is a finite
sequence of formulas such that for each formula at least one of the
following holds:
4.6 Some theorems of SC

1. It is an assumption line. Only two rules, RCP and RAA, allow us to introduce assumptions so that
2. It is a line inferred from previous lines that are not in the scopc proofs can get started. How this is done is abundantly illustrated below.
of assumptions already discharged using one of the ten basic
The collection of theorems is chosen for special purposes, namely they
inference rules.
are the major tautologies that should be studied until they become self-
evident. Where a theorem is listed without proof it is expected that the
A proof is a deduction from the empty set. That is, when in a deduction reader will demonstrate it as an exercise (if its corresponding argument
all the assumptions have been discharged, we have a proof. A theorem is pattern has not been proven in previous exercises). The reader is urged
the last line of a proof. to work carefully through each of these proofs, making sure that he
Theorems in SC will only be tautologies. In the next two sections a sees what is being done at each step and why it is a reasonable move to
number of theorems will be derived and two derived rules will be intro- make. Some discussion will follow some of the proofs, indicating
duced. Many of the theorems will be conditionals corresponding to abbreviations and reminding the reader of aspects of the strategy.
previously validated argument patterns, and the strategies followed in
validating the argument patterns are and should be used with the Tl I- P-+P Law of identity
theorems. We will continue to number the lines of a proof and to indicate l.P A
2. Tl RCP
origin and rules on the right of each line, though such procedures are not
part of the definition of a proof in SC. But since this discussion is
intended for human beings rather than machines, we shall take this Though this proof is quite simple, it illustrates a basic strategy. In
proving a theorem (abbreviated T) ofthe form A -+ B, the form of most
liberty.
of our theorems, the antecedent is taken as an assumption and the
consequent is derived. Having done this, RCP gives us the theorem.
'When the theorem to be proven has the form A -+ B, as Tl has, we assume
Exercises
A, derive B, and use RCP to obtain the theorem. If the theorem to be
proven has the form A -+ (B -+ 0) we assume A and B and derive 0,
1. How many theorems are there in SC? and so on, just as we did with argument patterns. We note that in the
above proof all assumptions have been discharged when we come to
2. 'Vhy are lines 1, 2, and 4 in this sequence not sequences in a proof? line 2; thus, given the statement of a proof in SC, 2 is a theorem. It
1. "-' P A "-'Q should again be pointed out that we will regard any line as derivable
2. ,,-,(P A Q) from itself. Thus the appropriateness of applying RCP on line 2. The
3. P A
sign I- is placed between P -+ P and T1. This indicates that P -+ P is a
4. "-'Q
theorem or is provable in SC. From this point on we will regard the
theorem numbers, 'Tl' in this case, as doing the job of 'Tn 1-' when they
3. Provide an inference rule that will allow the introduction of +-t into proofs.
appear before a formula.
Would this alone be adequate for all tautologies using +-t?
90 Natural Deduction System SO Derived Rules 91

T2 P ---+ P vQ Addition Tl7 Pv(QvR)---+(PvQ)vR

I~' P A 1. P V (Q V R) A
L2 PVQ 1, Add 2. P A
3. T2 ROP 3. PVQ 2, Add
4. (P V Q) V R 3, Add
T2 corresponds to the rule of addition. If SO is to have all tautologies 5. Q V R A
A

~
as theorems, then a theorem corresponding to each rule, excluding A, 'Q
7. P VQ 6, Add
must be provable. 8. (P V Q) V R 7, Add
9. R A
Excluded middle 10. (P V Q) V R 9, Add

·
T3 P V ""p
1. ",,(P V ""P) A 11. (PV Q) V R Dis
A 12. (P V Q) V R Dis

U
P
3. P V ""P 2, Add 13. T17 ROP
4. (3) A (1) 1,300nj
In this proof we first assume the antecedent of the theorem, line 1.
5. ""p RAA
6. P V ""P 5, Add '['hen from line 1 we take each disjunct and try to derive (P V Q) V R,
7. (6) A (1) 6, 1 Oonj the consequent of T17, by Dis. We can easily get this from the first
8. r....;r....;(P V r....;P) RAA disjunct P. But to get it from Q V R, line 5, we must again use Dis
9. P V ""p 8,DN assuming Q, line 6, and R, line 9.
T18 (P V Q) V R ---+ P V (Q V R)
In this interesting proof we cannot use ROP since the theorem is not a T19 P A (Q A R) ---+ Q A (P A R)
conditional. T20 Q A (P A R) ---+ P A (Q A R)
T21 (P V Q) A I'.JP ---+ Q Disjunctive syllogism (DS)
'['21 may be proven by following the RAA strategy found in the answer
for Exercise 4.4.1 (j). As indicated earlier, any strategies used to prove an
A argument pattern that corresponds to a theorem can be adopted to a
A proof for the theorem. It will prove useful to abbreviate the name of
1,2 MT T21 as DS and indicate the theorem by name as we will have occasion to
ROP do with other theorems.
ROP
Exercises
Since T6 is an instance of A ---+ (B ---+ 0), we have taken A and B as 1. Oonstruct a proof for each of the underived theorems. It is permissible to
assumptions in our attempt to obtain O. omit proofs for those theorems whose corresponding argument patterns have
been proven in previous sections or in the exercises.
T7 (""Q ---+ ""P) ---+ (P ---+ Q)
T8 P V Q ---+ Q V P 4.7 Two derived rules
T9 Q V P ---+ P V Q
TlO P A Q ---+ Q A P To simplify the construction of proofs we will introduce two derived
TIl Q A P ---+ P A Q 1'ules. A derived rule allows us to do no more than what we can do with
T12 P V P ---+ P the ten primitive rules. The usefulness of derived rules is that they allow
T13 (P ---+ (Q ---+ R» ---+ (P A Q ---+ R) Importation us to do a proof in fewer steps, or they allow easier proofs. Whenever a
T14 (P A Q ---+ R) ---+ (P ---+ (Q ---+ R» Exportation
Transitivity of line in a proof is introduced by a derived rule, this will mean that if we
T15 (P ---+ Q) ---+ ((Q ---+ R) ---+ (P ---+ R»
T16 (P ---+ Q) A (Q ---+ R) ---+ (P ---+ R) implication had chosen, we could delete the derived rule and use some combination
}
(Trans) of the ten primitive rules.
"'-I.
 l'latUral uelluctlOn System Derived Rules 93

The first derived rule may be stated as follows: this proof T2I-disjunctive syllogism-is used to separate Q from
p VQ given ,""P. No substitution is needed in using T21. If we had to
RULE OF THEOREM INTRODUCTION (TI): separate P --+ R from P V (P --+ R) given ,""P, then we would need to
A theorem or substitution instance of a theorem may be entered on make a substitution in T21. Again, since TI is a derived rule the proof
any line of a proof. T24 can be done using only the primitive rules. One way is to inject
into the above proof the proof for T2I, obtaining line 4. Another way is
It will be recalled from the previous chapter that one obtains a
illustrated in the proof for the argument pattern P V Q f- "'-'( '""P A ",-,Q)
substitution instance of a formula when one uniformly substitutes a
well-formed formula for a single sentential variable. The use of TI is found in Section 4.4.
illustrated in the next proof.
T25 ",-,(,,-,P V Q) --+ P A "-'Q
T22 P ---+ (Q ---+ P) Law of Clavius
'P A To shorten proofs, we will collapse introducing a theorem (with or

O
2. Q A
without substitution) and MP into a single step. Thus we may make this
3. P ---+ P TI (TI)
4. P 1,3, MP mOve, for example:
5. T22 RCP (2)
1. "'-' "'-'P --+ P
No rule so far allows us just to repeat line 1, so TI along with TI plus MP 2. P--+Q
is used at this point. The use of TI in proving T22 can be abandoned 3. "'-' "-'P --+ Q 1,2 Trans
if we like. This is a necessary feature of a derived rule: it is dispensable:
In effect TI (T 1) at the justification place on line 3 means 'at this point rrrans here collapses several steps, which may be explicitly unpacked as
insert the proof for TI'. Thus we have not a proof, but a proof outline. follows:
A proof of T22, one using only the }}asic rules, would be:
1. "'-' "-'P --+ P

~
. P A 2. P--+Q
2. Q A 3. (P --+ Q) --+ ((Q --+ R) --+ (P ---+ R)) TI(TI5)
cl P A 4. (1) --+ ((2) --+ (5)) "'-' "-'PIP, P/Q, QIR in 3
4. P ---+ P RCP 5. "'-' "-'P --+ Q 1,2,4 MP (2)
5. P I,4MP
6. T22 RCP (2) T26 (P ---+ Q) --+ ,,-,p V Q
1. P--+Q A
A
As can be seen, the earlier proof of TI has merely been injected at the
point where TI(TI) appears.
~3. ~(~PvQ)
P A "-'Q 2, T25
4. P 3, S
T23 ",-,p ---+ (P ---+ Q) Law of Duns Scotus 5. Q I,4MP
T24 P V Q ---+ ",-,(",-,P II '"'-'Q) 6. "'-'Q 3, S
1. PVQ A 7. Q A "'-'Q 5,6 Conj
2. r--;p II "'-'Q A 8. "-' ,,-,(,,-,P V Q) RAA
3. '""-'P 2, S 9. ",-,p V Q 8,DN
4. (P V Q) II ",-,p ---+ Q TI (DS) 10. T26 RCP
5. (P V Q) A ,,-,P 1, 3 Conj T27 (,,-,P ---+ Q) --+ P V Q
6. Q 4,5 MP
7. "-'Q 2, S The definition for ~ can be used to obtain biconditional formulas
8. Q II "-'Q 6, 7 Conj
9. "-' ("-' P A ,,-,Q) RAA from previously proven theorems. The strategy when one has a theorem
10. T24 RCP of the form A ~ B is to construct a proof for A ---+ B and a proof for

"""/-,
94 Natural Deduction System SC 95

B -+ A and then use df to obtain A +-t B. This strategy is illustrated in introduction of the JYI +-t N line. Thus we may move directly from 2 to 4
the proof of the next theorem. with the justification being, simply, T28. And we may move from 4 to 6
with the justification being, simply, Idem. Placing 'Idem' at the
T28 P +-t "'-' ",-,P Double negation
'ustification spot in such a move leaves it ambiguous whether TI and
1. P -+ "'-' ",-,p T4
2. "'-' ",-,p -+ P T5 ~IP or R is being used, but this is no matter of concern.
3. (1) 1\ (2) 1, 2, Conj
T38 P V (Q 1\ R) -+ (P V Q) 1\ (P V R)
4. T28 3, df
T29 P V Q +-t Q V P} T39 P V Q+-t r_+_"P 1\ ",-,Q)}
Commutation T40 "'-' (P V Q) +-t "'-' P 1\ "'-'Q
T30 P 1\ Q +-t Q 1\ P de Morgan's Theorems
T41 ,,-,(P 1\ Q) +-t ",-,P V ,......,Q
T31 P V P+-t P} Idempotent P 1\ Q +-t r_+'_' P V ",-,Q)
T32 P 1\ P+-t P T42
T43 P -+ (Q -+ R) +-t P 1\ Q -+ R Export-Import (E-I)
T33 (P V Q) V R +-t P V (Q V R)}
Association T44 "'-'( P 1\ ",-,P) La w of contradiction
T34 (P 1\ Q) 1\ R+-t P 1\ (Q 1\ R)
T35 (P +-t Q) +-t (P -+ Q) 1\ (Q -+ P) Equivalence T45 P 1\ (Q V R) }
T36 P -+ Q+-t~P V Q Implication +-t (P 1\ Q) V (P 1\ R) Distribution
T46 P V (Q 1\ R)
+-t (P V
Q) 1\ (P V R)
The second derived rule will make use of biconditional theorems.
T47 (P V Q) (P -+ R)
1\
The rule will permit the replacement of A with B (and vice versa) 1\ (Q -+ R) -+ R Proof by cases
whether A and B are alone or in a formula if A +-t B is a theorem. To T4S P -+ Q +-t "'-'Q -+,.......P Contraposition
state the rule, let M and N be well-formed formulas. Let AM be a T49 (P-+Q) 1\ (R-+S)
formula such that M is A or part of A. Let AN be the result of replacing 1\ ("'-'Q V ",-,S) -+,......,P v,.......R Destructive dilemma (DD)
one or more, but not necessarily all, occurrences of M in AM by occur- T50 (P-+Q) 1\ (R-+S)
1\ (P V R) -+Q V S Constructive dilemma (CD)
rences of N, then
T51 P V (P 1\ Q) +-t P }
RULE OF REPLACEMENT(R):
T52 P 1\ (P V Q)+-tP
T53 (P -+ Q) 1\ (P -+ R) Absorption
If f- M +-t Nand f- AM, then f- AN.
+-tP-+QI\R
In using this rule we must have previously proven biconditional T54 ("",P -+ R 1\ ",-,R) -+ P } . .
T55 (P 1\ "'-'Q -+ R 1\ ",-,R) -+ (P -+ Q) Proof by contradIctIOn
theorems or substitution instances of them so as to obtain M +-t N. Let
us illustrate the use of this rule.

·
Exercises
T37 (",-,P -+ P) -+ P Conseq1lentia M imbilis 1. Construct a proof for each of the underived theorems.
~P-+P A 2. Since any line is taken as derivable from itself, shorten the proof of T22.
2. ~,-.."PV P 1, T26
3. "-' r-..Jp +-t P T28 3. The ten primitive rules are not all independent. We can dispense with some

U 4. Pv P
5. P V P+-t P
6. P
7. T37
3,2, R
Idem
4,5, R
RCP
and still obtain the same set of theorems. If one can, say, derive a theorem
corresponding to one of the rules not using this rule, then this theorem plus
theorem introduction, substitution, and MP allows one to do all you can do
with the rule. Thus the rule is a dependent rule relative to the other rules.
For some of the rules derive a theorem corresponding to the rule not using
Line 3 is the double negation theorem. In the first application of R, the rule or derived rules.
M +-t N is line 3. AM is line 2 and AN is line 4. In the second application 4. Give a proof using only the ten primitive rules for one of the proofs in which
of Equiv, M +-t N is line 5. AM is line 4 (here A is M) and AN is line 6 you used derived rules.
(here A is N). A much shorter proof of T37 can be constructed using
5. ]<"'or cach argument derive the conclusion from the premises, using any of
RAA (see exercise 4.3.1 (i)), but this longer proof usefully illustrates two the rules, listed theorems, and derived rules of SC. See answer to (g) below
uses of R. In using R we may shorten the proofs further by omitting the for procedure.
96 Natural Deduction System SC 4.8 Soundness, Consistency, and Completeness of SC 97
(a) If there is no test, then the students will be happy only if the professor
stays in bed. The professor is in class every day. If the students are
happy, then there is no test. Therefore the students will not be happy. 4.8 Soundness, consistency, and completeness of SC
(b) Either the students will be happy and there will be no exam or the
students will be happy and class will be dismissed. If there is no exam,
However confident we may be with the version of the sentential calculus
then the students will be unhappy. Therefore class will be dismissed.
(c) If either the students are happy or classes are dismissed then Roger we are calling system SC, we need to raise and provide answers to these
loses his bet. If Roger loses his bet, then Stan gains. Classes are to be questions:
dismissed and Stan will not gain anything. So the students are not going
l. Is the system consistent?
to be happy.
(d) If you are interested in history or in scenery, then you will leave the 2. Are the only formulas provable in the system tautologous?
main road. If you are not interested in history, then you won't leave 3. Are all tautologous formulas provable in the system?
the main road. Therefore if you are interested in scenery, you are
interested in history. Such questions are questions about system se and are spoken of as
(e) Logic makes no sense. Therefore either I give it up or logic makes sense metalogical questions. The theses in se are exhausted in the theorems of
iff I give it up. se. Thus the statement 'se is consistent' is a truth that goes beyond se
(f) The instructor will be pleased if Anderson takes the test and passes. But and is about se and is not a thesis in se.
if Anderson takes the test and writes illegibly, the instructor will not be There are several senses of the term 'consistent' when applied to a
particularly pleased. In fact Anderson will pass the test only if he writes system, but the basic notion of consistency is what is called negation
illegibly. (Then the instructor won't be able to see how bad his exam is.)
So Anderson won't both take the test and pass. consistency or simple consistency. A system is negation consistent if
(g) If JB has his way and KF is defeated, then the Fitzwigg dynasty will and only if the system does not permit the proof of a theorem of the form
come to an end. But it is not true that the Fitzwigg dynasty will come A 1\ ,.....,A. When a system has as theorems only acceptable formulas, the
to an end if KF is defeated. Therefore if KF is defeated, JB will not system is said to be sound. If it can be proven that only tautologies are
have his way. theorems of se, se will be proven sound.
6. For each valid argument found in exercise 2.5 and exercise 3.3, derive the A system is said to be complete, in one of the senses of this term, if
conclusion from the premises, using SC. all the acceptable well-formed formulas ofthe system are theorems of the
system. To say se is complete is to say that all tautologous formulas
Answers are theorems. Thus to prove se complete is to provide a decisive answer
3. For example, MT can be gotten from the other rules minus MT. So can MP, to the third important metalogical question above.
Add, and half of DN. Why is negation consistency a desired feature for system Se1 If se
were inconsistent, then it would not serve the purpose for which it was
5. (g) Let L: JB has his way
R: KF is defeated constructed, viz., providing or establishing tautologies. For if se lacks
F: Fitzwigg dynasty will come to an end consistency, then any formula would be a theorem. One way to show
The proof of the validity of this argument may be given as follows: this is to suppose that an inconsistency, say P 1\ ""P, is a theorem of se.
Prove: L 1\ R --+ F, ",,(R --+ F) I- R --+ ""L Since disjunctive syllogism is a theorem of se, we could produce this
1.LI\R--+F A proof:
2. ",,(R --+ F) A 1. P I\,....,P TI
3. R A 2. P 1, S
4. ",,(,-.....;R V F) 2, Imp 3. PvQ 2, Add
5. "" ""R /\ ""F 4, DM 4. ""P 1, S
6. ""F 5, S 5. Q 3,4 DS
7. ,-.....;(L 1\ R) 6, 1 MT
8. ""L V ""R 7, DM
And if Q is a theorem, then given theorem introduction, any well-
9. """"R 5, S
10. ,-.....;L 8,9 DS formed formula is a theorem of se. If this obtained, se would be a
11. R --+ r--I L RCP useless theory.
98 Natural Deduction System SC 4.8 Soundness, Consistency, and Completeness of SC
99
How are we to go about showing that SO is consistent? Olearly,
Exercises
our not having already derived P /\ ",-,P is not sufficient to establish
the consistency of SO. We need some systematic method. One such 1. Suppose a nontautologous formula to be a theorem and show how this would
procedure is to show that every theorem of SO must be a tautology. In make SO inconsistent.
other words if we can show that SO is sound, then it must be consistent,
2. Supposing SO to be sound, show that SO is absolutely consistent (a system is
since no formula which is equivalent to F can be a tautology.
said to b~ abso~utely c?nsistent if not eve~y wff ?f the system is a theorem)
A precise proof of the soundness of SO will be taken up in Part II of and consIstent m Post s sense (a system IS consIstent in this sense if a wff
this book. At this time let us consider some of the notions relevant to consisting of a sentential variable alone is not a theorem).
a proof of the soundness of SO.
We note that when a theorem is obtained in SO, then either ROP or 3. If SO is complete,. sh~w why the premises of a sentential argument logically
RAA is used and all assumptions are discharged. If either rule is so imply the conclUSIOn rEf the corresponding conditional of the argument is an
instance of a theorem of SO.
applied, then the result, a line from no assumptions, is a tautology if the
derived line-the B from A or B /\ ",-,B from A-is a tautological 4. Show for each rule in SO that if it is applied to a line or lines, the result is a
consequence from the final discharged assumption. Oonsider ROP and tautological consequence from these lines.
RAA in turn:
5. There is an interesting alternative method for demonstrating validity using
ROP rules called the method of semantic tableaux, 01' the tree method. The
Supposing from assumption A there is obtained the line Band B is a method is quite simple. Begin with premises and the denial of the conclusion.
tautological consequence from A, then A --'>- B is a tautology. Note For each compound statement (now we treat denials of simple statements as
that if B is not a tautological consequence from A, then A --'>- B is noncompound statements) we indicate below all the conditions under which
not tautologous. the statemer:t is true. If there a.r~ more than two conditions, we employ a
fork; otherWIse we stack the condItIons. An illustration will show how the t~'ee
RAA is built. The tree for
Supposing from A there is obtained the line B /\ ",-,B as a tauto-
logical consequence, then A --'>- B /\ ",-,B is a tautology. If A --'>- B /\ P --'>- Q I- """Q --'>- ",-,P
r-->B is a tautology, A must always be F; thus r-->A is a tautology.
would be as follows:
If when ROP or RAA is used, all assumptions have been discharged,
and if the derived line is a tautological consequence from the final 1. P --'>- Q
2. "'-' (r-->Q -+ "'-' P)
discharged assumption, and if a theorem can only be obtained by using 3. "'-'Q
ROP or RAA in this way, then all theorems must be tautologies. If all 4. P
theorems are tautologies, then SO is sound. And if all theorems are
5. ",-,p
AQ
tautologies, no theorem equivalent to F, no inconsistency, is a theorem.
Proving that SO is complete is also a task that will be saved for Part x x
II. We note that neither SO's soundness nor its consistency implies that
SO is complete. However if SO were inconsistent, then SO would be Line 1 is the premise. Line 2 is the denial of the conclusion. Lines 3 and 4
are the. conditions in which line 2 would be true-both r---;Q and P must be
complete. For if SO were inconsistent, any well-formed formula would
true. Smce both """Q and P must obtain, "'-'Q and P are stacked. We now
be a theorem and thus all tautologies would be theorems. t~rn to the only remaining compound statement, line 1. Line 1 is true iff
Why is the completeness of SO desirable? SO is constructed to eIther ",-,p or Q; thus we use a fork. 'Ve find two paths on the tree. If we had
establish tautologies. If SO were incomplete, then it could not be used to another compound sentence to break up, the truth conditions would have
establish all tautologies. Furthermore we wish to say that the premises ~o be ~dded to each path of the fork. Going up each path, we encounter an
~n~?nsI~tency-thus the use of 'x' at the end of each path. A path with an
of a sentential argument logically imply the conclusion iff the corre-
x IS saId to be closed. An argument is valid iff all of the paths of its tree are
sponding conditional of the argument is an instance of a theorem of SO. closed. All compound sentences in paths must be broken down unless all
This claim would be false if SO were not complete. the paths have closed. Ajinished tree is one with all closed paths, or one in
100 Natural Deduction System SO

which all compound sentences have been broken down (or both). Consider
these next examples of finished trees.
PI\QI-PvR P -+ Q, P V Q I- P 1\ Q
PI\Q P-+Q
r-.I(P V R) PvQ
Not all valid arguments are valid
,--,P ,--,(P 1\ Q) predicate because the conclusion is a tauto-
,--,R ~ logical consequence of the premises.
P ,--,p "-'Q
Q ~ ~ language To illustrate consider the two fol-
x P Q P Q lowing arguments:
x A /\ x
All men are mortal. All Greeks
,--,p Q ,.....,P Q
x x are men. Therefore all Greeks are
mortal.
The first is valid; the second is invalid. All circles are figures. Therefore
(a) Use the tree method to test the validity of some of the exercises in 4.3 whoever draws a circle draws a
or 4.4. figure.
(b) Use the tree method to establish the validity of some theorems of SO.
Note A -+ B is valid iff ,,-,(A -+ B) has a tree with each path closed. Each argument is valid, for it is
(c) If a path does not close in a finished tree, this path provides a counter- evident that the conditionals corre-
example. Read off some counterexamples from trees for invalid argument sponding to each argument are
patterns. necessarily true. However, if we
(d) Why does the tree method work? replace each simple statement with
sentential variables and supply the
Answers
appropriate statement connectives,
2. If A is a single sentential variable, then A is a nontautologous wff. Since all the result would not be valid argu-
theorems of SC are tautologous, A is not a theorem; thus SC is consistent in ment patterns. Neither
both senses.
P,QI-R
PI-Q
nor their corresponding conditionals
can be established using the system
SC for the sentential calculus .. What
this indicates to us is that the
validity of the above arguments is
not merely a matter of how simple
statements are related by statement
connectives, but depends on the
inner structure of the simple state-
ments. In the first argument, for
example, the validity depends on
the sense of 'all' as well as the way
the words 'men' and 'mortal' are
related.
The object of this chapter is to
make additions to the sentential
,.
101
102 Predicates 103
language so that the inner structure By singula-r term will be meant any expression that names or picks out a
of simple statements can be re- single thing-whether or not that individual exists in the realm we have
flected. The result will be called the under consideration. The above examples indicate that proper names,
prediwte language. Later (Ohapter
demonstrative expressions, and what are called definite descriptions can
6) the system for establishing valid
all be used as singular terms. Abstract nouns are also singular terms.
formulas that reflect the inner
structure of simple statements is ]l'or example, the subject term of the next sentence is a singular term.
introduced.
Courage is a virtue.

vVhen we have expressions that contain one or more individual

5. 1 Predicates
variables and that produce a statement when a singular term (or terms)
replaces the variable(s), we have what are called open sentences. Thus
Several kinds of symbols are needed for the predicate language. One
'x is a person' and 'x > 7' are open sentences as well as 'x is married to y'
such symbol is what in this book will be called a predicate term.
and 'x = y'.
To begin the explanation of the notion of a predicate term, let us
One way to obtain open sentences is to replace the occurrences of
consider the following expressions:
names with individual variables. Thus from the sentences
x is a person.
George Washington is alive in Argentina.
x> 7.
The number 6 > 5.
These are neither true nor false. But one could obtain true or false
we may obtain open sentences as follows:
statements from these expressions in this way: Substitute a name for
one of the things the x ranges over in each of these expressions. It is x is alive in Argentina. x is alive in y.
safe to say that in this first expression x ranges over individuals such as The number x > 5. The number y > 5.
Barry Goldwater, Bertrand Russell, and Tiny Tim. Thus if we supply The number x > y. George vVashington is alive in x.
a proper name for a person we obtain a statement from the first ex-
pression. For example, if we supply 'Shirley Temple Black' for x in the or from
first expression we obtain the true statement:
Straws on knows Jones is a fink.
For all x, x > O.
Shirley Temple Black is a person.
we may obtain the open sentences
Similarly, if we supply numerals for x in the second expression, we can
obtain a statement, for the x here is naturally taken to range over x knows Jones is a fink.
numbers. For example, if we supply 8 we obtain a true statement, but For all x, x > y.
if we supply 7, we obtain a false statement.
The letters x, y, z, ... used as variables for names, as x is used above, Open sentences are represented in a special way in the predicate
are called individual variables. Individual variables will be one kind of language. An open sentence is indicated by a capital letter followed by
symbol that is necessary for the predicate langugage. Individual the individual symbols involved. Thus if the open sentences are
variables may take singular terms as values. Thus fDr the first expression,
x is a person, we may obtain the following instances: x is a little girl. x melts.

Strawson is a person. they will be represented in this way:

That fellow is a person.
The philosopher who wrote The Bounds of Sense is a person. Lx Mx
~I·,
104 Predicates 105
where L is taken to stand for 'is a little girl' and M is taken to stand for cannot do is use different singular terms for the same kind of individual
'melts'. Similarly, if we allow Nto stand for' is next to , variable.
x is next to y becomes Nxy. To indicate the open sentence that, say, Fx It wiII prove convenient from time to time for us to indicate
represents, we wiII use the following form: explicitly what the individual variable ranges over in an open sentence
by using the sign for the universe of discourse or domain of discourse in set
Fx:xis---
theory, V. If the x ranges over cities, this may be indicated in this way:
V: cities. But if the x ranges over physical objects or persons, this may
Thus to indicate that Gxyz is the open sentence 'x gave y to z', we would be indicated, respectively, in this manner: V: physical objects; and V:
write: persons. If we restricted V to the set {I, 2, 3}, that is, if our universe of
discourse is the set made up of only three individuals, the numbers 1,
Gxyz: x gave y to z.
2, and 3, then the open sentence
It makes no difference, we should note, what individual variables we use Ox: x is odd.
to represent open sentences. Thus the last may be written: would yield three statements, when we replace x with a name of one of
the individuals in {I, 2, 3}, only one of the statements being false.
Gzyx: z gave y to x.
Exercises
An expression such as Lx is also known as a one-place predicate. 1. Obtain open sentences, and thus n-place predicates, from the following
Nxy is a two-place predicate, and Gxyz is a three-place predicate. A sentences.
predicate may be defined as what occupies the F position in (a) 10 is greater than 5.
(b) 7 +5= 12.
(c) Americans prefer Democrats to Republicans.
F( (d) Chicago is larger than Phoenix.
(e) This book = this book
where ( indicates the positions to be occupied by individual (f) It is raining.
variables. An n-place predicate is an open sentence containing n 2. Indicate if the expressions below are one-, two-, or three-place predicates.
different kinds of individual variables. In Lx, n = 1; thus Lx is, as (a) x <y
indicated above, a one-place predicate. In Nxy, n = 2; thus Nxy is a (b) x < 10
two-place predicate. But if we had Nxx, we would have a one-place (c) x +y= y +x
predicate since only one kind of individual variable is involved. When a (d) x +y < x +6
(e) x +x = x
predicate contains more than one occurrence of the same kind of variable, (f) x+y = z
it is understood that the same name of an individual is to be supplied (g) For any x, x + y = 10
for each occurrence. To illustrate, 3. If possible, obtain true propositions from the following n-place predicates by
supplying singular terms.
x +x = 10. (a) V: everything: x is smaller than y, x exists, x is red and green all over
(b) V: positive integers: x + 2 = 5,x + x = x,x + y < x + 6,x(y + 2) =
is a one place predicate from which we can get statements such as xy + xz
2 + 2 = 10 and 10 + 10 = 10. However, (c) V: human beings: x is the present President of the United States, x is
older than y and y is older than z, everyone loves x, x is wise and is a
United States Senator
x +y = 10.
Answers
is a two-place predicate from which we can get statements such as 1. (d) x is larger than y, Chicago is larger than y, x is larger than Phoenix.
2 + 3 = 10, 10 + 7 = 10, and 2 + 2 = 10. It is permissible to use the 2. (a) 2, (b) 1, (c) 2, (d) 2, (e) 1, (f) 3, and (g) 1.
same singular term for different kinds of individual variables. What one 3. (c) J. William Fulbright is wise and is a U.S. Senator.
 Predicate Language 107
106
where, naturally, the universe of discourse for 'rIxPx is numbers and the
universe of discourse for 'rIx,....,Hx is persons. In other words, in saying,
5.2 Quantifiers for example, 'rIxPx, we are saying

There is a second way to obtain a statement from an open sentence For all x's (in the class of numbers), Px.
besides supplying a singular term for the individual variable. We can
prefix expressions such as 'for all x' and 'for some x'. For example, The quantifier 'for some x' is called the existential quantifier and will
supposing the universe of discourse is persons, the open sentence be denoted by this symbol: 3x. Each of the following statements may be
paraphrased as 3xMx:
x is mortal.
For somc x, x is mortal. }
turns into a statement in either of these two ways: Some x is such that x is mortal.
There exists an x such that x is mortal. 3xMx
There is an x such that x is mortal.
For all x, x is mortal. There is at least one x such that x is mortal.
For some x, x is mortal.
where, it seems likely, the universe of discourse is persons. Similarly
The expressions 'for all x' and 'for some x' are called quantifiers, since
they indicate the quantity of the individuals being considered in the There is at least one number that is prime.
universe of discourse. Since the universe of discourse for the above At least one person is not happy.
sentences is persons, the last statement is, in effect:
may be paraphrased into
For some x (in the class of persons), x is mortal.
3xPx where Px: x is prime
3x"",Hx where Hx: x is happy
If, say, the universe of discourse were the gods, then we would have:
where the universe of discourse of the first is numbers and of the second ,
For some x (in the class of gods), x is mortal.
persons.
Syllogistics, traditional Aristotelian logic, treats four types of
The quantifier 'for all x' is called the unive1'sal quantifier' and for this
propositions, which are called categorical propositions. Categorical
quantifier we shall use this symbol: 'rIx. Using the one-place predicate
propositions were regarded in traditional logic as sUbject-predicate
'x is mortal' (Mx) and the new symbol, each of the statements below
statements and were classified as affirmative or negative depending on
may be paraphrased as 'rIxMx: whether the subject was affirmed or denied of the predicate, and uni-
versa~ or particular. depending on whether all or some of the things
For all x, x is mortal. }
des~l'lbed by the subject were under consideration. Allowing S to be the
For every x, x is mortal. 'rIxMx
For each x, x is mortal. subject and P the predicate, the four forms of a categorical proposition
For any x, x is mortal. are:

Again using predicates and the universal quantifier, we obtain these Affirmative Negative
Universal A: All Sis P E:NoSisP
paraphrases: Particular I: Some S is P 0: Some S is not P
All numbers are prime numbers. 'rI xPx where Px: x is a prime
where A, E, I, and 0 are used t~ identify the four forms.
number.
'rIx,...., Hx where Hx: x is If. we Suppose the universe of discourse to be everything and let
Everyone is unhappy.
happy. x IS Sand Px: x is P, then the particular affirmative propositions can
108 Quantifiers 109

be paraphrased of Bx, and the second quantifier, naturally, only has the scope of ,,-,Bx.
The parentheses thus serve as punctuation marks to indicate what the
3x(Sx A Px) quantifier has within its scope. If we wished to symbolize:

read "there is an x such that x is S and x is P." The particular negative (5) If anything is digestible, it is eaten.
categorical proposition, 0, would be rendered:

3x(Sx A ,",-,Px)
we would do the following symbolic paraphrasing:

"there is an x such that x is S and it is false that x is P." If our universe \:Ix(Dx -+ Ex)
of discourse is everything, then to say 'All S is P' may be construed
as saying of anything if it is S then it is P. The A proposition thus rather than
goes over in our notation to
\:IxDx -+ Ex
\:Ix(Sx -+ Px)
for the 'it' in (5) clearly refers back to the 'anything', so the quantifier
"for any x if x is S then x is P" and the E naturally becomes must be made to cover the whole compound. Also note that (5) is a
statement, whereas \:I xDx -+ Ex is an open sentence and is neither true
nor false. We will adopt the convention-which we have in fact been
following-of not using parentheses when the scope of a quantifier is a
"for any x if x is S then it is false that x is P." After some further pre-
single predicate expression. Thus to indicate \:Ix(Fx) we will merely
liminaries, we will return to categorical propositions and raise the
question concerning the logical relationships between A, E, I, and
propositions.
° write \:IxFx.

In using quantifiers the scope of the quantifier is of prime importance. Exercises

The scope of a quantifier is the formula to which the quantifier applies,
and it is indicated by parentheses. To illustrate, in ordinary discourse 1. Symbolize each of the following, assuming that the universe of discourse is
we can easily see the disjunction between saying everything.
(a) Only idiots drink paint thinner.
(b) Some numbers are prime and odd.
(1) Everything is blue or not blue. (c) All that glitters is not gold.
(d) Nothing is a horse unless it is a mammal.
which is true, if we are talking about visible objects, and (e) If something is a horse, then it is a mammal.
(f) A thing is a pen only if it holds ink.
(2) Everything is blue or everything is not blue. (g) Only nonspiders are insects.
(h) At least one plant is edible.
which is clearly false. In the notation just introduced, (1) and (2) are (i) Not all philosophers are logicians.
symbolized respectively: (j) All Americans except New Yorkers are friendly.
(k) None but the brave deserve the fair.
(1) If there is a frustated dean, then something is frustrated and something
(3) \:Ix(Bx v,......,Bx)
is a dean.
(4) \:IxBx V \:Ix "-' Bx
(m) Everyone is good and kind if and only if everyone is good and everyone
is kind.
In (3) the quantifier has the scope ofthe area marked by the parentheses.
The quantifier in (3) covers the whole compound or has the scope of the 2. Distinguish (a) \:Ix,,-,Fx from l"o.'\:IxFx
entire compound. However, in (4) the first quantifier only has the scope (b) 3x,-.....,Fx from '"'-'3xFx
""'/-,
IlO Predicate Language r~t~rpretation of Quantifiers III
3. An individual variable can occur either free or bound in a formula. This be understood to mean that at least one ofthe statements Fa, Fb, Fc, ...
distinction may be put as follows:
An occurrence of an individual variable is bound if and only if it is
is true. In other words, 3xFx can be interpreted as a disjunction of
within the scope of a quantificational expression that contains an singular sentences as follows:
occurrence of that individual variable. An occurrence of a variable
is free if and only if it is not bound. 3xFx Fa V Fb V Fe V '"
For example in
Vx((Fx /\ Oy) -+ 3y(Gy /\ (Hx V Mz))) Disjunctions and conjunctions can only be finite in length. Often
the first y and the z are free, but the rest of the individual variables are we wish our universe of discourse to be made up of an infinite number of
bound. Which variables are bound and which are free in the following individuals. So, in general, the interpretation for the two quantifiers may
formulas?
be indicated in this way. Let Ax be any predicate that at least has x
(a) VxFx -+ Gx
(b) 3x(3yFxy -+ Gyx) not under the scope of a quantifier. And let us say that an individual,
(c) Fx -+ (VxFx -+ VyJ/x) say a, in a universe of discourse, V, satisfies Ax iff Aa is a true statement.
(d) 3x(Fx /\ VyFy -+ x = Y /\ Gx) With this in mind we may indicate the interpretation for the two
4. When in the natural language we use the form 'all Sis P', is this equivalent quantifiers as follows:
in meaning to Vx(Sx -+ Px)? Note that Vx(Sx -+ Px) has a truth-value
whether or not there are x's that are S. Also evaluate the following claim: V xAx is true in V iff all the individuals in V satisfy Ax.
The statement-making job done by 'all S is P' in the natural language is done 3xAx is true in V iff at least one of the individuals in V satisfies Ax.
adequately by Vx(Sx -+ Px) and when the two have different meanings the
statement in the natural language has failed in its statement-making To see how this interpretation for the quantifiers works out, let Lx:
function. x is less than 5, and Suppose that our universe of discourse is {l, 5, lO}.
'rhus to say VxLx would be to say Ll /\ L5 /\ L10. Since L10 is false
Answers the conjunction is false and thus V xLx is false. In short, since all th~
1. (j) Ax: x is an American, Nx: x is a New Yorker, Fx: x is friendly. individuals in V do not satisfy Lx, L10 being false, VxLx is false. But
Vx(Ax /\ "",Nx -+ Fx). to say 3xLx would be to say something true since to say 3xLx is to say
2. (a) Vx,,-,Fx says "nothing is F" whereas "",VxFx says "something is not F." L1 V L5 V L10 and L1 is true. In other words, 3xFx is true since there
is at least one individual in V that satisfies Lx, namely 1. To say
Vx Lx or 3x "'" Lx would amount to:
,-..0

5.3 Interpretation of quantifiers Vx"",Lx "",L1 /\ "",L5 /\ ,...."LlO

~x""" Lx ,....,L1 V "",L5 V "",LlO
In sentential logic an interpretation was given for the statement con-
nectives. This interpretation removed the possibility of ambiguity 'I'hus Vx "'" Lx is false, but 3x "'-' Lx is true. In turn, we obtain the
arising from their use. We wish to provide an interpretation for the relationships below in this example:
quantifiers that will similarly result in their unequivocal use. We desire
"",(L1 /\ L5 /\ LlO)
a fixed meaning for the quantifiers.
"",(L1 V L5 V LlO)
Suppose our universe of discourse consisted of a finite set of in-
dividuals a, b, c, ... To say VxFx, everything is F, will be so understood Thus """VxLx is true, but ,....,3xLx is false.
to mean that Fa and Fb and Fc and .... In other words VxFx can be For the universe of discourse
interpreted as a conjunction of singular sentences as follows:
{China, Cuba, U.S.S.R.}
VxFx Fa /\ Fb /\ Fc /\ ...
letting Ox,
' . ,x is a commums
. t country, we can produce statements such as
By a singular sentence is meant a predicate followed by n number of
names for individuals. In turn, the statement 3xFx, something is F, will VxOx, 3xOx, ,....,Vx "'" Ox, "",3x "'" Ox
112 113

whose truth-value, given the interpretations for the quantifiers, would be 5. We know by DM that the following biconditionals are valid:
T for each. Finally, if Fy: 3x(x + y = 10) and V: {O, 2, 5, 10}, then
Fa! V Fa 2 V ... V Fan<c---+ ,-.....,(,,-,Fa! A ,-.....,Fa 2 A ••• A ,-.....,Fa n)
(a)
(Fa! A Fa 2 A ... A Fan)<c---+ ,,-,(,.....,Fa! V ,,-,Fa2 V ... V,-.....,Fa n )
(b)
the following statements would have the indicated truth-value: Now let a! ... an represent all the objects in the universe of discourse.
Replace both sides of the biconditionals with existential or universal
F2 A F5 -+ FlO (T) statements.
"IyFy (F)
3yFy (T) Answers
1. (c) Both "IxFx and 3xFx are false.
Exercises 2. (d) ?xFx,""'" "IxFx, ,.....,"Ix,-....., Fx, 3x,....., Fx.
1. Let Fx: x is less than 10. Below are three universes of discourse. In each (e) ?
universe indicate the truth-value of "IxFx and 3xFx, and "Ix,,-, Px and Mx: x is a male
3. (d) "Ix(Mx A Ax -+ Bx)
3x"-' Px. "Ix(Mx A Ax -+ ,-.....,Bx) Ax: x is an American
(a) V: {,,-,2, 0, 2, 5.5}
3x(Mx A Ax A Bx) Bx: x is ambitious.
(b) V: {O, 5, 10} 3x(Mx A Ax A ,-.....,Bx)
(c) V: {15, 23, 10}
4. (f) "Ix(Px -+ Gx).
2. For each open sentence and its associated universe of discourse give four
true statements using quantifiers.
(a) Fx: x is odd. V: {,-.....,1, 0,1, 2}
(b) Fx: x is a free nation. V: {Cuba, U.S.A., New Zealand}
(c) Fx: x +2 is less than x +
3. V: {,,-,1, 0,5, ,-.....,3} 5.4 Valid predicate formulas
(d) Fx: 2x - 5 + x 2 = 3. V: {.....,2, 0,.2}
(e) Fx: x is a great leader and V: {Nixon, Ted Kennedy, Lindsay, Formulas in sentential logic can be classified as valid, inconsistent, or
a moral man. McCarthy} contingent. A similar classification can be made for predicate formulas.
Below are examples of each kind of predicate formula.
3. Put the following into symbolic form for each of the given universes.
Every male American is ambitious. valid: "IxFx -+ 3xFx
No male American is ambitious. inconsistent: 3x(Fx A ,.....,Fx)
There is a male American who is ambitious. contingent: 3xFx
There is a male American who is not ambitious.
(a) V: All male Americans It is necessary to give a precise account of what it means to speak of a
(b) V: All Americans
(c) V: All males predicate formula as valid. It is also useful to have an account for the
(d) V: Everything notion of validity so that it can apply to specimens of both sentential
formulas and predicate formulas. To this end, we first modify slightly
4. Let the universe consist of all positive integers and let the notion of a valid sentential formula. To assign a truth-value to the
Ex: x is even. components of a sentential formula will be to interpret a sentential
Ox: x is odd.
formula. Valid formulas will be those formulas that have the truth-value
Px: x is a prime.
Gx: x is greater than 7. true no matter what interpretation is supplied. A predicate formula can
Paraphrase each of the following into symbolic form. be said to be valid iff it is true on all interpretations. And one obtains an
(a) Some positive integers are odd. interpretation of a predicate formula, if the formula is composed out of
(b ) No integer is both even and odd. quantifiers, n-place predicates with bound variables, and sentential
(c) Some primes are odd. variables, in the following way: First, select a universe of discourse that
(d) 1 is even.
has at least one member (the reason for this proviso is examined at the
(e) Some integers are greater than 7.
(f) If any x is a prime then x is greater than 7. end of this section). Second, assign to each n-place predicate symbol an
~I ,
 1I5
Predicate Language

n-place predicate. Third, assign a truth-value to the sentential variables , . any'-"'" that is found in front of the quantifiers. Thus
leave 111
if any.
The formula VxFx -+ 3xFx is declared above to be a valid pre-
dicate formula. (Soon techniques to establish the validity of predicate
formulas will be introduced.) Let us interpret the formula as follows:
V: living Americans and Fx: x is a long distance runner. We see that the
,-....,F -+ F = T -+ F = F
statement resulting from VxFx -+ 3xFx on this interpretation is true.
If VxFx -+ 3xFx is valid, as it is, a true statement will result no matter
what (nonempty) universe of discourse is selected and no matter what
one-place predicate we supply for Fx. VxFx -+ ,-...,Vx,-...., Fx
In explaining the notion of a valid predicate formula it is said that if
a formula is valid, then it will be true no matter what universe of dis-
bccomcS
course is selected~so long as there is at least one individual in the T -+ ,-....,T = T -+ F = F
universe. vVhat is the reason for this proviso?
All formulas in predicate logic that are classed as valid yield true '
vve 0 btaIll al
1 F , the formula is false in an empty domain of discourse.
statements whether or not an empty term is assigned to a predicate
variable. By an empty tel'm is meant a predicate that is such that no Exercises
individual in the selected universe of discourse satisfies the predicate. I. 'rry to classify the following as valid, inconsistent, or contingent predicate
For example: formulas:
(a) Vx,-...., Fx -+ 3x,-...., Fx
(b) Vx(Fx /\ Gx) -+ VxFx /\ VxGx
(1) VxFx -+ 3xFx
(c) 3xFx /\ 3xGx -+ 3x(Fx /\ Gx)
(d) Vx(Fx -+ Gx) -+ (VxFx -+ VxGx)
comes out true even if Fx is an empty one-place predicate. For if no (e) converse of (c)
individual in V satisfies Fx, then VxFx is false and thus (1) is true. (f) converse of (d)
However, what happens if the universe of discourse itself is empty? 2. Give an interpretation for each of the predicate formulas in exercise 1.
For example, what if we have a statement of the form (1) and the V has no
3. Which of the following are valid even in an empty domain?
members~there are no individuals in V? The situation that develops is
(a) Vx(Fx /\ Gx) -+ VxFx /\ VxGx
simply this. If V is empty, then to say either 3xFx or 3x ~ Fx is to say (b) ,-....,3xFx -+ ,.....,Vx,-...., Fx
something false. Since ~VxFx is equivalent to 3x ~ Fx, it follows that (c) Vx(Fx -+ Gx) -+ (VxFx -+ 3xGx)
if 3x ~ Fx is false, then VxFx is true. If VxFx is true and 3xFx is (d) 3x(Fx /\ Gx) -+ 3xFx /\ 3xGx
false, then (1) is false. Not all the predicate formulas that we wish to (e) VxFx /\ Vx(Gx V Hx) -+ 3x(Fx /\ Gx) V 3x(Fx /\ Hx)
classify as valid break down in this way if the domain of discourse is 4. A model for a predicate formula A is an interpretation M for A such that A
empty. For example, VxFx -+ ~3x ~ Fx is true on interpretation even is true in M. A formula is inconsistent if it has no model.
if the universe of discourse selected is empty. (a) Give a model for some invalid consistent predicat~ formulas.
(b) Prove that the following sets of formulas are conSIstent:
It is customary in predicate logic to designate as valid any formula
,-....,VxFx Vx(Fx -+ Gx)
that is valid in any domain excluding an empty domain. This restriction 3xFx Vx(Fx -+ ,-....,Gx)
is not unreasonable, for a moment's reflection shows that we do not have
5. What must the nature of V be if we make an assertion that has an empty V?
many occasions to make assertions that have domains of discourse with
no members.
Answers
It is a simple matter to determine whether a formula holds in an
empty universe of discourse. Merely replace all universal quantified 1. Only (c) and (f) are invalid.
predicate expressions with T and the existential ones with F, being sure 3. Only (a) and (d).

~I·,
·I..lU Predicate Language 117

4. (a) If M is V: {1,2}, Fx: x is even, Gx: x < 3, then 3xFx A 3xGx __ the pattern accordingly, remembering again that universal
3x(Fx A Gx) is satisfied in M. ts expand into conjunctions, whereas existential statements
statemen . .
5. An inconsistent predicate determines V. _",~"r'" into disjunctions. If V: {a, b} then (3) IS eqmvalent to

Fa V Fb ~ Fa A Fb
5.5 Proving the invalidity of predicate formulas 'i'reating the '~' as a '--*', w~ may use th; redueti,o truth-table. test to
determine its validity. We wIll find that T sand F s can be consIstently
From the definition of a valid predicate formula it immediately follows . ed so that the premise is T and the conclusion F. If Fa is T and
that a predicate formula is invalid if we can give an interpretation that ~p . .
Fb is F (or vice versa), then the premISes are true and the conclUSIOn
results in a false statement. For example a proof of the invalidity of Thus V: {a, b}, Fa is T and Fb is F provides an abstract counter-
(1) 3xFx A 3xGx --* 3x(Fx A Gx) example to (3). A concrete counterexample is obtained by selecting two
individuals for a and b and supplying an interpretation for Fx so that Fa
is obtained when V: drawn figures, Fx: x is a square, and Gx: x is a circle. is true and Fb is F. For example: {I, 2} and Fx: x is odd.
For under this interpretation (1) becomes the false statement: If some To prove the invalidity of predicate formulas, we may also give
drawn figures are squares and some figures are circles, then some abstract interpretations. To illustrate:
figures are squares and circles. Or if V: {I, 2}, Fx: x is even, and Ox: xis
odd, then (1) is a false statement. Vx«Fx --* Gx) A Gx --* Fx)
To show that an argument pattern made up of predicate formulas is
is proven invalid when V: {a} and Fa is F and Ga is T. To prove the
invalid we can produce a counterexample. A counterexample for an
invalidity of the next formula
argument pattern is an interpretation that results in true premises and a
false conclusion. To illustrate, we demonstrate the invalidity of the 3x(Hx A Gx) A 3x(,-.....,Hx A Gx) AV x(Fx --* Hx) --* Vx(Fx --* Gx)
following argument pattern
using the abstract method, demands that we have three individuals in
(2) Vx(Fx --* Gx) ~ 3x(Fx A Gx) the domain of discourse.
with this interpretation: V: living animals, Fx: x is a saber-toothed tiger,
Exercises
Ox: x has claws. For given this interpretation 3x(Fx A Ox) is false-
there are no saber-toothed tigers among living animals, whereas I. Demonstrate the invalidity of the following formulas or patterns. Provide
a concrete counterexample for at least one.
Vx(Fx --* Ox) is true. To see that Vx(Fx --* Ox) is true let a, b, e, ... be
(a) Vx(Fx --* ,...,Gx) --* 3x(Fx A ,-.....,Gx)
names for living animals. Since universally quantified statements (b) Vx(Fx V Gx) --* VxFx V VxGx
expand into conjunctions in a finite universe, we obtain this relation- (c) (VxFx --* VxGx) --* Vx(Fx --* Gx)
ship: (d) ""VxFx ~ Vx,-....., Fx
Vx(Fx --* Gx)<.---+ (Fa --* Ga) A (Fb --* Gb) A (Fe --* Ge) A ... (e) 3x(Fx --* P) ~ Vx(Fx --* P)
(f) VxFx --* P ~ Vx(Fx --* P)
and since FCl, Fb, Fe, . .. are all false-no living animal is a saber- (g) ",,(Vx(Fx --* Gx) A Vx(Fx --* ,-.....,Gx»
toothed tiger-each conditional is true, and thus the conjunction is true. (h) 3x(Fx A Gx) V 3x(Fx A ,.......,Gx)
We have successfully produced a counterexample to (2). 2. If the categorical propositions A, E, I, 0 are represented as VxFx, V x Fx, r-.J

Truth-table procedures can be usefully employed to find abstract 3xFx, and 3x,-....., Fx, respectively, then the following are valid formulas:
counterexamples. Suppose we wish to demonstrate the invalidity of the A --* I (A implies I)
next pattern: E --* 0 (E implies 0)
,-.....,(A A E) (A and E are contraries)
(3) 3xFx ~ VxFx I V 0 (I and 0 are sub contraries)
,-.....,(E <.---+ I) (I and E are contradictions)
We may do this by supposing a universe of two individuals {a, b} and ,-.....,(A <.---+ 0) (A and 0 are contradictions)
/-,
 lIS Predicate Language Proving the Validity of Predicate Formulas 119
If A, E, I, and 0 are represented as Vx(Fx -+ Gx), Vx(Fx -+ '"'-'Gx), validity of these formulas depends on the meaning ofVx and 3x and
3x(Fx A Gx), and 3x(Fx A ,,-,Gx) , respectively, then only the last two formulas
are valid. Demonstrate the invalidity of the first four formulas in this latter }S no t »1
>YO erely a matter of how simple statements
.
are.
related by statement
case. (Note (a), (g), and (h) in the first exercise.) oonncctl'ves . Thus they cannot be obtamed from tautologIes.
One way to establish the validity of these formulas is to use a
Answers .1 l'o nd ab8~t1'dum technique. For example, consider (1) and suppose
t'elbUC .~ cc
1. (c) Abstract counterexample: V: {a, b}, Fa is T, Fb is F, Ga is F, and Gb is 1'.
it is false:
Concrete counterexample: V: {I, 2}, Gx: x is even and Fx: x is odd.
(f) V: {a, b}, P is F and Fa is true and Fb is false. ,-...,(VxJ?x -+ 3xFx)
V: {I, 2}, P: 1 + 2 = 2, Fx: x is less than 2.
If this is true, then VxFx must be true and 3xFx must be false. If
,-...,3xF;r is true, then Vx "-' Fx. If VxFx is true, then Fa is true where ct
is an individual in our universe of discourse. Similarly, if Vx "'-' Fx is
5.6 Proving the validity of predicate formulas
t.rno, then ,-...,Fct is true. Thus we obtain a contradiction Fct A ,-..."Fct, so
If a predicate formula has the form of a tautology, then it is valid. An our original assumption must be false. To take another example, let us
indefinite number of valid predicate formulas can thus be obtained from informally prove the validity of (6) by this RAA method, using suc-
sentential logic by uniform substitution of quantified formulas for cessively numbered lines as follows:
sentential variables. This may be illustrated by our making the 1. ,-...,(3x(Fx V Gx) -+ 3xFx V 3xGx)
following substitu.tions for the tautology (P -+ Q) A P -+ Q: 2. 3x(Fx V Gx)
3. ,-...,(3xFx V 3xGx)
VxFxjP, (VxFx -+ Q) A VxFx -+ Q 4. Fct V Gct
VxFxjP, VxGxjQ (VxFx -+ VxGx) A VxFx -+ VxGx 5. ,-...,3xFx
3xFxjP, VxFxjQ (3xFx -+ VxFx) A 3xFx -+ VxFx 6. VX"'-' Fx
7. ,-...,Fct
A second way to obtain valid predicate formulas directly from tautologies 8. Fa A ",,-,Fa
is to substitute predicates for sentential variables uniformly and prefix
the result with a universal quantifier as the next examples illustrate. In this informal proof, 2 and 3 are true if 1 is true. 4 follows from 2,
our now taking 'ct' as the individual in V for which 2 is true. 5 comes from
PAQ-+P
3 by SO (DM, DN and S), 6 from 5 since (10) is valid; 7 from 6-if
Fxj P, GxjQ Fx A Gx -+ Fx
Vx(Fx A Gx -+ Fx) Vx "'-' Fx then for any individual, including 'ct', "'-'x is F. 8 comes from
7 and 4 by SO.
There are, however, valid predicate formulas other than those The valid predicate formulas above have something in common.
obtainable in this direct way from tautologies. Here is a list (there is an '1'hey contain only one-place predicates. For such formulas a truth-table
infinite number) of such valid predicate formulas: decision procedure is possible. It has been established that a predicate
(1)VxFx -+ 3xFx formula with only one-place' predicates is valid iff it is valid in a domain
(2)(VxFx -+ VxGx) -+ (VxFx -+ 3xGx) of 2" individuals where n equals the number of kinds of one-place
(3)Vx(Fx -+ Gx) -+ (3xFx -+ 3xGx) predicates in the formula. Thus formula (10) is valid iff it is valid in a
(4)Vx(Fx -+ Gx) -+ (VxFx -+ VxGx) domain of two individuals. Formula (2) is valid iff it is valid in a domain
(5)Vx(Fx A Gx) -+ VxFx A VxGx
of four individuals. We may then establish the validity of (10) in this
(6) 3x(Fx V Gx) -+ 3xFx V 3xGx
(7)VxFx V VxGx -+ Vx(Fx V Gx) manner. First, write the equivalent to (10) in a V of two individuals as
(8)VxFx A 3xGx -+ 3x(Fx A Gx) follows:
(9)VxFx_ "",3x "'-' Fx
(10) Vx,,-, Fx_ ,,-,3xFx
(ll) ,,-,YxFx_ 3x "'-' Fx Second, use truth-table techniques to see if the resulting formula is a
(12) "-,Vx",,, Fx_ 3xFx
tautology. If it is, (10) is valid and if it is not, UO) is invalid.
 120 Predicate Language 121
Such truth-table methods have their limitations in predicate logic. A one-place predicate determines a class. For example, Fx de-
It has also been demonstrated that there is no truth-table decision pro- the class of those things and only those things that have
cedure for all predicate formulas; in fact, there is no effective or
mechanical decision procedure for all predicate formulas. But this fact
~~l
.ty F
. If we let Fx: x is odd and V be the positive integers, then.
. determines the class whose members are 1, 3, 5, 7, .... DyadIc
does not prevent us from constructing a proof system similar to SO redicates determine a special kind of class~what is called a class of
that can establish the validity of all valid predicate formulas. Such a P pairs. Fxy determines the class made up of x's and y's that stand
system is introduced in the next chapter, and, as we will see, it makes use . the relationship Fxy. Thus if Fxy: x loves y and V: living persons,
of the moves involved in the RAA proofs above, along with other moves. In determines the class of x's and y's such that x loves y. The sym-
Exercises bolization of dyadic predicates makes clear the order of the individuals
or the direction of the dyadic predicates. Thus Fxy makes clear that x
1. In the following, establish the validity of the valid formulas by using one f is F to y, and Fyx makes clear that y is F to x. Similar considerations
the three methods discussed. Be sure to use each method at least once. F~r
those formulas that are invalid, demonstrate their invalidity. hold for the other polyadic predicates.
(a) VxFx V 3xGx ->- 3x(Fx V Gx) There are many occasions when we shall find it necessary to expose
(b) (3xFx ->- VxGx) ->- Vx(Fx ->- Gx) polyadic predicates in our symbolizing arguments in order to obtain the
(c) Vx(P ->- Gx) ->- (P ->- VxGx) logical structure relevant to the validity of the argument. For example,
(d) Vx(Fx ->- Gx +-+ ,,-,Fx V Gx) mcdievallogicians were particularly troubled over a valid argument that
(e) (VxFx ->- VxGx) ->- (,,-,VxGx ->- "-'VxFx)
defied validation in syllogistics. One version of the celebrated argument
(f) Vx(Fx ->- Gx) ->- (3xFx ->- VxGx)
(g) (P V 3xFx ->- VxGx) ->- (P ->- VxGx) 1\ (3xFx ->- VxGx) appeared at the beginning of this chapter and is
(h) Vx(Fx V (Fx V Gx) ->- Fx)
(i) "-'3xFx ->- Vx(Fx ->- Gx) (1) All circles are figures. Therefore whoever draws a circle draws a
figure.
2. Is the statement
If all men are mortal, then some men are mortal. If (1) were symbolically paraphrased in this manner:
of the form VxFx -:+ 3x!x-a vaIidformula-or of the form Vx(Fx ->- Gx)--;.
3x(Fx 1\ Gx)-an InvalId formula? Can it be an instance of both forms? Is vx(Cx ->- Fx) I- V x(Dx ->- Ix)
the statement analytic?
where Ox: x is a circle, Fx: x is a figure, Dx: x draws a circle, and Ix:
x draws a figure, it is intuitively obvious that its conditional would not
5. 7 Polyadic predicates exemplify a valid predicate formula. In fact what we have just done to
(1) is an incorrect symbolization since it violates the principle for
So far most, though not all, of the n-place predicates we have con- correct paraphrase (see Section 3.3). The trouble arises in leaving 'x
sidered are one-place or monadic predicates. As noted earlier, one-place draws a circle' and 'x draws a figure' as one-place predicates, rather than
predicates are predicate expressions with only one kind of free individual introducing a two-place predicate 'x draws y'. In many cases, for purposes
variable. But we may have two-place (dyadic) predicates, three-place of establishing validity, one can leave relations unexposed; indeed one
(triadic) predicates, and so on. Such n-place predicates are called can represent a statement as P, but not in this case. A polyadic or
polyadic predicates. Examples of such predicates would be: relational predicate needs to be introduced to display the structure of
the conclusion relevant to the validity of the argument.
Gxy: x is greater than
Exy: x is equal to y
Y) How can (1) be correctly paraphrased~ If we select this inter-
Lxy: x loves y two-place predicates pretation:
Rxy: x + 2y = 5
V: natural things
Lxyz: x loves y more than z} .
Gxyz: x +y > y +z three-place predICates Cx: x is a circle
Fx: x is a figure
Pxyzw: x pays y to z for w four-place predicate Dxy: x draws y
'""/-,
 122 Predicate Language P"lvad.1c Predicates 123
then (1) can be correctly symbolized. The first premise is as above Cy /\ Dxy. The rule to follow, however, is to change the individual
Vx(Cx -+ Fx). One way to go about symbolizing the conclusion usin~ variable as each new quantification is introduced when one is sym-
the two-place predicate Dxy is illustrated as follows: bolizing the structure of a simple statement. In some cases no harm
results from violating this rule, but in other cases it does (as with the
Whoever draws a circle draws a figure. inner expression), so a good policy is always to follow the rule. The
1. Vx(x draws a circle -+ x draws a figure) final step 3 is to put the various pieces together to form the statement,
2. 3y(Gy /\ Dxy) 3z(Fz /\ Dxz)
maJdng sure the Vx has the scope of the entire formula since we wish the
3. Vx(3y(Gy /\ Dxy) -+ 3z(Fx /\ Dxz»
final x in Dxz to link up with Vx.
Let us now use this procedure in symbolizing some more examples
The procedure is to first ask oneself if the statement is a universal or an using the interpretation given at the right of each example:
existential statement. That is, does the statement have the form
Vx( -+ ) or does it have the form 3x( /\ )? Clearly the statement
Every student who passes knows something.
'Whoever draws ~ circle draws a figure' has the first form. To say
Vx(x is a student who passes -+ x knows something)
whoever draws a CIrcle draws a figure is to say if anything is a drawer of a Sx /\ Px 3yKxy
circle then it draws a figure. Having decided that our statement has the Vx(Sx /\ Px -+ 3yKxy) V: person
form Vx( -+ ), we write the statement, injecting this form, putting Sx: x is a student
in the individual variable x, and leaving in the rest of the English. What Px: x passed
we obtain is line one. The second step is to symbolize the remaining Kxy: x knows y
parts. In the antecedent within the parentheses we have
Some student is liked by all professors.
3x(x is a student /\ x is liked by all professors)
x draws a circle
Sx Vy(Py -+ Lyx)
3x(Sx /\ Vy(Py -+ Lyx» V: person
To say 'x draws a circle' is to say 'x draws some circle or other', not that Sx: x is a student
x draws every circle. Given the above interpretation, the former is Px: x is a professor
symbolized: Lxy: x likes y

(2) 3y(Gy /\ Dxy)

There is no greatest prime number.
,.....,3x(x is a prime /\ x is greater than any other prime number)
Px Vy(Py /\ ~Ixy -+ Gxy)
and the latter would be symbolized ,.....,3x(Px /\ Vy(Py /\ ,.....,Ixy -+ Gxy» V: positive integers
Px: x is a prime
(3) Vy(Gy -+ Dxy) Gxy: x> y
Ixy: x = y

We want the former, (2). It is useful to note at this point that (2) does not
There is a mother and a father for every child.
say that x draws some particular circle, but that x draws some circle or Vx(Gx -+ x has a mother and father)
other. (2) is true if x draws circle a or x draws circle b, etc. The same 3y3z(Myx /\ Fzx) V: persons
considerations apply to symbolizing the consequent found within the Vx(Gx -+ 3y3z(Myx /\ Fzx» Gx: x is a child
parentheses of 1. It should not be overlooked that in line 2 we put a y Mxy: x is the mother of y
rather than an x in the first inner expression and a z rather than an x or y Fxy: x is the father of y
for the second inner expression. The outside x has the scope of the
formula, so we needed y in the first expression to avoid ambiguity (and In an interpretation it makes no difference what individual variables
for other reasons). We need not have used z; we could have used y, for we use. For example, the x and y in Fxy indicates only that we have a
example, in the second inner expression since the 3y only has the scope two-place predicate. This could be indicated by Fyz or, if we like, FCD®.
""/ ..
124 Predicate Language
125
To continue with some more examples of paraphrases of sentences
It will be observed that each of the following formulas is valid:
using predicate symbols, consider:
(4) 3yVxFxy ---+ Vx3yFxy
Politicians prefer wealth to power. (5) 3xVyFxy ---+ Vy3xFxy
Vx(Px ---+ x prefers wealth to power) (i.e., any wealth to any power)
VyVz(Wy II Oz---+Exyz) V: persons . that the converse of each is not valid. Much is to be learned from this
Vx(Px ---+ VyVz(Wy II Oz ---+ Exyz)) Ox: x is power
concerning the effect that the order of Vx and 3x can have on meanin?
Exyz: x prefers y to z
Px: x is a politician Let us try to bring out the sense of the antecedent and consequent.m
Wx: x is wealth each of the above conditionals and see exactly why (4) and (5) are valId,
hereas the converse of each is not. We begin with (4).
Sisters have a common parent. \V Suppose that our universe of discourse consisted of just three
VxVy(x is the sister of y ---+ x and y have a common parent)
individuals: a, b, and c. And let Fxy: x loves y. In this case the general
Sxy 3z(Pzy II Pzx)
VxVy(Sxy ---+ 3z(Pzy II Pzx)) V: persons proposition below would mean the same as the accompanying string of
Sxy: x is the sister of y disjuncts.
Pxy: x is the parent of y
3yVxFxy VxFxa V VxFxb V VxFxc
Consider, finally, the statement:
To say 3yVxFxy is to say that some individual a or b or c is loved by
Every gentleman prizes a blond. every individual. If everyone loves a or b or c, then 3yVxFxy would be
true. Thus 3yVxFxy reads "everyone loves the same person" or "some
This statement can be understood in at least these two ways: person is loved by everyone." With the consequent of (4) we obtain
(i) There is some blond who is prized by every gentleman. Vx3yFxy 3yFay II 3yFby II 3yFcy
(ii) For every gentleman there can be found some blond whom he
prizes.
which reads "everyone loves someone or other" or, simply "everyone
loves." Thus if, say, Faa, Fbc, and Fca are true, Vx3yFxyistrue. Each
Using this interpretation: V: persons; Gx: x is a gentleman; Pxy: x
person does not have to love one and the same individual for Vx3yFxy
prizes y; and Bx: x is a blond, (i) and (ii) are paraphrased as follows:
to be true, but they do if 3yVxFxy is to be true. We thus see why (4)
(i) 3x(Bx II x is prized by every gentleman) is valid. However, if everyone loves someone or other, it does not
Vy(Gy ---+ Pyx) follow that everyone loves the same person. Thus the converse of (4)
3x(Bx II Vy(Gy ---+ Pyx)) is false.
(ii) Vx(Gx ---+ x prizes some blond or other) Turning to (5), the antecedent becomes
3y(By II Pxy)
Vx(Gx ---+ 3y(By II Pxy)) 3xVyFxy VyFay V VyFby V VyFcy

The reader is encouraged to study carefully each of the above examples. If a loves a and a loves b and a loves c, then 3xVyFxy would be true. In
The order of the quantifiers is unessential when the same quantifiers other words, if some one individual loves everyone, then 3xVyFxy is true.
are used, since the formulas below are valid: It thus reads: "someone loves everyone" or "everyone is loved by the
same person." The consequent of (5) becomes
VxVyPxy+--t VyVxPxy
3x3yPxy+--t 3y3xPxy
Vy3xFxy 3xFxa II 3xFxb II 3xFxc
However, the order with Vx and 3x is essential, since a change in order
which is true if everyone is loved by someone or other. For example, if
means a change in sense.
Fhc, Fbb, and Fac are true, then Vy3xFxy is true. Vy3xFxy thus reads
126 Predicate Language polyadic Predicates 127

"everyone is loved by some person or other" or, simply "everyone is To the interpretation above add a: Igloo, and symbolize the following.
(a) If no one loves Igloo, then he does not love himself.
loved." If everyone is loved by the same person, then 'everyone is loved'
(b) If anyone loves Igloo,. then it is Igloo.
is true-(5) is valid-but the converse is false. (c) If unless Igloo loves lumself, he loves no one, then Igloo loves someone
As some of these examples bring out, when we have only if he loves himself.
(d) If everyone whom Igloo loves, loves Igloo, then if Igloo loves everyone,
'Vx3yF-- or 3x'VyF-- everyone loves Igloo.
(e) If Igloo has a lover who has no lover, then Igloo is no one's lover.
one way to go about reading them is to start with the first quantifier and (f) If all the world loves a lover and no one loves Igloo, then Igloo isn't a
if the x is in the first blank, read F in the active voice. But if x is in the lover.
second blank, read F in the passive voice. Thus
4. A statement is analytic if it is an instance of a valid predicate formula. A
'Vx3y y causes x predicate formula using a term like a-an individual constant-is valid iff
3x'Vy y causes x it is true on all interpretations. We get an interpretation of a predicate
'Vx3y x causes y formula using a term like a when we satisfy the conditions set down earlier
3xVy x causes y (Section 5.4) and when we assign a name of an individual in the universe of
discourse to each individual constant. Try to figure out which of the state-
ments in exercise 3 are analytic.
read, respectively, "everything is caused by some y or other," "some x is
caused by everything," "everything causes some x or other," and "some
x causes everything." 5. Again, paraphrase with the interpretation given below.
V: the set of positive integers, Gxy: xis greaterthany, Qxy: 2x + y = 6,
Px: x is prime
Exercises
(a) For every integer there is a greater integer.
1. Paraphrase into logical notation the following statements using the suggested (b) There is an integer that is greater than every integer.
interpretation. (c) For every integer it is false that it is greater than itself.
V: the set of human beings, Pxy: x influences y (d) For every integer x, y, z, if x < y and y < z, then x < z.
(a) There is somebody whom someone influences. (e) There is a prime number greater than O.
(b) There is somebody whom everyone influences. (f) If 1 is less than any integer, then 0 is not an integer.
(c) Everybody is influenced by someone or other. (g) There is one particular integer that, if added to any integer multiplied
(d) Everybody is influenced by everyone. by 2, gives 6.
(e) Someone influences somebody. (11) If 2 is multiplied by any integer, it can be added to some integer or other
(f) Everyone influences somebody or other. to get 6.
(g) Someone influences everybody. (i) For any integer it can be added to some integer or other multiplied by 2
(h) Everyone influences everybody. to get 6.

2. Use the suggested interpretation and paraphrase each of the following into
logical notation. 6. Establish the validity of those formulas below that are valid by using the
V: the set of human beings, Ox: x is a coed, Px: x is a professor, Lxy: x RAA method discussed in Section 5.6. For those formulas that are invalid
loves y, Mxyz: x loves y more than z demonstrate their invalidity with abstract or concrete interpretations. Se~
(a) Every coed is in love with some professor or other. answers for how to proceed.
(b) Some coeds are in love with every professor. (a) 'Vx3yFxy -+ 3y'VxFxy
(c) Some coeds only love professors. (b) 3x(,-..,,(Fx -+ 'Vy(Gy -+ Fxy»)
(d) Some coeds love themselves more than any professor. (c) 3y'VxFxy -+ 'Vx3yFxy
(e) Some coeds do not love any professors. (d) 'Vy3xFxy -+ 3x'VyFxy
(f) Every coed loves some professor more than she loves herself. (e) 3x'VyFxy -+ 'Vy3xFxy
(g) Some coeds love a professor only if the professor loves coeds more than (f) 3x3yFxy -+ 'Vx3yFxy
himself. (g) 3x'VyFxy -+ ,-..,,'Vx'VyFxy
(h) There is a coed such that if any professor is loved by that coed then he (h) 3y3x'VzFxyz -+ 'Vz3y3xFxyz
" .loves her more than himself. (i) 'Vx(Fx -+ Gx) -+ 'Vx(3y(Fy /\ Hxy) -+ 3y(Gy /\ Hxy»
 128 129
Answers addition we need the following two kinds of logical constants:
1. (a) 3y3xPxy, (b) 3y'VxPxy, (c) 'Vy3xPxy, (d) 'Vy'VxPxy, (e)3x3ypxy
(f) 'Vx3yPxy, (g) 3x'VyPxy, (h) 'Vx'VyPxy , Statement connectives: "-', A, V, -+, and~
Quantifiers: 'V and 3
2. (d) 3x(Cx A 'Vy(Py ---+ Lxxy», (h) 3x(Cx A 'Vy(Py A Lxy ---+ Lyxy»
3. (d) 'Vx(Lax -+ Lxa) ---+ ('VxLax -+ 'V Lxa) We wish to count as well-formed formulas only those strings of these
(e) 3x(Lxa A 'Vy "-' Lyx) -+ 'Vx "-' Lax mbols that, on interpretation, have sentences expressing statements as
4. Only (b) and (e) are not analytic. :~stances. We should not overlook that this requirement does not
restrict formulas with free individual variables from being wffs. For
6. (a) If V: {a, b}, (a) becomes
(Faa V Fab) A (Fba V Fbb) -+ (Faa A Fba) V (Fab A Fbb) example:
FTTT TTFF FFTF TFF
which is false under conditions indicated. Fx, Fx -+ 3xFx, 'VxFxy
(b) If V: {a, b}, (b) becomes the following nontautological formula:
,,-,(Fa -+ (Ga -+ Faa) A (Gb ---+ Fab» V may be included as wffs since we can extend the notion of in~erpretation
"-' (.Fb -+ (Ga ---+ Fba) A (Gb ---+ Fbb)) sO that they yield statements. To provide an interpretation of a formula
(c) 1. ",,-,(3y'VxFxy ---+ 'Vy3yFxy) with a free variable, all that one needs to do is to replace free individual
2. 3y'VxFxy
variables uniformly with a name for an individual in the universe of
3. ,,-,'Vx3yFxy
4. 3x'Vy,,-, Fxy from 3 discourse. Thus if V: positive integers, Fx: x is odd, and Fxy: x = y,
5. 'Vy"-' Fay from 4 selecting I for x in the above formulas would give us:
6. 'VxFxb from 2
7. ,,-,Fab from 5 I is odd.
8. Fab from 6 If I is odd, then some integer is odd.
Note that 'a' on line 5 is taken as the something-the x ofline 4 such that All integers are equal to I.
3x'Vy,,-, Fxy. We now need a different name to remove the 3y of line 2;
thus b is put in at line 6. It is worthwhile noting that, given this V and these predicates, depending
on what numeral one substitutes for the free variable, the first formula
will sometimes give true statements and sometimes false ones, the
second formula will also give true statements, but the last formula will
5.8 Formation rules for the predicate language
usually give false statements. The second formula, in fact, is a valid
formula since no matter what interpretation one may provide, it always
Up to this time we have relied on the reader's having an intuitive under-
gives true statements.
standing of what is and what is not a well-formed formula in the predicate
The decision needs to be made whether we care to admit formulas
language. It is now time to provide formation rules by which one can
that involve vacuous quantification. Vacuou8 quantification occurs when
tell in all cases whether or not a formula made up of the terms of the
one has a quantifier without a corresponding individual variable coming
predicate language is well-formed or not. The notion of a well-formed
under its scope. Examples of vacuously quantified formulas would be:
formula in the predicate language needs, in other words, to be made an
effective notion.
'VxP, 'VxFy, 'Vx3yFxz, 'Vx3y3zFxy
First let us set down the symbols we wish to include in the predicate
language (omitting from now on the mention of parentheses). We want No harm would come from allowing such formulas to be well-formed.
these kinds of variables: However, such formulas contain quantifiers that do not work. So let us
state the formation rules so that vacuous quantification is avoided: Also
Sentential variables: P, Q, R, ... . we wish to rule out formulas such as
Predicate variables: F, G, H, ... .
Individual variables: x, y, z, ... . (I) 'Vx'v'xFx, 'Vx'VxFxx, 'v'x(Fx -+ 'VxFx)
130 Formation Rules for the Predicate Language 131

as wffs. We wish to rule out as well-formed, formulas such as \:fx\:fxFxx All well-formed formulas of SO are well-formed formulas of
iang ua ge .
since one of the quantifiers is doing no work, the two occurrences of x predicate language. In applying. these rules we continu~ to drop
after F being controlled by either quantifier. Eliminating each kind as eses according to the conventIOns we have been followmg.
paren th . .
well-formed formulas will simplify to some extent the statement of the To take an example, the followmg IS a wff:
rules to be introduced in the next chapter. If we said
\:fx(Fx A (",-,Gx -+ \:fy3z "'-' Fyz))
If A is a wff which contains an individual variable x, then
\:fxA is a wff.
because

this would rule out vacuous quantification but would allow the (l)'s to be Fx, Gx, and Fyz are wffs by 1
wffs. To avoid (1) we need to stipulate that all occurrences of x within the ,-...;Gx and "'-'Fyz are wffs by 2
3z,-...; Fyz is a wff by 4
scope of the quantifier must be free in A. As with vacuous quantification,
\:fy3z,-...; Fyz is a wff by 4
no harm results in allowing formulas like (1); however, as we have said, C"",Gx -+ \:fy3z "'-' Fyz) is a wff by 3
we wish our wffs to have straightforward statements as instances. (Fx A (",-,Gx -+ \:fy3z "'-' Fyz)) is a wff by 3
To state the rules, it will be useful first to introduce the notion of \:fx(Fx A (",-,Gx -+ \:fy3z I"-' Fyz)) is a wff by 4
an atomic formula. Let us mean by a te1'm an individual variable. An
atomic formula is either Finally, if we have a formula that contains, say, Fx and Fxy, these
are taken to be different predicate variables since Fx is a variable for
(a) a sentential variable, or one-place predicates, whereas Fxy is a variable for two-place predicates.
(b) a predicate variable followed by n number of terms With the introduction of the predicate language, we are not only in
position to introduce systems that can generate all of t~e valid fo~~ulas
We can now dispense with listing some capital letters as sentential in the predicate language, but we can; at last, prOVIde a prOVISIOnal
variables and some as predicate variables. We can simply say that an clarification for some of the terms left undefined in the first chapter.
atomic formula is a capital letter followed by n number of terms. Thus 'fhe final clarification will be found in Section 7.3. First, jOl'mulas in
examples of atomic formulas would be: logic are well-formed formulas of the predicate language. To say such a
formula is valid is to say that no matter what interpretation we provide,
P, Fx, Fxy, Fxyz
the result is a true statement. Any instance of a valid formula in logic
expresses a necessarily true statement of the kind called analytic. With
In the first case, P, n = O.
respect to any argument, we can say it is valid if the premises logically
The definition for a well-formed formula of the predicate language can
imply the conclusion. The premises of an argument logically imply the
be set down in terms of the following rules:
conclusion if and only if the corresponding conditional is analytic.
1. All atomic formulas are wffs.
2. If A is a wff, then ",-,A is a wff. Exercises
3. If A and Bare wffs, then (A A B), (A V B), (A -+ B), and 1. For each of the following formulas state whether it is a wff. If it is a wff, give
(A +-t B) are wffs. a demonstration of this.
4. If A is a wff that contains an individual variable v, then \:fvA and (a) \:fxFxy
3vA are wffs, provided all occurrences of v within the scope of the (b) (\:fyFy -+ \:fxGx)
quantifier are free in A. (e) \:fx(Fx -+ \:fxGx)
(d) VxP
In these formation rules v in 4 is understood to be a variable for individual (e) \:fy(Fy -+ \:fxGx)
(f) (\:fz\:fxFz -+ Fa)
variables. We should note that these rules also generate all the well-
(g) \:fz(Fxy -+ Gz)
formed formulas of the sentential language. These rules make clear that (h) (P -+ \:fxFx)
the predicate language is the result of making additions to the sentential (i) ",-,Fx
132

(j) VxVxFx
(k) 3x(Fx --+ Vy3xFyx)
(1) 3x(Fx --+ Vz(Gz --+ 31lHxyu))
(m) (Fa --+ VxFx)
(n) I- (VxFx --+ 3xFx)
(0) (1 < 2 --+ 3xx < 2)
In the previous chapter the point
(p) 3y(3xFx --+ Fy)
natural was made that there is no effective
2. On interpretation, vacuously quantified formulas and double quantified decision procedure for the validity of
formulas such as Vx(Fx --+ VxGx) do generate full-blooded assertions that
are true or false. Provide interpretations for the following and indicate if
deduction all well-formed formulas of predi-
cate logic. However, this does not
possible, their truth-value. '
(a) VxP system preclude the construction of a system
that is capable of generating all the
(b) VxFy
(c) Vx3yFxz pc valid predicate formulas as theorems.
Such a system is called the pre-
(d) VxVxFx
(e) VxVxFxx dicate calculus. In this chapter one
(f) Vx(Fx --+ VxFx) such system will be described.
(g) 3x(Fx --+ VxFx) The system of this chapter will
(h) Vx(Fx --+ 3xFx) be an extension of the natural
deduction system SO. As with se it
3. As indicated in an earlier exercise, a formula has a model when it has an'
will prove useful to introduce in-
interpretation that results in a true statement. Think up some arguments in
support of the following true generalizations. ference rules with reference to
(a) For any finite number n, there are formulas that do not have models in argument patterns. Later in this
a domain of discourse containing less than n individuals. chapter the system will be ex-
(b) If a formula has a model in some finite domain, then it has a model in an plicitly set out and some of the
infinite domain. theorems of the system will be
(c) If a formula contains only one-place predicates, then it has a model in proven. In other words, we will
some finite domain. follow the same order of exposition
4. Show that for there to be a model for the set of formulas below, the domain that was used for se.
of discourse must be an infinite domain. The basic strategy in construct-
VxVy(Fxy --+ r-..lFyx) ing proofs for argument patterns
VxVyVz(Fxy --+ (Fyz --+ Fxz)) whose validity is a matter of how
Vy3xFxy predicates are related by quanti-
fiers, such as
Answers
1. Only (a), (b), (e), (g), (h), (i), (1), and (p) are wffs. Vx(Fx --+ Gx),
Vx(Gx --+ Hx) I- Vx(Fx --+ Hx)

will be this: First, we will drop the

quantifiers from premise assump-
tions according to two elimination
rules, one for the universal quanti-
fier and one for the existential
quantifier. Second, we apply the
rules of system se to derive the
conclusion minus the quantifier.

133
 134 Natural Deduction System PC 135
Finally, we add quantifiers accord_ Ohapter 4. It will be well at the outset to make clear how the apparatus
ing to two introductory rules for SO can be applied. To do this we must first give the precise account of
each type of quantifier. These new
counts as a tautology.
rules for taking off and adding
quantifiers constitute the primal'
A formula will be spoken of as basic iff it is an atomic formula or if it
extension of SC that will give us tl1e
y is of the form V xA or 3xA. A formula P that is not basic is formed from
. formulas called basic components of P. The basic components of a
macIunery to construct proofs for
all valid predicate formulas. These formula must be related by statement connectives. A formula is a
rules together with the ten rules f t(Putology iff it takes the value T for every assignment of truth-values to
SC will make up a natural deductio: its basic components. Thus, for example, the following are tautologies:
system for predicate logic that will
be called system PO after 'predicate P -+ P V Q, r--;VxFx -+ r--JVxFx, Fa V r--;Fa, Fa -+ Fa V 3xFx
calculus' .
We will be allowed in the new system PC to make use of SC since, as
6. 1 Individual constants and tautologies we will see later, PC is constructed by making additions to SC. At this
time the apparatus of SC may be used either to make tautological
In o~'der to simplify to some extent the statement of the four quanti- transformations of the basic components of a formula or to infer a
ficatIOn rules found in PC, it will prove useful to introduce individual tautological consequence from a formula. As was indicated in Chapter
constants into our bag of symbols. (Why the four new rules can be 4, B is a tautological consequence of A iff A -+ B is a tautology, and A
stated in a simpler form, with the consequence of increasing the number or a part of a formula containing A may be tautologically transformed
of terms and wffs in PC, will be indicated in exercise 6.4.2.) We will let into B if A t--t B is a tautology. And, as just indicated, A -+ B or
~, b, c, ... be individual constants. An individual constant is a term that A <--+ B are tautologies iff they take the value T for every assignment of
IS taken as a name for an element in the universe of discourse. If truth-values to their basic components.
pre~icate form~la contains an individual constant, an interpretation i: The import of these remarks for what is at hand is that we may
achIeved by umformly supplying a name for an individual in the universe make inferences such as those below since the second formula is a
of discourse. Thus an interpretation for tautological consequence of the first.

(1) Fa /\ Ga Fa /\ Ga VxFx /\ Fa Fa
r--; r--; r--; 3x F x Fa Fa V VxFx
might be V: posi~ive integers, Fx: x is odd, Gx: x is even, a: 1; or it might
be V: natural obJects, Fx: x is solid, Gx: x is heavy, a: Earth. Individual Exercises
const~nts differ from individual variables in the fact that they cannot be
1. Which of the following now are win
quantIfied, though they can come within the scope of quantifiers. (a) VxFxa
In the formation rules given at the end of the last chapter, an atomic (b) VxP
formula is a capital letter followed by n number of terms. We will now (c) VxFm
count as terms not only individual variables but also individual constants. (d) Vx(Gx -+ VzHxz)
To illustrate, we now count the formulas below as wffs: (e) 3xFx -+ VzGxz
(f) 3x(Gx /\ Ga)
Fa, Fa /\ Gb, 3z(Fz /\ Gaz), Fa -+ P, Fb -+ 3xFxy, VxFxa 2. Which of the following are valid formulas? Demonstrate the invalidity of
the invalid formulas.
(a) Fa -+ VxFx
At the beginning it was said that the basic strategy in constructing
(b) Vx Fx -+ Fa
proofs for predicate argument patterns is first to remove quantifiers (c) Fa -+ 3xFx
and then make tautological inferences using the apparatus of SC in (d) 3xFx -+ Fa
 136 Natural Deduction System 1'C Universal Elimination and Existential Introduction 137
3. Which of the following are tautologies?
(a) P /\ 'VxFx -+ P the domain is F. In other words, given the domain {a, b, c, ... }
(b) ,-...,'Vx(Fx /\ ,-...,Fx)
(c) 'V xFx /\ 3xGx -+ 3xGx ('VxFx -+ Fa) /\ ('VxFx -+ Fb) /\ ('VxFx -+ Fc) /\ ...
(d) Fa /\ P -+ Fa
(e) 'Vx(Fx /\ Gx -+ Fx) This general truth can be expressedin.the form of a.rule .comm~nl!
(f) 3xFx /\ 3xGx ~ ,-...,(3xFx -+ ,-...,3xGx)
lmown a S Universal Specification or Umversal QuantificatIOn EhmI-
4. In each pair, is the second formula a tautological consequence of the first? • . n To express this rule, let A be a well-formed formula of the
(a) 'VxFx . natlO . language, v a variable, an d t an III
redicate . d"d
IVI uaI t an
cons t . Then
,-...,,-...,AxFx ~(tlv) is the formula that results when each occurren~e of v in A is
(b) 'VxFx
Fa rep1ace d by an occurrence of t. In other words, A(tjv)
" IS the result of
bstituting t for v in A. To illustrate, suppose A IS 'VxFxy and v IS y.
(c) 'VxFx -+ 3xGx
,-...,3xGx -+ ,-...,'VxFx
(d) 'Vx(Fx -+ Gx)
;;t is a, then A(tJv) is 'VxFxa, and if tis b, then A(t/v) is 'VxFxb.

'VxFx -+ 'VxGx The rule of universal elimination is

(e) '-""('VxFx -+ 3xFx) DE 'IvA
'V xFx /\ ,-...,3xFx A (tjv)
(f) 'VxFx
3xFx
Examples of the use of this rule would be
(g) Fa -+ 'V xFx
,-...,(Fa /\ '-""'VxFx)
(h) 'Vx'VyFxy 'VxFx 'VxFx 'Vx3yFxy 'Vx(Fx -+ 3yFxy) 'VxFxa
'Vy'VxFxy Fa Fib 3yFay Fa-+ 3yFay Faa
(i) 'Vx,-..., (Fx V Gx)
'Vx(,-...,Fx /\ ,-...,Gx) In using this rule we must remember several things. First, t goes in
(j) Fa /\ Fb /\ ... for v, and not for any other variable. It only goes in for the v found in
'VxFx
'VvA. Thus the following inferences are erroneous:

Answers 'Vx3z(Fxz /\ Gxz) 'VxFxb

1. Only (a), (d), and (f) are wffs. 3z(Faz /\ Gaa) Faa

2. Only (b) and (c) are valid. To demonstrate the invalidity of (d), let V: since a went in for the second z and z is not v(x) in the first example, and
human beings, Fx: x is handsome, a: Charles DeGaulle.
in the second example a went in for band b is not v(x). Second, t goes in
3. All but (b) and (e). for each occurrence of v in A. Thus the following are erroneous:

4. The second lines in (a), (c), (e), and (g) are tautological consequences. 'Vx(Fx -+ Gx) 'Vx(Fx -+ Gx)
Fa-+ Gx Fa-+ Gb

Finally, the inferences below are also erroneous:

6.2 Universal elimination and existential introduction
'V xFx -+ 'V xGx 'VxFx -+ 'VxGx
W ebegin with a rule of inference that allows us to remove universal
Fa -+ 'VxGx Fa-+ Ga
quantifiers from formulas of the form 'V xA. The rule expresses the truth
that if we are given, say 'VxFx, then F _ follows no matter what term
since to UE we must begin with a formula of the form 'VvA, where the
we place in _. For if in a given domain every x is F, then each element
quantifier has the scope of A.
138 Natural Deduction System Pc Universal Elimination and Existential Introduction 139

To illustrate a simple use of DE, consider the proof below: We now illustrate some uses of EI and UE in the next two proofs.

Vx(Fx -+ Gx), Vx(Gx -+ Hx) f- Fa -+ Ha VxFx f- 3xFx

1. Vx(Fx -+ Gx) A 1. VxFx A
2. Vx(Gx -+ Hx) A 2. Fa 1, DE
3. Fa -+ Ga I, DE 3. 3xFx 2, EI
4. Ga-+Ha 2, UE
5. Fa -+ Ha 3, 4, Trans Vx(Fx -+ Gx) f- Fa -+ 3xGx
1. Vx(Fx -+ Gx) A
Fa A
The basic strategy of constructing proofs for predicate argument
3. Fa-+ Ga I, DE

IT
patterns is first to remove the quantifiers. This can be done with the 4. Ga 2,3, MP
premises of this pattern now that UE is available. Once t's are inserted 5. 3xGx 4, EI
for v's, the result is atomic formulas related by -+. 5 is a tautological 6. Fa -+ 3xGx RCP
consequence of 3 and 4.
The second rule of inference reflects the valid move of going from, Exercises
say, Fa to 3xFx. If an a in the domain of discourse has F, then some- 1. Construct proofs for the following, using DE and EI.
thing in the domain has F. The rule that warrants such a valid (a) VxFx f- Fa
(b) Vx(Fx A Gx) f- 3x(Fx A Gx)
inference is called Existential Quantifier Introduction or Existential
(c) VxFx f- VxFx
Generalization. The rule is (d) Faa f- 3xFxa
(e) ,.....,3xFxa f- r-.;Fba
EI A (t/v) (f) ,.....,Faa -+ Vx r-.; Fax f- Fab -+ Faa
3vA (g) VX(f".'Fxb -+ Fab), r-.;Fab f- r-.;3xFxb
(h) Fa, Vx(Fx -+ Gx) f- f'.'VX r-.; Gx
As before, A (tlv) is the formula that results when each occurrence of v in 2 If formulas such as V xV xFx and V xFa were admitted as well-formed, what
A is replaced by an occurrence of t. In determining the correct application . restrictions would be needed on the rules to prevent invalid inferences such as
of UE, one's attention goes from the bottom line to the top line, whereas Vx(Fx -+ 3xVyFxy) Fa-+ VxGaa
the reverse is the case with EI. With EI, one must see if t goes in for each Fa -+ 3xVyFay 3x(Fx -+ VxGxx)
occurrence of v in A. Examples of EI would be:
Answers
Fa Faa Faa Fa -+ VxFx I. (c) 1. VxFx A
2. r-.;V xFx A
3xFx 3xFxx 3xFxa 3y(Fy -+ VxFx)
3. (I) A (2) I, 2, Conj
4. "" ~VxFx RAA
Note the third pattern, which is valid and accords perfectly with EI. 5. VxFx 4, DN
Examples of erroneous uses of the rule would be: (f) 1. r-.;Faa -+ 'Ix r-.; Fax A
2. ",(Fab -+ Faa) A
3. ",(~Fab V Faa) 2, Imp
Fa -+ VxFx Fab Fab
4. r-.; r-.;Fab A '"'-'Faa 3, DN,DM
3x(Fx -+ VxFx) 3xFbx 3xFxx 5. r-.;Faa 4, S
6. 'Ix"", Fax 1,5, MP
In the first case 3x(Fx -+ VxFx) is not a wff; also each occurrence of 7. r-.;Fab 6, DE
v(x) in A(Fx -+ VxFx) is not replaced by an occurrence of t(a). In the 8. r-.; r-.;Fab 4, S
second case b was changed to a and we did not merely have b going in for 9. Fab 8,DN
10. Fab A r-.;Fab 9,7, Conj
x in Fbx. In the third case neither a nor b go in for each occurrence of x
II. r-.; ,,-,(Fab -+ Faa) RAA
in Fxx. 12. Fab -+ Faa 12, DN
 Y::E:xistentlaJ Elimination and Universal Introduction 141

, A (tfv) represents our substituting an indivi~ual constant, t, in

2. A (tjv) must be the formula that results when each j1'ee occurrence of v in A lace v occurs in A. Examples of the use of thIS rule are
is replaced by an occurrence of t. eacI1 P
3xFx 3xFxx 3yVxFxy 3xFxb
provided a is not limited
~ Faa VxFxa Fab

6.3 Existential elimination and universal introduction Erroneous uses of the rule are
Both UE and EI are valid without any provisos. However, the rules 3x(Fx -+ Gx) Vy3xFxy
_3xFx 1'f . l'ImI'te d ,
a IS
to remove existential quantifiers and to add universal quantifiers need Fa Fa -+ Gx VyFay
certain restrictions to insure their validity.
To introduce the rule that allows the removal of existential quantifiers If we were allowed to use EE with an individual constant that is
under certain conditions, we first consider this semiexplicit proof: limited, then we could "prove" the invalid pattern below:

3xFx, 3xGx I- 3x(Fx /\ Gx)

(1) Vx(Fx -+ Gx), 3xFx I- 3xGx
l. 3xFx A
l. Vx(Fx -+ Gx) A
2. 3xGx A
2. 3xFx A
3. Fa 1, EE
3. Fa from 2 4. Ga 2, EE (erroneous)
4. Fa-+ Ga 1, DE 5. Fa /\ Ga 3, 4, Conj
5. Ga 3,4, MP
6. 3x(Fx /\ Gx) 5, EI
6. 3xGx 5, EI
Since a is introduced on line 3 by EE, a becomes limited, and conse-
This derivation may be described as follows: Assume that whatever has uently in using EE on line 2 we cannot use a. Another term must be
F has G. Now assume that something has F, and let us assume that a use d , but this would then prevent our obtaining line 5. The idea here is
q
is the thing that has F (hence line 3). If a has F and if, given 1, whatever that if we let a be the individual that is F, given 3xFx, then we can
has F has G, then a must have G and thus there is something that has G. validly move from 3xFx to Fa, but then we cannot move from 3xGx
Such reasoning is sound. Note that in these reflections we take a as being to Ga. We need to introduce another term, which itself becomes limited
one of the things in V which has F given line 2-3xFx. The idea is that and could not then be used if another existential quantifier were
if F is true for some individual, then we can choose some individual, a, eliminated. We resume following the restriction on EE in the next proof.
for which F is true.
If an individual term is used to stand for some chosen individual, 3x3yFxy I- 3y3xFxy
then we will say the term is limited. When we validly move from 3xFx l. 3x3yFxy A
2. 3yFay 1, EE
to Fa, 'a' is limited. It should be noted that if an individual constant is
3. Fab 2, EE
limited it will not follow that the term is used to stand for some chosen 4. 3xFxb 3, EI
individual, for we wish to use the expression 'limitation' to cover this use 5. 3y3xFxy 4, EI
of individual terms and a second and third use to be discussed shortly.
As will become clear later, to say a term is limited is to say that we cannot In using EE we introduce certain constants into the proof. It is
do with it some of the things we do with other individual constants; important to avoid having these constants appear in conclusion lines.
we cannot use it in existential elimination and in universal introduction, To illustrate, note that the following argument patterns are invalid but
the fourth rule to be introduced shortly. could easily be established if we allowed limited constants introduced by
The rule of existential elimination can be stated as follows: EE in conclusion lines.

EE 3vA 3xFx I- Fa
where t is not limited Vx(Fx -+ Gx), 3xFx I- Ga
A (tjv)
"'>/ ••
 14" Natural Deduction System flif:~;~ti~I Elimination and Universal Introduction 143

As was pointed out earlier, in providing an interpretation for a formula in A. For example, the following are correct if we suppose that a
with an individual constant we supply a name for one of the individuals not limited:
in the domain of discourse. This being the case, a counterexample for
3xFx f- Fa-would be: V: positive integers, Fx: x is a prime number, a:4. Fa 3yFya Faa
The last rule permits the introduction of universal quantifiers under v;;FX , Vx3yFyx' VxFxx
certain conditions. Again let us consider a semiexplicit proof:
"IT~,n,,~~T<'1' these are erroneous:
Vx(Fx --+ Gx), Vx(Gx --+ Hx) f- Vx(Fx --+ Hx)
1. Vx(Fx --+ Gx) A Fx--+ Ga 3xFxa Faa
2. Vx(Gx --+ Hx) A Vx(Fx --+ Gx) , Vx3xFxx' VxFxa
3. Fa --+ Ga 1, DE
4. Ga --+ Ha 2, DE the first, a(t) does not go in each place where x(v) is found in A, and in
5. Fa --+ Ha 3,4, Trans the second, (tlv) is violated and Vx3xFxx is not a wff. In the third
6. Vx(Fx --+ Hx) from 5 attern, u(t) occurs in V IvA. As it will turn out, if a term appearing in a
p is not limited, then at that point it stands for an arbitrary element.
When we used UE in the above proof to obtain 3 from 1 and 4 from 2, We It has already been stated that t is limited when it is introduced by
used a. But we could have used any name for an individual in the Consequently, ur cannot be applied to a term with an EE source.
domain of discourse. In effect, a here is an arbitrarily selected element fro illustrate the danger of ignoring the EE source of a term when using
of the domain. Suppose, for example, there are aI' a 2 , • • • an individuals consider the erroneous "proof" below:
in the universe of discourse. From (1) by UE we can obtain
1. 3xFx A
2. Fa 1, EE
3. VxFx 2, DI (erroneous)
and from 2 by UE we could obtain
A somewhat more sophisticated violation of the proviso on ur is found
in this next example.

1. 3xFx A
which in this V means that Vx(Fx --+ Hx). In general, when u is intro-
2. Fa 1, EE
duced by UE and it is proven that a has some property F, then it A
r~' r--.;Fa
validly follows that VxFx, since in this context we could have proven Li: (2) 1\ (3) 2,3, Conj
that each individual in V or any arbitrarily selected individual in V has }!', 5. r--.;r--.;Fa RAA
The last quantification rule allows us to go from A (tlv) to VvA and is 6. Fa 5,DN
commonly known as Universal Generalization or Universal Quanti- 7. VxFx 6, DI (erroneous)
fication Introduction. To secure the validity of such an inference the
Introducing an individual constant by EE is not the only way in
rule must be stated with two provisos:
which a term can become limited; there are two and only two other ways.
The second way a term can become limited is by appearing in an assump-
DI A(t/v)
where t is not limited and does not occur in VvA tion line. But when the assumption is discharged, it should be stressed,
VvA
the term is no longer limited unless it is limited in virtue of being
introduced by EE (see the last proof). To illustrate, consider the example
Again, the idea of this rule is that if A is true for t and t is not limited, and
below.
it stands for an arbitrary element of the domain, then A is true for every
element of the domain. Again, A (tlv) indicates that the individual 1. Fa A
constant goes in each place where the individual variable in question is 2. VxFx 1, DI (erroneous)
j.,
 Natural Deduction System "_'"'",..",,. Elimination and Universal Introduction 145

In this example, a is limited since it occurs in an undischarged assumption a is not introduced by EE, nor does it appear in an undis-
line. However, in the next two proofs a is not limited when VI is applied assumption both b and a become limited when EE is applied.
chtLrge d '
since the assumption line in which a makes its appearance has been b is limited follows from the use of EE. That a becomes limited is
discharged. erhtLPS surprising until we see that Faa would not follow from
~:dyFXY and 3yVxFxy does not follo~ ~rom VX3yFxy. To prev~nt su~h
Prove: ,.....,3xFx f- "Ix,....., Fx

. 1. ,.....,3xFx
Fa
A
A
£:
iULel'eI
quant 1
lces a term which occurs withm the scope of an eXIstentIal
, . . .
.fier and which is later removed by EE wIll become lImIted. A

U
.
3. 3xFx 2, Er good rule to follow with resp~ct to applying EE, by the way, IS not to use
4. (1) A (3) 1, 3, Conj a t which previously occurs m the sequence.
5. ,.....,Fa RAA The converse of the above is a valid argument pattern and can easily
6. Vx,.....,Fx 5, Dr
be proven.
Prove: Vy(Fy ->- 3xFx) 3yVxFxy f- Vx3yFxy
I~. Fa A 1. 3yVxFxy A
L! 3xFx 2, Er 2. VxFxa 1, EE
3. Fa ->- 3xFx RCP 3. Fba 2, DE
4. Vy(Fy ->- 3xFx) 3, Dr 4. 3yFby 3, Er
5. Vx3yFxy 4, Dr
Note that in this proof "Ix cannot be used in VI on line 3 since
Vx(Fx ->- 3xFx) is not a wff. It is worthwhile to note that Note that the b in line 4 is not limited.
The proviso on both EE and VI is that t not be limited. Let us
Fa f- VxFx
conclude this section with a summary of how an individual constant, t,
is not a valid pattern. Given this interpretation: V: positive integers, can become limited. t can become limited iff either
Fx: x is even, and a: 2, we obtain a counterexample. It should also be (1) t is introduced by EE,
remembered that since a term is limited in an undischarged assumption (2) t appears in an undischarged assumption line, or
line, this term cannot be used in EE unless the assumption has been (3) t appears within the scope of an existential quantifier later
discharged. For from 3vA we can obtain A (tjv) only if t is not limited. removed by EE.
Failure to keep this in mind might result in the following unfortunate
sequence: Exercises
1. Find and explain the errors in the following items.
1. Fa A A
(a) 1. VxVyFxyx
2. 3xGx A 1, DE
2. VyFxya
3. Ga 2, EE (erroneous) 2, DE
3. Fxba
4. Fa A Ga 1,3, Conj 3, Dr
4. VxFxba
5. 3x(Fx A Gx) 4, Er 4, Dr
5. VyVxFxyy
(b) 1. ,.....,VxFx A
The reader can easily demonstrate the invalidity of Fa, 3xGx f-
2. ""Fa 1, DE
3x(Fx A Gx). 3. ,.....,Fa V Ga 2, Add
The third and final way in which a term can become limited is by 4. ""Fa vVxGx 3, Dr
appearing in a line that is introduced by EE. To illustrate: 5. Vx,-...., Fx V VxGx 4, Dr
(e) 1. Fab --+ VxGax A
1. Yx3yFxy A 2. Vx(Fxb ->- VxGxx) Dr
2. 3yFay 1, DE (d)r1. Fm A
3. Fab 2, EE lb VxFx 1, Dr
4. VxFxb 3, Dr (erroneous) 3. Fm --+ VxFx RCP
5. 3yVxFxy 4, Er 4. VyFy ->- VxFx 3, Dr

-"
 146
'''.n.f",!!'les in Using Quantification Rules 147
(e) l. 3xGxa A f) VxFx V VxGx f- Yx(Fx V Gx)
2. Gaa
3. 3xGax
1, EE
2, EI
~g)VxFx A VxGx f- Yx(Fx A Gx)
4. 3y3xGyx 3, EI Using EE and EI, construct proofs for the following patterns.
(f) l. YxFx A (a) Fab - 3xFxb, Fab f- 3yFyb
2. Fa 1, VE (b) 3x(Fx - Fa), ""Fa f- 3x "" Fx
3. Fb 1, VE (c) r-I3xFxb --+ Fab, ""Fab f- 3y3xFyx
4. Fa A Fb 2,3, Conj (d) 3x{Fx V Fx) f- 3xFx
5. Yx(Fx A Fx) 4, VI (e) 3xFx f- 3x3y(Fx A Fy)
(g) l. YxYyFxy A
2. Faa 1, VE (2) Answers
3. 3xFxa 2, EI (f) Only the use of VI is erroneous since a{t) does not go in for each occur-
4. Fba 3,EE rence of x(v).
5. YxFbx 4, VI (g) Only the use of VI is erroneous since a becomes limited on line 4.
6. 3yYxFyx 5, EI
3. In Section 6.6 see T2 for (g).
2. Numerals are proper names for numbers and thus are singular terms. Rath (d) l. Vx(Fx --+ Gx) A
than let an individual constant stand for a numeral, we will leave the numera~r
as they are. Find and explain the errors in the following.
(a)ul. 1 >>
2. 3x(x 0 0) A
1, EI
s li: :::"G:Gx
4.
5.
Fa- Ga
r-IFa
A
2, VE
1, VE
3,4,MT
3. a> 0 2, EE 6. 'Ix"", Fx 5, VI
4. Yx(x > 0) 3, VI 7. 'Ix r-J Gx - Yx 1"'0.1 Fx RCP
5. (b > 0) 4, VE A
4. (e) 1. 3xFx
6. Yx(x> 0) 5, VI 2. Fa 1, EE
7. (1) - (6) RCP 3. FaA Fa 2, Idem
(b) l. "",(a> 3) A 4. 3x3y(Fx A Fy) 3, EI (2)
2.4> 3 A
3. 3x(x > 3) 2, EI
4. a> 3 3, EE
5. P A "",P - Q SC theorem 6.4 Strategies in using quantification rules
6. (1) A (4) - (7) 5, substitution
7. 0 > 1 1, 4, 16, MP There are several strategies that should allow the construction of proofs
8. "",(a> 3) _ 0 > 1 RCP for most predicate argument patterns. We review some of these
9. "",(0 > 1) A
strategies at this time.
10. a > 3 8, 9, MT
II. Yx(x > 3) 10, VI STRATEGY ONE. If the conclusion is of a conditional forni, A --+ B,
(c) l. 2> 1 A then assume antecedents for an RCP proof, as was done in SC. This
2. 3x(x> 1) EI strategy was used on examples and in some of the exercises. To illus-
3. a> 1 2, EE
trate the strategy with an only slightly more complicated proof,
4. a > 1 - a > 0 A
5. a > 0 3, 4, MP consider:
6. Yx(x> 0) 5, VI Yx(Fx _ Gx) f- 3xFx - 3xGx
7. (1) - «4) - (6» RCP (2) l. Yx(Fx _ Gx) A
3. Vsing VG and VI, construct proofs for the following. 2.3XFX A
(a) YxFx f- YyFy 3. Fa 2, EE

Q
(b) P A YxFx f- Yx(P A Fx) 4. Fa _ Ga 1, VE
(c) Yx(Fx _ Gx) f- YxFx _ YxGx
5. Ga 3,4, MP
(d) Yx(Fx - Gx) f- Yx "'" Gx _ Yx"" Fx 6.3xGx 5,EI
(e) Yx(Fx ~ Gx) f- YxFx ~ YxGx 7. 3xFx - 3xGx RCP
~I·,
 149
• ' Using Quantification Rules
Natural Deduction System Po strategieS III
then a is not limited in ",-,Pa, so we can use UI.
In general, any of the strategies used in constructing proofs in SO discharged, F can con]' oin this to line 1 and use a final
can be used in PC. For example, Dis can often be used when the inner sed Dr on "'-' a, we . .
u . 1 fl' 2 Let us continue wIth thIS strategy.
structure of a quantified formula is of the form A V B. To illustrate, to get the dema 0 me .
consider:
' Fa A

G:
~
4 3xFx 2, EI
3x(P V Fx) I- P V 3xFx 2,4, Conj
(2) /\ (4)
1. 3x(P V Fx) A RAA
6. ",-,Fa
2. P V Fa 2, EE 6, UI
7. Vx"'-' Fx
.P A 7,1, Conj
8. (7) /\ (1)
4. P V 3xFx 3, Add RAA

U
9. "'-' ,-..;3xFx
5. Fa A 9,DN
10. 3xFx
6. 3xFx 5, EI
7. P V 3xFx 6, Add . hen the conclusion is of the form VxA or 3xA we. may find
8. P V 3xFx Dis sometImes w . . RAA strategy. The next proof Illustrates
. usef u1 to deny A m usmg an
Jt
STRATEGY TWO. Sometimes the conclusion will have the form strategy.
VxA or 3xA, with A having the form B ->- O. Often B with t going in VxFx A 3xGx I- 3x(Fx A Gx)
for v can be taken as an additional assumption in an RCP proof. To 1. VxFx /\ 3xGx A
illustrate: 2. 3xGx 1, S
2, EE

.
3 Ga
Vx(Fx --+ Gx) I- 3x(Fx ->- Gx) 4' VxFx 1, S
1. Vx(Fx ->- Gx) A 5: Fa 4, UE
. Fa A ro-'(Fa /\ Ga) A

[}
3. Fa ->- Ga
4. Ga
5. Fa --+ Ga
6. 3x(Fx --+ Gx)
1, UE
2,3, MP
RCP
5, EI
3xFx --+ P I- Vx(Fx --+ P)
U 7 ",-,Fa V r-JGa
8' ",-,Ga
9: Ga /\ r-JGa
10. r-J "",,(Fa /\ Ga)
11. Fa A Ga
6, DM
5,7, DS
3, 8, Conj
RAA
10, DN

.1. 3xFx ->- P

Fa
A
A
12. 3x(Fx A Gx) 11, EI

[1 3. 3xFx
4. P
5. Fa --+ P
2, EI
1,3, MP
RCP
Exercises
1. Usmg
.
. UG UI EG and EI construct proofs for the followmg: (A 11- B
, " 'd b th 'ays)
means the derivation is purporte to go 0 " .
6. Vx(Fx --+ P) 5, UI
(a) 3xFx V 3xGx I- 3x(Fx V Gx)
(b) Vx""" Fx I- ,-.....,3xFx
Note that with Fa (line 2) discharged in both proofs, a is no longer (0) 3x r-J Fx 11- ,--,VxFx
limited. (d) Vx(Fx V Gx) I- VxFx V 3xGx
STRATEGY THREE. RAA strategies can be employed in interesting (e) VxFx V 3xGx I- 3x(Fx V Gx)
(f) ......,3x,-....; Fx 11- VxFx
ways. An interesting example appears below. (g) Vx3y(Fxy --+ Gxy) I- 3xVyFxy --+ 3x3yFxy
(h) Vx(Fx ->- P) 11- 3xFx ->- P
,.....,Vx,....., Fx I- 3xFx (i)vy(Fy -+ Gy) I- Vx(3y(Fx A Fxy) ->- 3y(Gy A Fxy»
1. ""Vx",-, Fx A
. the four rules then the
,2. ,....,,3xFx A 2. If individual variables were taken as terms m tl stat~d'
inferenoes below would be justified by the rules as presen Y .
One possibility is to introduce an assumption Fa that will lead to a Vx3yFxy 3yVxFxy
contradiction so that we can get f'..IFa with Fa discharged. If Pa has 3yFyy VxFxx
 151
150 Natural Deduction System Pc System PC

and yet these inferences are invalid (demonstrate this). The trouble is that have been working with, formulas such as
when t goes in, it becomes accidentally bound. Why can't this happen with
individual constants? What proviso must be added to which rules to prevent Fx--+ 3xFx
the rules from being invalid if individual variables are allowed as instances Fx --+ 3xFx~ Fa --+ 3xFx
of t?
woU Jd b e well-formed
'
and they are valid formulas. However, though
. .
Answers ~e could derive, say, Fx --+ Fx, we could de1'lve neIther of the above
1. (h) only individual constants are instances of t in the four rules. We

·
1-1. Vx(Fx --+ P)
3XFX
3. Fa
A
A
2,EE
could a11ow I
'ndividual variables as values of t, but then the rules would
the adjustment discussed in exercise 6.4.2. We wish to keep the
as simple as possible. So the decision is. made to keep the r~les as

U 4. Fa --+ P
5. P
6. 3xFx--+ P
1, VE
3,4,MP
ROP
are. Not a great deal is lost in not havmg such theorems smce as
biconditional suggests, whenever we have a valid forn: ula with a

. -no 3xFx --+ P

Fa A
A individual variable this is equivalent to a formula WIth the free
individual variables uniformly replaced by an individual constant (and
[i 3. 3xFx
4. P
5. Fa --+ P
2, EI
1,3,MP
ROP
versa). Nevertheless the PC with the previously discussed for-
lUation rules could not be complete. . .
6. Vx(Fx --+ P) 5, VI To provide for the possibility of PC's completeness we w~ll SImply
not count formulas with free individual variables as wffs. Agam no loss
2. Proviso added to VE and EE: t must be free for V. t is free for v iff ,vhen results since the equivalence described just now obtains. .
t/v, t does not become bound by a quantifier in A. The rules of inference are those for SO plus the followmg, where,
again, A(tfv) is a formula that results when each occurrence of v in A
is replaced by an occurrence of t:
6.5 System PC
VvA 3vA
VE EE where t is not limited
A (t/v) A (t/v)
We turn now to the construction of the system that will have valid
predicate formulas as theorems. This system, PO, will be in part made A (t/v) A (t/v) where t is not limited and does
EI VI not occur in V vA
up of the sentential calculus, specifically system SO. Thus the system 3vA VvA
will have theorems that are tautologies and theorems that are valid
predicate formulas. These are the only kinds of theorems the system will We should note that the metavariables in the ten rules taken from
have. In addition, all tautologies will be theorems of the system and all SO are now variables for any well-formed formula in PC. This fact
valid predicate wff of the system will be theorems. Since PC will be allows us to make tautological transformations and tautological in-
built on SC, we count among the symbols of the system all those of SC ferences on basic components of formulas.
plus The definition for the biconditional is also carried over from SO.
A proof in PC has the same description as a proof in SC, with the
1. Predicate variables: F, G, H, ... . rider that a theorem line cannot contain a limited term introduced by
2. Individual variables: x, y, z, ... . EE (see Section 6.3). We recall that a theorem is obtained when all
3. Individual constants: a, b, c, ... .
4. Quantifiers: V, 3 assulUptions have been discharged.

It will be advisable to alter slightly the formation rules in setting up Exercises

PC. We wish PC to be complete. This means we wish all valid well- 1. What changes need to be made to the primitive basis of PO if V is taken as
formed formulas to be theorems of PC. Now given the formation rules the only primitive quantifier?
 Natural Deduction System p() 153
2. List some other valid formulas containing predicate expressions that ar r T4 ,,-..,3xFx I- "Ix """ Fx
and not theorems of PC. What is the counterpart of each where indiev:da ld T5 r--.JVx""" Fx I- 3xFx
constants are used? I uaJ

3. Set down the formation rules for PC. ~~.

r;: :~:]i':
"Ix""" Fx
Ji'x !2, TI (T4)
Answers Li (1) 1\ (3) 1, 3, Conj
5. "" ,......,3xFx RAA
2. For example VxFx ->- Fx and VxFx ->- Fa 6. 3xFx 5, DN
7. T5 RCP

Note that TI can be eliminated from this proof and the steps of the proof
6.6 Theorems of PC for T4 can be inserted. (T5 can also be proven using a different strategy
,vithout T1 or T4-see Section 6.4.) Such eliminability is a necessary
In this section we will prove a few theorems, introduce the two d . feature for a derived rule such as TI. Also note that we have again
rules of PO, and list some major theorems. e1'lved explicitly collapsed introducing the theorem (with or without substi-
tution) and using MP into one step, and we have justified this with TI
T1 Y,x(Fx 1\ Gx) ->- VxFx 1\ VxGx
(T4).
1. Vx(Fx 1\ Gx) A
The Rule of Replacement (R) from SC may be applied to welI-
2. Fa 1\ Ga 1, VE
3. Fa 2, S formed formulas in PO. As R is a derived rule in SO, so it is a derived
4. VxFx 3, VI rule in PC. For example, having proven T3 we can now distribute
5. Ga 2, S quantifiers within quantified conjunctions and justify this by appealing
6. VxGx 5, VI to R, or we can omit using R or T3 (with or without substitution) and
7. VxFx 1\ VxGx 4,6, Conj MP and go through the laborious process of recreating an instance of the
8. T1 RCP proofs of T1, T2, and T3. In proofs it is sometimes convenient and time
T2 saving to use R. For example, having proved
\f.xFx 1\ VxGx ->- Vx(Fx 1\ Gx)
1. VxFx 1\ VxGx A
2. 'VxFx 1, S
3. VxGx 1, S
4. Fa 2, VE we could make the following move:
5. Ga 3, VE
6. Fa 1\ Ga 4,5, Conj
7. Vx(Fx 1\ Gx) 1. 3x(Fx ->- Gx)
6, VI
8. T2 2. """"Ix"",, (Fx ->- Gx) 1, R (T6)
RCP

The first derived rule will be a carry-over of Theorem Introduction In this case Fx ->- Gx is a substitution for Fx in T6. The move from 1
(TI) from SO: Now, however, we understand theorems to include not to 2 above could also be made by TI and MP. However, if 1 appeared
only tautologIes .but all theorems of PO or any instances of them. The as part of a longer formula, R would have to be used. Often when TI or
next two proofs Illustrate the use of TI. R are used, substitutions are made in formulas. The rule that makes
the notion of substitution effective in PC is a difficult rule to state.
T3 Vx(Fx 1\ Gx)~ VxFx 1\ VxGx Since.the student intuitively substitutes correctly when he uses R or TI
1. T1 TI with substitution, and since both Rand TI are luxury rules, things we
2. T2 TI can get along without, we will not set down the rule for substitution.
3. T1 1\ T2 1, 2, ConJ. Also in constructing proofs for formulas, normally the only occasion we
4. T3 df have for using TI or R is to change existential quantifiers into universal
""f.,
 Natural Deduction System nUf,lanIBSS, Consistency, and Completeness of PC 155
quantifiers (and -:ice versa) in order to drop the quantifier. In other Prove the following theorems.
words, TI or R WIll normally be used with T6 plus the theorems below. (a) Vy(YxFx -- Fy)
(b) 3y(3xFx -- Fy)
T7 3x,...."",Fx~,...."",YxFx (c) Vy(Fy -- 3xFx)
T8 Yx,...."", Fx ~ """""3xFx (d) 3y{Fy -- YxFx)
T9 ,...."",3x,...."", Fx ~ YxFx
Establish the analyticity of those statements that are analytic in exercise
5.7.3.
TI-T9 a~ong with the theorems below make up the major theorems
the predICate calculus. of
Answers
1. f"'oooJ3y(Fy -- YxFx) A
TIO YxFx--* Fa 2. Yy,....., (Fy -- YxFx) I, R (T8)
TIl YxFx --* 3xFx 3. f"'oooJ(Fa -» YxFx) 2, UE
TI2 YxFx~YyFy 4. ,.....,(,.....,Fa V YxFx) 3, Imp
TI3 YxYyFxy~ YyYxFxy 5. f"'oooJ ,.....,Fa /\ ,.....,YxFx 4,DM
TI4 Fa--* 3xFx 6. f"'oooJ,-."JFa 5, S
TI5 3xFx~3yFy 7. Fa 6,DN
TI6 3x3yFxy~ 3y3xFxy 8. YxFx 7, UI
TI7 3xYyFxy --* Yy3xFxy 9. YxFx /\ ,.....,YxFx 5,8, S, Conj
TI8 Yx(Fx --* Gx) --* (YxFx --* YxGx) 10. ,.......,,.....,3y(Fy-- YxFx) RAA
TI9 Yx(Fx~ Gx) --* (3xFx~ 3xGx) II. Td 10, DN
T20 Yx(Fx~ Gx) --* (YxFx ~ YxGx)
T21 Yx(Fx --* Gx) --* (3xFx --* 3xGx)
T22 (YxFx --* YxGx) --* 3x(Fx --* Gx)
T23 Yx3y(Fxy --* Gxy) --* (3xYyFxy --* 3x3yGxy) 6.7 Soundness, consistency, and completeness of PC
T24 YxFx /\ 3xGx --* 3x(Fx /\ Gx)
T25 Yx(Fx V Gx) --* YxFx V 3xGx
T26 If one of the uses of PC is to establish predicate formulas as valid, then
YxFx V YxGx --* Yx(Fx V Gx)
T27 YxFx V 3xGx --* 3x(Fx V Gx) PC must be consistent; otherwise it could be used to establish any
T28 3x(Fx --* Gx)~ YxFx --* 3xGx predicate formula. Also the formulas it establishes as theorems must be
T29 (3xFx --* YxGx) --* Yx(Fx --* Gx) valid if PC is to have this function. In other words, PC must be sound.
T30 3x(Fx /\ Gx) --* 3xFx /\ 3xGx Finally, we wish all valid predicate formulas to be theorems of PC;
T31 3x(Fx V Gx)~ 3xFx V 3xGx that is, we wish PC to be complete. If PC is complete, then we can say
T32 Yx(P /\ Fx)~ P /\ YxFx
a predicate formula is valid only if it is a theorem of PC.
T33 Yx(P V Fx)~ P V YxFx
T34 3x(P /\ Fx)~ P /\ 3xFx To prove the soundness of PC one must first prove that each of the
T35 3x(P V Fx)~ P V 3xFx four quantification rules is valid. The next step is to show that in a
T36 3x(P --* Fx)~ P --* 3xFx proof each line is a valid consequence of its assumptions. If A is a valid
T37 Yx(Fx --* P)~ 3xFx --* P consequence of no assumptions, A is valid. Since theorems in PC are
T38 3x(Fx --* P)~ YxFx --* P assumption free lines, all the theorems would be valid.
If one proves PC sound, then it follows that it is negation consistent
Exercises since no theorem can have the form A /\ ,.....,A. However, one can prove
PC negation consistent without relying on a proof of the soundness of
l. Using .the derived rules as needed, construct proofs for the unproven theorems PC. Let us outline one such proof.
fol~owmg the numbered order. It is permissible to pass up any theorem for To each formula A we assign a formula A', called an associated
whI~h the corres~onding argument pattern has been proven in the previous sentential formula (asf). We obtain the asf of a quantified formula by
sectIOns or prevIOus exercises.
simply deleting all the quantifiers, individual variables, and individual
 156 Natural Deduction System PC 157
):Intllldne.;~, Consistency, and Completeness of PC

constants, leaving everything else, including all the "-' signs, and PO is not only consistent but is sound and complete. We will
substituting sentential variables for the predicate variables. Thus, the to the question of the soundness of PO in Part II. The proof for
asfs of
completeness of PO or of any predicate calculus is now generally
'zed as the demarcation point between beginning and advanced
VxFx 3xFx, rccog nI
--? "-'Vx3y(Fx V Gy)+-+ VxFx V 3yGy . and will not be attempted in this book.
One final note. Earlier it was pointed out that there is no mechanical
are, respectively, .dCCISl
. 'on procedure for determining whether any predicate formula is
or not. This means that we could not program a computing
P--? P, ,.....,(P V Q)+-+ (P V Q) achine such that it would always give the answer 'yes' if the formula is
m and the answer 'no' if the formula is invalid. However, we could
We note that. all the theorems proven or to be proven in this chapter use a system like PO to program a machine that would give a 'yes'
have tautologiCal asfs. All the formulas of SO are already in transformed answer if the formula is valid. We could program a machine that would
form, and all theorems of SO are tautologies. If it can be shown that all construct a proof for all well-formed predicate formulas that are valid.
theorems have tautological asfs, then A /\ ,,-,A cannot be a theorem However, with respect to some formulas the machine, no matter how it
and PO is negation consistent. All the theorems of PO would have is programmed, would never give a 'yes' or a 'no' answer. In effect, the
tautological asfs if the rules will only allow theorems that have tauto- machine would be inadequate with respect to 'no'. Some formulas that
logical asfs. If we consider each predicate rule, we see each one has the are invalid would be greeted with no answer at all by the machine. Logic,
asf form in a word, can never be fully mechanized. This important truth was
first noted (in 1936) by the American logician Alonzo Ohurch, when
A' he proved that there is no mechanical decision procedure for predicate
A' logic. Human intuition-providing counterexamples-will thus always
be necessary for recognizing invalidity ..
Any asf derived by such rules must be a tautological consequence of
the asf assumptions. If an asf line derived from previous asf lines by the
rules of PO is a tautological consequence of its asf assumptions then Exercises
since theorems are assumption free lines, each theorem must have ~
tautological asf. I. Give the asf of the first ten theorems of PO and demonstrate that they are
If PO is consistent, this does not imply that all the theorems of PO tautologies.
are valid-that PO is sound. To show this we need but note that the
2. Would proving PO complete imply that PO is sound? Why?
asf form of the consistency proof would work even if the restrictions
were removed from EE and Ur. PO with unrestricted quantifications 3. Discuss how one might go about proving that each of the quantification rules
rules can be proven consisten t. However if the restrictions were removed , is valid.
invalid formulas such as those below would be theorems.
4. Establish the validity of each of the following arguments that are valid using
any of the theorems or derived rules of PO. If the argument is invalid,
Fa --? VxFx
establish the invalidity of its argument pattern by providing a counter-
3xFx --? V xFx example.
3xFx /\ 3xGx --? 3x(Fx /\ Gx)
Vx3yFxy --? 3yVxFxy (a) No student who is a member of YSA is dishonest. Some students are
dishonest. Therefore, not all students are members of YSA.
The asf of each of these formulas is a tautology, we should note. Oonse- (b) Some students are athletes. Some students like all athletes. So some
students are not disliked by all students and some athletes are liked by
quently, eliminating provisos on the quantification rules does not
some students.
affect the consistency of the system, but it does affect the soundness. (c) Students are animals. So students' tales are animals' tales.
 Consistency, and Completeness of PC 159
Natural Deduction System

(d) Converse of (c). To take another example:

(e) Some students are radicals. Some faculty members do not like 'v'xFx --+ P f- 3x(Fx --+ P)
any radicals. Therefore some students are not liked by all faculty 'v'xFx --+ P
members. ,.....,3x(Fx --+ P)
(f) Anyone who can please any student can please anyone any student can ~
So if Homer cannot please Judy and Judy can be pleased by some student' ,...,'v'xFx P
then Homer cannot please all students. ' 3x ,....., Fx 'v'x "-' (Fx --+ P)
,.....,Fa ,.....,(Fa --+ P)
(g) If anyone pleases anyone, then someone dislikes both of them. No on
dislikes anyone unless he knows them. Judy pleases Homer. Therefore 'v'x,....., (Fx --+ P) Fa
Judy is disliked by someone who knows her and Homer. e ,.....,(Fa-+ P) ,.....,P
Fa x
5. In the exercises for 4.8 the tree method for sentential logic was described. ,.....,P
Th~s ~ethod ca~ be expanded so as to 0 b~ain a techniqu? for establishing the
valIdIty of predIcate formulas. To do thIS we need to mtroduce two rules. . cXh path does close in this finished tree, the argument is valid. Note
Smce ea . .
n go on forever-and thus are mvalId-for example, the tree
The first relates to dropping universal quantifiers when we have formulas of some t re es ca .
the form 'v'vA and can be stated as follows: for Vy3xFxy (EE introduces a new name, thus UE demands we put I?-
UE Given an open path with a sentence of the form 'v'vA for another 3xFxt where t is the new name, and so on). Thus the tree method If
each individual constant, t, that appears anywhere in the path programmed on a computer would have the result that for some sets ~f
write a sentence of the form A (tfv). If no t appears in th~ J! ulas the machine would give no answer. The method, however, IS
path, choose a term. form lete' thus if the set of entered formuI as .IS ,mconsIsten,
. t th e mach'me
The rule for dropping existential quantifiers is: ::rd gi;e a 'yes' answer. Remember.: ~ is inc.onsistent iff all its ~aths ?lo~e
EE Given an open path with a sentence of the form 3vA, write on a tree. P f- Q is valid iff P A ""Q IS mconsIstent, and P --+ Q IS valId Iff
a sentence of the form A (tfv) where t is an individual constant ,...,(P -+ Q) is inconsistent.
that does not previously appear in the path. (a) Use the tree method to establish the validity of some argument patterns
When a formula has either the form ,,-,'v'vA or the form ,,-,3vA, we remove in this chapter.
the denial sign by employing (b) Use the tree method to establish the validity of some theorems in PC.
,,-,'v'v A +-+ 3 v "-' A (c) When a finished tree has an open path, this provides us with a counter-
"-'3vA +-~'iv "-' A example. Read off some counterexamples from trees for invalid argu-
as if they were rules. We also remove double denials whenever they may ments.
occur. The EE rule must be used before the UE rule for the method to work (d) Why does this method work? .
adequately in all cases. To illustrate, examine the following: (e) Use the tree method to establish the validity of the valid arguments m
'v'x(Fx --+ Gx), 3xFx f- 3xGx exercise 6.7.4.
1. 'v'x(Fx --+ Gx)
2. 3xFx Answers
3. "-'3xGx
4. 'v' x "-' Gx 2. No. A negation inconsistent logic system is necessarily complete. PC also can
5. Fa be complete but have invalid formulas as theorems.
6. Fa --+ Ga
7. ,,-,Ga
~
8. ,,-,Fa Ga
x x
In using the tree method to establish an argument pattern we continue to
enter the premises-lines 1 and 2-and the denial of the conclusion-line 3.
Line 4 comes from 3 by using one of the denial equivalences. 5 comes from 2
by EE. 6 comes from 1 by UE. Since only 'a' appears in the path we write
Fa --+ Ga. If 'b' had also occurred, then we would write Fb --+ Gb under
Fa --+ Ga. 7 comes from 4 by using UE, and 8 comes from 6. An argument is
valid iff each path closes. The x's indicate that both paths close, establishing
the validity of the argument pattern.

~I·.
 161

Let us suppose someone said:

(1) The figure in the upper left hand corner of the diagram is the
figure to the left of a triangle with a circle below it.
(2) The figure in the upper left hand corner of the diagram is the
figure in the lower right hand corner.
predicate To express that a loves b or that a
caus~s b, we can Use the two-place (1) identity would be expressed, whereas in (2) identity of type would
predICates Lxy: x loves y, and
logic Oxy: x causes y. To express that
he expressed. It is the use of 'is' in (1) that is our interest. In (1) we
have two expressions, called definite descriptions, that can be said to
is identical with b we could Use Ix ~
with identity ~ is i~entical with y, or the sign f~l: name the same thing, but in (2) we have two definite descriptions that
name two different things that are exactly alike. To say line segment AB
Iden~Ity, =. x = Y is a two-place
predIcate constant just like Lxy is congruent with line segment DO would be an example of this second
and Oxy. notion of 'identical' or 'is'. These two uses of 'is' could be displayed by
The identity relation has some using the names given to the squares as follows:
rather special properties that will
be discussed in this chapter. This a is a (identity)
a is b (identity of type)
relation is also useful in expressing
how many things there are that
meet a certain description. This will
If a happened to have two names-a! and a 2-then identity would be
be looked into. In addition, in this expressed by saying
chapter, the system of natural
deduction for predicate logic PC
will be expanded so that the~rern~
Identity thus is expressed in saying
containing = can be proven. But
first we consider the interpretation Lewis Carroll is Charles Dodgson.
for =.
where we have two names for the same person.
7. 1 Identity The symbol = will be used for the identity predicate. We may
interpret = as follows: a = b is true iff a and b are the same. What
~he notion of identity, which is of particular interest in logic, can be needs to be kept in mind is that the identity sign is used to express that
mtroduced by considering the following diagram. two singular terms are names for one and the same thing. It is to be
recalled from an earlier discussion (Section 5.1) that a singular term is any
expression that names or is used to pick out an individual whether or not

OD
that individual exists in the universe of discourse. Proper names,
demonstrative expressions, definite descriptions-expressions that take
the form 'the so-and-so'-and abstract nouns are all used to pick out
individuals and are singular terms.

""/ ..
00 Exercises
1. Distinguish between the 'is' of one-place predication, the 'is' of identity, and
the 'is' of type identity in the following:
(a) Beauty is truth.
(b) The worst president in U.S. history is yet to be born.

160
 163
(c) Arizona is a state.
(d) Aristotle is a philosopher who taught Alexander. A relation may be such that whenever it holds between one individual
(e) 2 + 2 is 4. and another in some domain it does not hold between the second and the
(f) These exercises are easy. first. For example, 'is the uncle of', 'is taller than', and '>' are likethis.
(g) This car is the car they made back in 1930. A relation that has this property is said to be asymmetrical. Thus a
(h) This is the dress you wore at the party. relation R is asymmetrical in V iff
(i) The author of The Sound and the Fury is Faulkner.
(j) The only worthwhile thing in life is love. VxVy(Rxy -+ ~Ryx)
2. Is identity the same as equivalence?
A wodd in which no one loved an individual in return would give us a
Answers domain in which 'loves' would be asymmetrical. In mathematics the
1. (a) predication, (b) predication, (c) predication, (d) identity (e) ident"t relation'S;;' (being less than or equal to) is such that if x and yare
(f) predication, (g) can be understood to be type identity, (h) 'can be un~ y~ distinct objects and x S;; y, then ,,-,(y S;; x). We may speak of such a
stood to be type identity, (i) identity, and (j) identity. er relation as being. antisymmetric. R is antisymmetric is some V iff
2. No. Equivalence is a relation between statements, whereas identity is a
relation between names. VxVy(x =F y II Rxy -+ f".IRyx)

It is interesting to note that identity is antisymmetric since VxVy(x =F

y II x = Y -+ y =F x). A relation can thus be symmetrical and anti-
7.2 Some properties of two-place predicates
symmetrical.
All relations are either symmetrical, asymmetrical, or neither. If a
There are several important properties of relations. A relation may be
relation has this last property, it is said to be nonsymmetrical. A relation
such that in some given domain of individuals, some V, whenever it holds
R is nonsymmetrical in some V iff:
between one individual x and another individual y, it also holds between
y and x. For example, the relation 'next to' has this property in the 3x3y(Rxy II Ryx) II 3x3y(Rxy II ,,-,Ryx)
domain of natural objects. If, for example, object a is next to object b,
then b is next to a. A relation that has this property is said to be a An example of a nonsymmetrical relation in propositional logic would be
symmetrical relation. In general, a relation R is symmetriwl in some 'implies'. There are some propositions such that if x implies y then y
V iff: implies x, but there are others such that x implies y but y does not
VxVy(Rxy -+ Ryx) imply x. 'Brother of' is also nonsymmetrical in our world, as is 'loves'.
A relation may be such that whenever it holds between x and y and
Other examples of symmetrical relations would be 'conjoined to' (in between y and z it holds between x and z. Such a relation is said to be
domain of woodworking) and 'sibling of' (in domain of historical persons). transitive. That is, a relation R is transitive in some V iff:
In the latter domain we may note that 'brother of' is not symmetrical.
If x is the brother of y, then it does not follow that y is the brother of VxVyVz(Rxy II Ryz -+ Rxz)
x-y may be a sister. However, if we selected as our domain male
persons, then 'brother of' would be symmetrical. Similarly, in some 'Implies', 'parallel to', '>', 'stronger than', and 'subset' are all transitive.
people's ideal world 'loves' would be symmetrical, though as almost all Identity is transitive in any universe of discourse.
discussions of symmetry point out, love, tragically enough, is not On the other hand, the relation 'is the mother of' is not transitive.
symmetrical in our world. In any domain, however, identity is sym- In fact, 'is the mother of' is such that in any domain if Mxy and Myz,
metrical. In other words, in any universe of discourse then ~Mxz. If Mxy and Myz, then x is the grandmother of z. The
relation 'is the mother of' is said to be intransitive. A relation R is
VxVy(x = y-+y = x) intmnsitive in some V iff
is true.
VxVyVz(Rxy II Ryz -+ ~Rxz)
 operties of Two-Place Predicates
165
Some P r

There may be some relations in a given V that are neither transitive may do this as follows:
nor intransitive. They are said to be non transitive. A relation R is
1. 'v'x'v'y'v'z(Fxy A Fyz -+ '"'-'Fxz) A
nontran8itive in a given V iff 1, DE (3)
2. Faa A Faa ---+ r--JFaa
3. Faa ---+ '"'-'Faa 2, Idem
3x3y3z(Rxy A Ryz A Rxz) A 3x3y3z(Rxy A Ryz A "-'Rxz) 4. '"'-'Faa V '"'-'Faa 3, Imp
4, Idem
5. '"'-'Faa
'Loves' in our world can again be used to illustrate this property, along 6. 'v'x r--J Fxx
5, DI
with 'two inches from' and 'defeats'. Every relation is either transitive,
intransitive, or nontransitive. e relational properties of identity, the following are true
To sum up th
A relation R may be such that for every x in a domain, Rxx. Such , ny universe of discourse:
In a
relations are said to be reflexive. Identity is reflexive in any domain
i.e., 'v'x(x = x) is true in any universe of discourse. We may thus sa; 'v'x'v'y'v'z(x = yAy = z ---+ x = z}
that a relation R in a given domain is reflexive iff 'v'x'v'y(x = y ---+ y = x}
'v'x(X = x}
'v'xRxx 'v'x'v'y(x =f. y A x = Y -+ Y =f. x}

Other examples of reflexive relations are' :::;;', 'is the same age as', and Identity is transitive, symmetrical, reflexive, and, oddly enough, ant~-
. 1 Each of these formulas will be a theorem when PC IS
probably 'loves' in the domain of natural persons. A relation R is symmet rIca. . .
irreflexive in a domain iff d d 'n the next section to include IdentIty.
expan e 1

'v'x "-' Rxx Exercises

. d . 1 'fy the following relations according to the relational
I. Supposmg a omam, c aSSI
Examples of such relations are 'is the mother of', '<', and 'runs faster properties they do and do not have.
than'. Finally, a relation may be neither reflexive nor irreflexive. Such (a) Rxy: x < y + 1
a relation is said to be nonreflexive. A relation R in a given domain is (b) Rxy: x is the mother of y
nonreflexive iff (0) Rxy: x :::;; y
(d) Rxy: x =f. Y
(e) Rxy: x is a female br~ther of y
3xRxx A 3x "-' Rxx (f) Rxy: x is good and y IS good

Examples of nonreflexive relations would be 'loathes' and 'is the square 2. Give examples of relations whioh are, in some domain,
of' (1 is the square of 1 but 2 is not the square of 2). (a) transitive and no~symm~trical
(b) symmetrical and Irreflex.lve
A relation may have various combinations of these properties. For (0) transitive and nonreflexlve
example, a symmetrical relation such as 'is equal to' may be transitive, (d) transitive and intransitive
but the symmetrical relation 'lives next door to' is intransitive in the
natural world, given linear construction procedures, and the sym- 3. Prove the following, if true: . . .
(a) Every irreflexive and transitive relatIOn IS nonsym~etrI.c. .
metrical relation 'is one foot from' is nontransitive. However, some (b) Every relation that is both .symme~rical and tranSItIve IS also refleXIve.
properties entail the presence of other properties. For example, all
(0) Every asymmetric relation IS reflexlv.~. and irreflexive is also asym-
intransitive relations are irreflexive. This can be formally demonstrated (d) Every relation that is b oth t ranSl t Ive
as follows. We wish to prove that if Fxy is intransitive, then it is metrical.
irreflexive. In other words, we wish to prove: (e) No relation can be intransitive and reflexive.
(f) Every asymmetrical relation is irreflexive. .
(g) No relation can be transitive, reflexive, and asymmetrIC.
'v'x'v'y'v'z(Fxy A Fyz -+ r-...JFxz) f- 'v'x"-' Fxx

j.,
 167
l:I.,,,H~,ate Logic with Identity: System PCI
Predicate Logic with Identity
by two In . d'IVI'd u al variables or constants is a well-formed
Answers f familiarity these two definitions are then
For purposes 0 , " ' bl
1. (a) transitive, asymmetrical, and irreflexive. be the same or different mdIvIdual vana es or
(f) nonreflexive, symmetrical, transitive. (let t an d t 1

2. (a) x follows from y.

(b) x is different from y. t = tl =df Ittl
(c) (f) in 1. = t I ) =df r-Jlttl
,-....;(t
(d) any relation in which in the domain r-v3x3yFxy.
also write ---,(t = t I ) as t -=1= t I , " .
3. (d) 'v'x'v'y'v'z(Fxy 1\ Fyz --+ Fxz) 1\ 'v'x,-.., Fxx I- 'v'x'v'y(Fxy --+ ,-..,Fyx) may t d ,'ve all valid well-formed formulas uSIllg =, It IS
enable us en
To I l the first rule let t be any m IVI uaI
0 , d"d
1. 'v'x'v'y'v'z(Fxy 1\ Fyz --+ Fxz) A
a.: ' nt to add two ru es. n
2. 'v'x,-.., Fxx A SUlliCle
3. Fab 1\ Fba --+ Faa I, DE (3) constant, then
4. ,.....,Faa 2, DE
5. ,.....,(Fab 1\ Fba) 3, 4, MT
Y INTRODUCTION (II): . f
IDENTIT . t d ce into a proof at any time any Illstance 0 t = t.
6. Fab --+ ,-..,Fba 5, DM, Imp, DN We may III ro u
7. 'v'x'v'y(Fxy --+ ",-,Fyx) 7, UI (2) . b d the validity of 'v'x(x = x). The second rule is
This rule t~e t~:~h ~:at if a = b, then if a has property F, then b has
based on F Let At be a wff containing the term t and let At l . be a wff
prope:t! . h t t The rule may be stated as follows, lettmg t and
e
conta~lll~g !h s::e ~~ different constants and letting Atl be At with tl
7.3 Predicate logic with identity: System PCI

The logician's intuition, to a large extent, seems to control whether = is tl agaI~ e. e more occurrences but not necessarily all occurrences:
replaCIng t m one or '
classed as a logical constant along with 'v' and 3 and statement con-
nectives. It must be so classed to obtain predicate logic with identity, IDENTITY ELIMINATION (IE):
If I- t = tl and I- At, then I- AtI ·
and it is convenient in mathematics and the sciences and in establishing
the validity of arguments to use a predicate logic with identity. The system resulting from these ~odifications is usually called the
With identity introduced into logical formulas, we need to add to predicate calculus with identity (abbre~Iated POI).
what constitutes an interpretation of a formula. We get an inter- Let us noW prove some theorems m POI.
pretation for a formula in predicate logic when we follow the previously
given instruction, and, in addition, we assign to = the identity predicate T1 'v'x'v'y(x = Y --+ Y = x)
A
::~
U
where a = b is true iff a is the same individual as b. 2: II
A formula is valid iff it is true for all interpretations. From allowing 3. b = a 1,2, IE
= into formulas and from this expanded definition of an interpretation, 4. a = b --+ b = a RCP
5. T1 DI (2)
it follows that formulas such as
In this proof we follow the familiar strat~gy ~f ass~min~ the ante~de~t
'v'x(x = x)
within the quantified formula. IE appbes smce III thIS case t - tl IS
'v'x'v'y'v'z(x = y 1\ Y = z --+ x = z)
a = b, At is a = a, and Atl is b = a.
are valid formulas.
The system of predicate logic, PC, may be expanded so as to include
such formulas as theorems. The usual method of doing this is first to
introduce Ixy into the list of primitive terms. Ixy is introduced as a
T2

.
U
'v'x'v'y'v'z(x = Y 1\ Y = z -->- x = z)
1. a=bl\b=c
a = b
3. b = c
4. a = c
A
1, S
1, S
2, 3, IE
predicate constant and interpreted to mean identity. The rules for a
5. (1) -->- (4) RCP
well-formed formula are then adjusted so as to allow formulas using 6. T2 DI (3)
Ixy. This adjustment is merely a matter of indicating that I immediately

",?,
 ·
rreUlCate Logic with
T3 169
Fa ---+ 3x(x = a II Fx)
Fa . VxVy(x = Y ~ (Fx Fy))

U
A f--+
2. a= a (!) VxVy(Fx II x = Y ~ Fy II x =
3.
4.
a = a II Fa
3x(x = a II Fx)
II
1,2, Conj () Fa II ,....,.,Fb
(f) ~ 3x3y(x * y)
y)

3, EI
5. T3
RCP
T4 VxFx ---+ 3x(x = a II Fx)
T5 PrOD f.s for arguments
a = b ---+ (Fa ---+ Fb)
T6 Fa II ",-,Fb ---+ "'-'(a = b) · easI'1y see n . Often
nce of PC with identity to argumen t s IS . the

~ FaA~Fa
. Fa II ",-,Fb
A re1eva
.c: ula that correspon s d to an argument can convemently
. be
2. a = b A
>

3. ",-,Fb 101m . _
by employIllg -. n A d when = is employed III such a way,
t'
4. "'-'Fa
I, S .;l"flV~'<"''-'~ is made POSSI'ble b Y USIll.
. g PCI For example , this argumen IS
2,3, IE
[ 1,4, S, Conj trans pa rently valid:
6. "'-'(a = b) RAA
7. T6 RCP LewIs
. Carro11 was Charles Dodgson. Carroll wrote Alice in Wonderland.
So Dodgson wrote Alice.
One could construct a proof for T6 without RAA by using T5.

T7 (a = b Va = c) II Fa ---+ Fb V Fe A d validation is forthcoming using a simple a pp!icatiodn o~:! once :.he

T8 VxVy(Fx II x = Y ---+ Fy) n a
argument IS , t ly sym bolized using =. Here IS the eSll e proo.
. correc
rI. Fa II a = b A
L1: Fb I, IE Prove: c = d, Ac ~ Ad
3. Fa II a = b ---+ Fb RCP I. c=d A V: natural world
4. T8 3, DI (2) 2. Ac A c: Lewis Carroll
3. Ad 1,2, IE d: Charles Dodgson
T9 VxVy(x = y ---+ x = x II Y = x) Ax: x wrote Alice in Wonderland
TlO VxVy(~(x = y) -+ ~(y = x))
. I y pI'oven identity theorems are used in the next validation
'l'wo prevIOUS
The definitions for some of the working notions used in this book
of an argument.
may now be unconditionally stated. Formulas in logic are well-formed
formulas of the predicate calculus with identity. A statement is analytic
Chicago is crowded. Any crowd~d city has its slums. This city has
if and only if it is an instance of a theorem of PCI. Form accounts for no slums. So this city is not ChICago.
analyticity, and a form of a statement is obtained by obtaining a formula Prove: Ga, Vx(Gx ~Sx), '"'-'Sb ~ '"'-'(b = a)
in logic exemplified by the statement. Often analytic statements are I. Ga A V: natural world
spoken of as logically true. Finally, the premises of an argument logically 2. Vx(Gx~Sx) A a: Chicago
imply the conclusion if and only if the corresponding conditional of the 3. ,...,Sb A Gx: x is crowded
argument is analytic. 4. Ga~Sa 2, DE Sx: x is a slum
5. Sa 1,4, MP b: this city
Exercises 6. Sa II '"'-'Sb
~'"'-'(a=b) T6
1. Prove the unproven theorems in this section. 7. '"'-'(a = b) 6,5,3, Conj, MP
8. '"'-'(b = a) 7, T10
2. Prove the following theorems.
(a) VxVyVz(x =1= y II Y = z ---+ x =1= z)
(b) Vx(Fx ---+ 3y(x = Y II Fy)) TI and substitution are used III . t 1'0 d u~ t'IOn of T6. on1'f
. the III line'ustifica-
6. TIO,
(c) Vx(Fx ---+ Vy(x = y ---+ Fy)) UE d MP are used to obtain line 8 from hne 7. To SIm~ 1 y J
tion,' an .
these relatively ObVIOUS moves ale, not explicitly indIcated.
-"I·,
 Predicate Logic with Identity Using the Identity Sign 17l
!lVIl~"oU"'''''!l
In this next somewhat longer validation, another theorem is used:
. ne person who is identical with everyone. Therefore everyone
(f) There IS 0 d' .
. . everyone is not Ivme. .
No one but Ivan and Joe had the plans. Someone who had the plans is dIvme or !'t.. 'f they are not one and the same then eIther one
took the briefcase. So either Ivan or Joe took the briefcase. (g) For any two po 1 IClhans.I
Prove: Vx(Px -+ x = a V x = b), 3x(Px II Bx) f- Ba V Bb . rupt or the ot er IS corrupt . Therefore either any politician, if he
IS cor . corru t or Anderson is corrupt.
1. Vx(Px -+ x = a V x = b) A V: natural world is not Ande~s~~, IS' f he i~u"t Anderson then he is honest. Therefore for
2. 3x(Px II Bx) A Px: x is the holder For any
(b) any poll~tt~c~an
two po 1 'f they are not one and the same, then either one is
1 !Clans, 1
3. Pe II Be 2, EE of the plans
4. Pe -+ e = a Ve = b 1, DE a: Ivan
honest. . h t For any politician, if he his nott Anderson, then he
5. Pe 3, S b: Joe (1') Anderson IS ones.
. h ones.,
also IS t So happily , all politicians are ones.
6. e = a V e = b 4,5, MP Bx: x is a briefcase
7. Be 3, S taker
8. Ba V Bb 6, 7, T7 Answers
3x(x =f- 10 II Ex) f- 3xEx II (ElO ---+ 3x3y(x =f- y II Ex II Ey»
(e) 1. 3x(x =f- 10 II Ex) A E

We have omitted the explicit introduction of a substitution instanoe of 2, aT-.L 10 II Ea

A E
1,

T7, and the use of MP. Indirect proof using IE can be used to construct a
validation for the argument that ends this section.
r::
3 ElO

Li a =f- 10 II Ea II ElO
3x3y(x =f- y II Ex II Ey)
2,3, Conj
4, EI (2)
6. (3) ---+ (5) RCP
The agent who found the bomb was in the hotel. Now if anyone Was E 2, S
7. a 7 EI
in the hotel he was in the city. If anyone was in New York, then he 8 3xEx ' .
was not in town. In fact, Bond was in New York. Therefore Bond 9: (8) II (6) 8, 6, Con]
was not the agent who found the bomb. f 3 Vx(x - y) f- VxDx V Vx I " . ' Dx
Prove: Ha, Vx(Hx -+ Ox), Vx(Nx -+ "-'Ox), Nb f- "-'(b = a) ~ » /xVy(x -; y ---+ Ox V Oy) f- Vx(x =f- a ---+ Ox) V Oa
1. Ha A V: natural world (f) Vx(x =f- a ---+ Hx) f- VxVy(x =f- y ---+ Hx V Hy)
2. Vx(Hx -+ Ox) A Hx: x is a hotel (i) Ha, Vx(x =f- a ---+ Hx) f- VxHx
3. Vx(Nx -+ ,,-,Ox) A Ox: x is in the city
4. Nb A Nx: x is in New York
5. b = a A a: the agent who found the
6. Nb -+ ,,-,Ob 3, DI bomb 7.5 Symbolizing using the identity sign
7. ,,-,Ob 6,4, MP b: Bond
8. ,,-,Oa 4, 5, IE
9. Ha-+Oa 2, DI The introduction of the sign for identity ~llow~ us. to expr~ss a ~etw
10. ,,-,Ha 9, 8, MT array of statement forms. Specifically, the IdentIty SIgn can e use 0

II. Ha II ,,-,Ha 10, 1, Conj express statement forms such as:

12. "-'(b = a) RAA
(i) At least n things are F.
Exercises (ii) At most, n things are F.
(iii) Exactly n things are F.
I. Establish the validity of the following arguments.
(a) There is an integer equal to 5 and odd. Therefore 5 is odd. Let us consider these in this order.
(b) 2 is even. Therefore if any integer is 2, then it is even.
(c) Jones is Smith. Therefore Jones hit the car iff Smith hit the car.
(i) At least n things are F.
(d) Only Jones and Smith are unemployed. Jones and Smith are both
sleeping. So everyone unemployed is sleeping.
(e) There is an integer that does not equal 10 and is even. Therefore there is To say at least one thing is F, all we need to do is to use this form:
an integer that is even and if 10 is even then there are at least two
integers that are even.
3xFx
 173
Symbolizing Using the Identity Sign
Predicate Logic with Identity
d d to show that we are speaking of all numbers other than
but if we wish to say that at least two things are F, then X y is nee e G' To obtain our statement we merely negate (5).
x when we say xy.
(I) 3xFx 1\ 3yFy or 3x3y(Fx 1\ Fy)
(ii) At most, n things are F.
will not do. For (I) is true if only a in V has F, and 'At least two things
. 's F is to say something that is true if no
are F' is not true if only a has F. The inadequacy of (I) is shown by at most one tlllng 1 . . F
To say thO . F Thus to say at most one t,hmg IS
remembering that the following are valid: F 'f only one mg IS .
things are or 1 t I t two things are F. In other words, at most one
is to deny that a eas
3xFx 1\ 3xFx~ 3xFx . . F can be expressed by
3x3y(Fx 1\ Fy)~ 3x3x(Fx 1\ Fx)~ 3xFx 1\ 3xFx thIllg IS
(6) r-./3x3y(Fx 1\ Fy 1\ x =1= y)
We can adequately indicate that at least two things are F by using
the identity sign in this way: (6) is equivalent to
(2) 3x(Fx 1\ 3y(Fy 1\ x =1= y)) (7) VxVy(Fx 1\ Fy -----l- X = y)

Note that since x =1= y appears at the end, 3x must have the scope of the . h t mary way in which 'at most one thing is F' is para-
entire formula and 3y must have the scope of Fy 1\ x =1= y. (2) is read: and (7) IS t e cus o"F 11' d all y if x and yare both F, x is identical
ase d (7) reads: ora x a n , .
"there is an x such that x has F and there is a y such that y has F and x pl:trl ,; To express 'at most two things are F', we WrIte
WIly.
is not y". (2) is equivalent to:
VxVyVz(Fx 1\ Fy 1\ Fz -+ x = y Vx = z Vz = y)
(3) 3x3y(Fx 1\ Fy 1\ x =1= y)
(iii) Exactly n things are F.
In turn, 'at least three things are F' is expressed by:
To say exactly one thing is F is to say
(4) 3x3y3z(Fx 1\ Fy 1\ Fz 1\ x =1= y 1\ x =1= z 1\ Y =1= z)
at least one thing is F
Note that it is not enough to have x =1= y and x =1= z, for z may be the and
same thing as y. at most one thing is F
If we wish to say that something is F and has a property, but that
everything else that is F lacks this property, then x =1= y is indispensable. Thus 'exactly one thing is F' can be phrased:
Suppose we wish to paraphrase
(8) 3xFx 1\ VxVy(Fx 1\ Fy -+ x = y)
It is false that there is a number greater than any other number.
. I b t we can use a shorter formula
(8) adequately expresses what we WIS 1, u
'There is a number' would be which is equivalent to it, namely:

3xNx 3x(Fx 1\ Vy(Fy ~ Y = x))

Now we wish to say that this number is greater than any other number, . rea d : "Tllere I'S something that has F, and anything which has
. h IS
w 1nc
so if we let Gxy be the two-place predicate 'x is greater than y', we may the property F is that thing." To paraphrase
write
There are exactly two honest politicians in 'Vashington.
(5) 3x(Nx 1\ Vy(Ny 1\ x =1= y -----l- Gxy))

"").,
 175
Descriptions
nd y if x is y then the successor of x is the
any two numb ers x a ,
with V: persons and Hx: x is an honest politician, we have: (c) F 0 r
successor of Y x it is false that x is less than O.
ber
3x3y(Hx /\ Hy /\ Vz(Hz -+ z = y V z = x) /\ x =F y) (d) For any numb "f 's less than or equal to 0, then x is zero.
(e) For any num er x, 1 x 1
We conclude this section by giving some further examples of para- <J ShoW that . )
.,. 3x(Fx /\ Vy(Fy -+ y = x)
phrasing into predicate symbolism employing the identity sign.
and
There is one and only one building in Yuma and it is older than any 3xVy(Fy<:-'t Y = x) h . that there are at least three ways to
. lent to (8) thus s owmg
building in Phoenix. are eqUlva ' F'
3x(Yx /\ Vy(Yy -+ x = y) /\ Vz(Pz -+ Oxz)) expresS 'there is exactly one .

where V: objects; Yx: x is a building in Yuma; Px: x is a building in Answers

Phoenix; and Oxy: x is older than y. 3x3 (Px /\ Py /\ x =F y /\ Bxy).
1. ~;~ ';'C(P; --+ 3y3z(y =F z /\ Cy /\ Cz /\ Byx /\ Bzx)).
Everyone admires himself more than he admires anyone else.
Everyxadmires himself and Vy if Y =F x then x admiresxmorethan y. 2. (a) Vx(x = 0 V 3y(x = yl»~
Vx(Axx /\ Vy(y =F x -+ Mxxy») (e) Vx(x < 0 V x = 0 -+ x - 0).

where V: persons; Axy: x admires y; and M xyz: x admires y more than z.

1.6 Definite descriptions
Everyone wants to make himself happy; only somB, not everyone,
want to do this for others. . f h f the following is a definite description.
VxMxx, and some people want to make others happy, and not all '1'he subject expressIOn 0 eac 0

want to do this.
(1) The yak in Nebraska is fat.
VxMxx /\ 3y3x(x =F y /\ Mxy) /\ ,...../v'x3y(x =F y /\ Mxy) The largest mammal at Berkeley plays football. .
hn Kennedy was an anarchIst.
The man wh 0 shot Jo
where V: persons; Mxy: x wants to make y happy. The author of Waverley was Scottish.

Exercises then called upon to paraphrase into symbols singular

1. Symbolize the following propositions using the indicated interpretation. Tn the. ~as '':th bJ'ect terms that are definite descriptions, we have
proposItIOns WI su d d
V: set of buildings in the U.S.; Ax: x is of architectural merit; Px: x is in used the form Fa. For example (1) may be ren ere
Phoenix; t: Tribune Tower; Bxy: x looks better than y; Cx: x is in Chicago;
x = y: x = y Fx: x is fat; a: the yak in Nebraska.
(a) There is only one building of architectural merit in Phoenix. (2) Fa
(b) At least one building in Phoenix is better looking than the Tribune Tower.
(c) There are at least two buildings in Phoenix. Let us use the upside-down iota 1 and rea d 1 as "the.
"
(d) There is one building in Phoenix that is better looking than any building
in the U.S. 1XYX Yx: x is a yak in Nebraska.
(e) There are no two buildings in Phoenix such that one is better looking
. a yak in Nebraska." (2) may thus be
than the other. will read: "the x such t h at x IS
(f) For every building in Phoenix there are at least two buildings in Chicago further paraphrased as:
that are better looking.
2. Symbolize the following. F·/xYx
V: positive integers; x = y: x = y; x': x' is the successor of x; 0: 0; x < y:
x is less than y A puzzle or paradox suggests itself when one reflects on nega:~ng
(a) For any x, either x is zero or x is the successor of some y. statements of the form F1xGx. At first sight it seems that the nega IOn
(b) For any x, x plus 1 is the successor of x.

,.
 .L/U
Definite Descriptions 177
of, say,
If (4) is taken to be a proper paraphrase of (I), then our paradox is
(I) The yak in Nebraska is fat. resolved. (3) becomes

is (5) 3x(Yx /\ 'v'y(Yy ---+ Y = x) /\ r-->Fx)

(3) The yak in Nebraska is not fat. Now (4) and (5) are contraries-they both cannot be true-not contra-
dictories; thus (5) is not the negation of (4). The negation of (I) IS
Now if this is so, then (I) V (3) is an instance of the tautology p V simply obtained as follows:
and thus (I) V (3) must be true. But there are no yaks in Neb k ,,-,p
. N b k ras a not
e~en m eras a zoos. This being the case, both (I) and 3 are' (6) r-->3x(Yx /\ 'v'y(Yy ---+ y = x) /\ Fx)
smce, apparently, both (I) and (3) imply that there is a yak' (N) b false
If (I) d m e raska
an (3) are both false, then (I) V (3) is false. This then is . '(I) or (6)' is thus an instance of P V r-->P, not '(I) V (3)'. (4) and (5)
puzzle: (I) V (3) seems to be both true (since it is an instance of pour both imply 3x Y x, but (6) does not. In fact, if r-->3x Y x, then (6) can be
and false (since there are no yaks in Nebraska). V "-'P) seen to be true.
The c~stomary logical analyses of such statements as {l) and (3 In proving the validity of the arguments that follow, we will
resol~e tIns paradox, and others, by maintaining that (3) is not th~ occasionally paraphrase propositions of the form F1xGx as 3x(Gx /\
negatIOn of (I). ~ clue to the analysis of (I) comes from notin that. 'v'y(Gy --+ x = y) /\ Fx). Some logicians have questioned such a pro-
any of the followmg are true. g If cedure on the grounds that any statement of the form Fa is neither true
nor false if there is no a. Some, for example, have argued that if one
(i) There are no yaks in Nebraska. utters statements of this form and there is no a, for example, if one
~~~) There is more than one yak in Nebraska.
(m) The yak isn't fat. utters (I) and there is no yak in Nebraska, then the utterance (I) is
neither true nor false. A necessary condition for the utterance being
then (I) would be said to be false in a familiar sense of '.cal ' W either true or false, it is argued, is that the substitution for 1XGX have a
d' 'd i. se. e may bearer. There are serious difficulties with this account, but there is no
IVI e statements into those for which there is a complex I'n th .
f d' e UllIverse doubt that often, though not always, in the nonformal use of language,
o Iscourse corresponding to what is asserted in the st t t
th £. . a emen , and ifthe description or singular term fails to designate anything existing, the
ose or wInch there IS no complex-the former can be called t rue d
the latter false. And it is in this sense of 'false' that we ma sa IaI~ statement as a whole would not be said to be either true or false. The
false if any of (i) to (iii) are true. y y ( ) IS explanation for this is quite simple. Typically we would understand
'" If (i), (ii), and (ii~) im~ly (I) is false, then the denial of (i), (ii), and
It is false that the yak in Nebraska is fat.
(m) o.ught to appear m (I) s logical analysis to bring out the validity of
such mferences. As we saw in the last section, this expression 'There is
exactly one yak in Nebraska', may be symbolized: ' as implying that the yak is lean, so we would hesitate in saying this if we
knew that there are no yaks in Nebraska. In other words, the negation
3x(Yx /\ 'v'y(Yy---+y = x)) of the form Fa is commonly understood to imply a is not F and this
implies 3x r--> Fx (by EI). So if there is no a, to avoid possible mis-
~h~ property of being fat can be conjoined within the scope of 3x thus understanding, one does not very well want to negate Fa. However, to
gIvmg us ' say Fa is neither true nor false if there is no Ct means, it seems, that
neither Fa nor a is not F is true. It does not mean that neither Fa nor
(4) 3x(Yx /\ 'v'y(Yy---+y = x) /\ Fx) ""Fa is true. If this is so, Fa is neither true nor false if there is no
a-Fa and a is not F are both false-and yet r-->Fct would be true.
:vh~cl~ reads: "Some individual is a yak in Nebraska and only one Specifically, r-->Fa is true when there are no a's.
mdlvldual is a yak in Nebraska and that individual is fat."
Paraphrasing definite descriptions in the manner above will enable
 179
Definite Descriptions
that we can paraphrase F1xGx as 3x(Gx /\ Vy(Gy -+ x = y) ./\ Fx)
us to establish the validity of certain arguments. 'ro illustrate: <J 'fhe theory 11' d . ly one among a number of such theOrIes for
"" t d Russe s an IS on . ., h n1
is Ber ran. definite descriptions. The above obJectIOn IS not teo y one.
The philosopher who taught Alexander was a Greek. Aristotle was paraphrasmlg ld one take all of the following to be true:
the philosopher who taught Alexander. Therefore Aristotle was a Flor examp e, cou f 1
(a) F1XGx is always either true or. a se.
Greek. (b) The God of Islam does not eXIst.
Prove: 3x(Px /\ Vy(Py -+ x = y) /\ Gx), Pa f- Ga (c) Malcolm X worshipped the God of Islam.
1. 3x(Px /\ Vy(Py A V: natural world hold to Russell's account?
and yet
-+ x = y)/\ Gx) Px: x is a philosopher who
A taught Alexander A ShoW that G
2. Pa "'. , 3x(Fx /\ Vy(Fy -+ y = x) /\ Xi
3. Pb /\ Vy(Py -+ b 1, EE Gx: x is a Greek
is equivalent to .
= y) /\ Gb a: Aristotle 3x(Gx /\ Vy(Fy+-t Y = x») . , F . G'
4. Vy(Py-+b = y) 3,8 . 'that there are at least two ways of expressmg the IS a or
5. Pa-+b = a 4, UE
thus showmg . ,
'the one and only F IS a G . . . If
6. b = a 2, 5, MP . I'd f rmula 3x(x = a) can be derived from It (do thIS).
7. Gb 3,8 5. Vx(x ~ x) IS advapl. 0 I'd then'Q is valid. But is 3x(x = a) valid? How can
8. Ga 6,7, IE P implIes Q an IS va 1 ,
this puzzle be resolved? . .
' t I gic with identity applies to mathematIc reasomng,
Exercises l! 'fo see how pre d ICa e 0 d th
u. k me derivations in a theory calle group eory,
let uS ma e so h
1. Using the indicated interpretation of exercise 7.5.1, symbolize the following. The axioms (abbreviated A) of the t eoryare
(a) The one and only building in Phoenix is worse looking than a building Al VxVyVz(x + +(y z) = (x + + y) z)
in Chicago. A2 Vx(x + 0 = x)
(b) Only the building of architectural merit in Phoenix is better looking than A3 Vx(x + x = 0)
. r r'etation that we will select is as follows: V: set 0 POSI Ive ~n
f 't' d
the worst looking building in Chicago.
(c) There are no buildings in Phoenix.
T~le I~te p + 0 d _ have their usual arithmetical interpretatIOn.
(d) If there are any buildings in Phoenix, then the best looking building nega~~e m!~~e~~antifi~:~on and identity rules are used, 0 is a term (a~
looks worse than the worst looking building in Chicago. .
mstance o
en t) Wh n _ is used with a term the result is a term. When
f . e . d t
+ IS
used between two terms the result IS treate as a erm.
2. Validate the following arguments. The formula
(a) The philosopher who taught Alexander was a Greek. Therefore some Tl VxVyVz(x + z = y+ z-+x = y)
philosopher was a Greek. . . . . b he three axioms of the theory. It is called a theorem,
is logICally ImplIed ylt (bb . ted Tl) We establish that Tl is logically
(b) There is exactly one unscrupulous politician. Harry Fogg is clearly an specifically, theorem a reVIa '
unscrupulous politician. Therefore Fred Zilch is not an unscrupulous implied by the axioms as follows.
politician. 1. VxVyVz(x + + (y z) = (x + + y) z) !~
(c) The teacher who wrote that book writes only logic texts. Therefore that 2. Vx(x + 0 = x) A3
book is a logic book. 3. Vx(x+ x = 0) A
(d) The teacher of logic at the university is very smart. 80 it is false that 4, a + +c= b c
some teachers of logic at the university are not smart. 5. (a+ + c) +
c = (a c) c + II
(e) For any x there is exactly one y such that y = x. (Prove as a theorem.) 6. (a+ + c) +
c = (b c) c + 5,4, IE
(f) The only brother of Mary is the father of Harry. Therefore Mary is 7. a + + (c 8) = (a+ +c) c 1, DE
Harry's aunt. 8. a + + (c +
c) = (b c) c + 6,7, IE
(g) The highest mountain in Washington is higher than any mountain in 9, b + + (c +
c) = (b c) c + 1, UE
California. Mount Rainier is in Washington and Mount Shasta is in 10. a + + (c c) = b + +
(c 8) 8, 9, IE
California. Mount Rainier is the highest mountain in Washington. 11 +-
. c c=
0 3, UE
10 11 IE
Therefore Mount Rainier is higher than Mount Shasta. 12. a + +0= b 0 21m'
(h) Every integer added to 0 equals itself. For any two integers x and y,
x + y = y -+ x. Therefore there is exactly one integer such that for
13. a + 0= a 2' UE
any integer x, x +
y = x. (Remember numerals are terms.)
14. b + 0= b 12, 13, 14, IE
15. a = b _ RCP
(i) Every boy can jump higher than any boy who is smaller. Therefore if
there is a tallest boy then there is the highest jumping boy.
16. a+ c = b + c -+ a - b 16 Dr (3)
17. T1 '
-, ..
 180 Predicate Logic with Identity Definite Descriptions 181
It is the practice in mathematics to shorten such proofs by not expli 'tl Similarly, we can obtain these results, supposing the universe of discourse
' d own. t h e premIses,
settmg . t h e aXIOms,
. Y
but entering them as needed withClDE again to be everything.
already applIed. These proofs can be shortened even further if we go from 3F3xFx: "Some property belongs to some individual."
M~ ,
VFVxFx: "Every property belongs to every individual."
+ +
n. (a c) + +
c = (b c) c 3FVxFx: "There is a property which belongs to every
directly to individual.' ,
a+ + (c + +
c) = b (c c) 3xV F Fx: "There is an individual which has every property."
with our justification being merely 'from n by AI'. In effect we collapse the V F3xFx: "Every property has at least one instance."
two uses of DE and two uses of IE into the justification 'n, AI'. VxVy3F Fxy: "Any two things bear some relation to each other."
. We may make use of th~orem one in another proof since proven theorems The defining feature of the lower predicate calculus ~s that in its well-fo:med
logICally follow from the aXIOms. ~ rmulas all predicate variables are free. The defimng feature of all hIgher
We will use the above short cuts in the next proof. o~der predicate formulas is that one also has bo.und predicates. .
T2 Vx(x + + 0= 0 x) (a) Selecting the domain of discourse as everythmg, paraphrase the followmg
1. 0+ + (a + +a) = (0 a) a Al into logical notation.
2. 0+ + +
0 = (0 a) a 1, A3 (1) It is false tha~ there are t~o.things t~at .have the sar.ne properties.
+ +
3. 0 = (0 a) a 2, A2 (2) Every irreflexIve and tranSItIve relatIOn IS asymmetrIc.
4. a + + +
a = (0 a) a 3, A3 (3) Honesty is a virtuous prop~rty.
5. a = 0 + a 4, TI (4) Nothing is related to everythmg.
6. a + +0= 0 a 5, A2 (b) Construct an argument of some sort to show that the valid formulas
7. T2 6, UI below are valid and the invalid formulas are invalid.
Derive the remaining theorems below: (1) VxV FFx+--+ V FVxFx
T3 VxVyVz(x + + y = x z-+y = z) (2) V F3xFx -+ 3x3F Fx
T4 Vy(Vx(x + y = x-+y = 0» (3) 3x3y3FFxy -+ 3FVx3yFxy
T5 Vx(x + +
x = x x) (4) 3 FVxFx -+ Vx3F Fx
T6 VxVy(x + y = 0 -+ y = x) (c) What differences do you find between these two formulas:
T7 Vx(x = x) V FVxVy((Fx+--+ Fy) -+ x = y)
VxVy(V F (Fx+--+ Fy) -+ x = y)
7. The tree method discussed in previous exercises can be used to establish the
:alidity o~ formulas. ~nd argument patterns using the identity sign. All that Answers
IS needed IS the addItIOn of two rules corresponding to II and IE. In fact no
2. (h)
restatement of these rules is needed; they can be used with the tree method as
they are. IE will continue to be used to replace equals with equals. II will
Vx(x + 0 = x), VxVy(x + y = y + x) f- Vy(Vx(x + y = x) -+ 0 = y)
be used to close paths when a formula of the form t -# t appears. 1. Vx(x + 0 = x) A
Use the tree method to establish some of the argument patterns and 2. 'Ix Vy(x+ y = y + x) A
+ a = x)

~
theorems found in this chapter. . 'Ix (x A
4. a+ 0= a 1, DE
8. System PC is extended in this chapter so as to obtain the lowe?' predicate 5. 0+ a= 0 3, DE
calculus with identity. As the reader might suspect, if there is a lower 6. 0+ a = a+ 0 2, DE
predicate calculus there must be a higher predicate calculus. Let us outline 7. 0 = a + 0 5,6, IE
how lower predicate logic differs from higher order predicate logic. 8. 0 = a 4,7, IE
We have already had occasion to note that at times we wish to express 9. Vx(x + a = x) -+ 0 = a Rep
statements about all properties of a certain kind or some properties. For 10. Vy(Vx(x + y = x) -+ 0 = y) 9, DI
example, the rule of IE introduced in the earlier section can be expressed as 5. We must remind ourselves that in giving an interpretation of a formula we
follows: give a nonempty domain and assign to each individual constant of the formula
For any individual x and y, if x = y, then for any property, if x has it, an individual in the domain.
then y has it, and vice versa.
This can be symbolized by quantifying properties as well as individual 8. (b) Only (2) and (3) are not valid.
variables as follows, where the universe of discourse is everything.
VxVy(x = y -+ V F(Fx+--+ Fy»
 II
LOGICAL
AXIOMATIC
SYSTEMS

-,.
 Up to this point, we have two
formal methods of demonstrating the
validity of well-formed formulas in
aXiom predicate logic, the natural de-
duction system pcr and the tree or
systems tableau method (introduced in the
exercises). A very different approach
proceeds by assuming axioms and
some rules of inference and de-
ducing theorems-the theorems
being valid predicate formulas and
all predicate formulas being
theorems. This is a logical axiom
system for predicate logic. Such a
system will be described in this
part of the text.
There are at least two reasons
for introducing an axiom system
for predicate logic at this point.
First, it is useful to be acquainted
with the axiomatic technique.
Second, the important metalogical
questions concerning consistency,
soundness, and completeness can be
more easily answered with respect
to an axiom system than with a
natural deduction system.
The procedure we will follow
in Part II is first to describe the
notion of a formal axiom system and
discuss again the properties of con-
sistency, soundness, and com-
pleteness. This will be followed by
an axiom system for sentential logic
and one for predicate logic. For
each system the questions con-
cerning consistency, soundness, and
completeness will be raised. We will
stop short of a proof for the com-
pleteness of predicate logic.

185
186 187

derive a contradiction. Sacchieri had, in fact, developed a non-

8. 1 The development of geometry
Euclidean geometry.
The development of the axiomatic method has made a profound impact Non-Euclidean geometries were intentionally developed in the
on human thought. To many it seems the ideal way to organize some nineteenth century. In hyperbolic geometry, for example, through a point
field of knowledge. The axiomatic method was first applied to geometr not on a given line, always more than one line can be drawn parallel to
~nd the hist~ry of. the development of geometry provides many insighr~ the given line; and the sum of the angles of a triangle is always less than
mto the aXIOmatIC method. For our purposes, the most interesting two right angles. Out of the development of these geometries emerged
feature is how the development of geometry made people conscious of the pattern of how an axiom system should be constructed and developed.
the conditions for a formal axiom system. Let us very briefly review Since familiar spatial diagrams and intuition could not be used to
some aspects of this development. clarify the axioms of the systems, and neither could they be used as an
Most of us will recall that geometry is taught as a deductive dis- assistance to the derivation of theorems, a way had to be devised of
cipline. It is not presented as, say, a collection of contingent statements developing the systems independent of intuition. A proof of a theorem
that are adopted because they agree with observations. Rather a few must be a matter of what logically follows from what. We must pay
statements, axioms, are given without proof (e.g., 'a straight line can be nO attention to the meaning of the primitive terms of the system, but
drawn from any point to any other point', and 'things that are equal to think only about the logical form of the statement. But for this to be
the same things are also equal to one another') and from these, other possible, geometry had to be constructed as a formal axiom system.
statements, theorems, are deduced by using some logical rules. This We turn next to the notion of a formal axiom system, along with a more
procedure, begun by Euclid around 300 B.C. in The Elements, is called the precise account of the notion of theory, axiom, and theorem.
axiomatic method.
The ancient geometers saw two advantages in organizing geometry Exercises
in an axiomatic way. First, all the geometry truths that had been 1. What would be the consequence if each axiom of Euclidean geometry could
collected for hundreds of years by the Egyptians and others, could be be derived from any other axiom?
systematically organized. Second, if the axioms could be so selected that 2. At what points did intuition break down in the development of geometry?
they were necessarily true, then, if the deductions were valid, the truth 3. Pascal is said to have desired a system for superhuman minds in which all
of the theorems would be guaranteed, no matter how complex they terms would be defined and all statements demonstrated. Is such a system
might be and no matter whether they were self-evident. possible?
The Greeks regarded the axioms in Euclid's system as self-evident, 4. Must an axiom in a system have the feature of being intellectually self-
with the exception of the fifth postulate. The fifth postulate found in evident?
Euclid's system is equivalent to the assumption that through a point
Answers
outside a given line, only one parallel to the line can be drawn. It has
become known as the parallel postulate. 1. Only one axiom would be needed.
The ancient geometers made attempts to derive the parallel 3. No. Thus in a system some statements-axioms-must be supposed
postulate from the other axioms. All of these attempts failed, since, as it and some terms-primitive terms-undefined.
was later proven, the parallel postulate cannot be derived from the other 4. The logical status of an axiom is that of being an assumption from which
axioms. The effort to derive the postulate continued through the middle some set of statements are logically derived. Logically speaking, they need
ages, and in the eighteenth century an Italian mathematician, Gerilama not even be true.
Sacchieri, assumed the negation of the parallel postulate and tried to
deduce a contradiction. If he had succeeded in this reductio ad absurdum 8.2 Formal axiomatic theories
strategy, then he could have concluded that the parallel postulate is
deducible from the other axioms. Though Sacchieri succeeded in deriving A theory can be identified with two sets of statements. First, there is the
many counterintuitive and odd-sounding theorems, he was unable to set of statements concerning the subject matter of the theory. In

'">1·,
 188 189
geometry, all those statements about lines, points, and so forth, would are true in every admissible interpretation are, of course, the valid
make up this first set. In nuclear physics, all those statements about
formulas.
subatomic particles would ma~e up this first set. In sentential logic, aU In this book our primary interest is in formal theories for sentential
well-formed formulas would make up this first set. The second set is . nd predicate logic with identity. In Part I the formal _theory. for
the subset of the first set that is regarded as acceptable. All the state_ IogiC a . .
these logic systems is a natural deductIOn system. In thIS part we WIll be
ments regarded as true in geometry or nuclear physics fall into this interested in formal axiom systems for logical sys~ems. .
second set. All valid well-formed formulas of sentential logic fall into The difference between the natural deductIOn system for lOgIC
this second set.
presented in Part I and those presented in this part is not in terms of
A theory is an axiomatic theot·y when some acceptable statements are being formal. All are formal ~heories. Ra~her the difference re~ts in the
taken as axioms, as sta~ements acceptable without proof, and the fact that the natural deductIOn systems III Part I have no aXIOms but
remaining acceptable statements are theorems, or are the hoped for only rules. In an axiom system one starts with a, set of well-formed
theorems. A theorem is a statement deducible from the axioms. A formulas that are accepted without proof. Using rules, one then derives
theorem T is deducible from a non-logical axiom set S if and only if
fr om these axioms the theorems. , In a natural deduction system for
S ~ T is analytic, that is, if S ~ T .is an instance of a valid logical logic there are no axioms, but there are more rules. For a logical theory,
formula.
as the rules increase the axioms decrease (and vice versa) and when one
A theory is ajormal system when it has the following features: has the maximum number of rules (no axioms), then one has a natural
deduction system. In a natural deduction system for logic, we should
1. The notion of a statement of the system is effective, i.e., there is a note, there must be some rules that allow us to begin proofs by intro-
mechanical procedure by which in a finite number of steps one can ,ducing assumptions, rules like Rep and RAA. In axiom systems we can
tell whether some marks are or are not statements or formulas start proofs with axioms.
of the system. If the notion of a statement of the system is To illustrate some of the remarks in this section, let us set down as
effective, then a digital computer can be programmed to check
whether a string of marks of the system is or is not well-formed. simple a formal axiom system as wl} can imagine. l The primitive symbols
2. The notion of.a.proof is effective, i.e., there is an effective pro- of the system are these five: 1, 2, 3, =, and~. A well-formed string of
cedure for deCIding whether a sequence of statements is or is not these symbols may be defined as follows:
a proof. The last step in a proof is a theorem; thus a system is a
formal system only if checking a proof of a theorem can be 1. preceded by 1, 2, or 3 and followed by 1, 2, or 3 is a wff.
reduced to a mechanical procedure. 2. If A and Bare wffs then (A ~ B) is a wff.

Our only rule of inference is MP and the theory has exactly two axioms
If a system is formal, if one can check mechanically whether a string of
marks is well-formed or not, and if one can check mechanically whether a Al 1= 2 ~ (1 = 2 ~ 2 = 3)
sequence of lines is a proof of a theorem, then theorems can be proven in A2 1= 2
a system independent of what meaning one gives to the marks of the
system. If a computer can check whether a sequence is a proof, then a A proof in the theory is a finite sequence of formulas such that each
proof does not depend on the meaning of the marks but relies only on the formula is either an axiom or is inferred from previous lines by the rule
form and the rules for the manipulation of forms. of inference. A theorem is the last line in such a sequence.
We can usefully distinguish the syntax from the semantics of a This is a formal axiom system. First the notion of a wff is effective.
theory. With logical formal theories the rules for a wff and proof are For example, Al and A2 are wffs, but
parts of the syntax of the theory. Syntactical questions in logi~ are
questions independent of any interpretation given to the symbols of = 1, 1 2 = 10, 2 = = 1, 2 = 2 = 3, 4 = 2 + 2
the system. Semantics in logic consists in giving interpretations to the
1 The idea for this and for many of the definitions in this section is taken from
symbols, e.g. basic truth-tables for statement connectives, and in Angelo Margaris, First 01'der Mathematical Logic (Waltham, Mass.: Blaisdell
specifying when formulas are true in a given interpretation. Formulas Publishing Co., 1967).
 191
190 VB
SYMBOLS:
are not wffs. The definition for a wff settles in all cases whether a string
of symbols is ~ wff or not. The notion of an axiom is effective since .A 1. a, b, and m
is an axiom iff it is either Al or A2. Finally, the notion of proof is 2. A and I
effective.
The system also has exactly four theorems, AI, A2, 1 = 2 ~ 2 = 3 FORMATION:
1 A or I preceded by a, b, or m and followed by a, b, or m is a wff.
and 2 = 3. And this can be determined even though no meaning has 2: If A and Bare wffs, then ",-,A, (A A B), and (A --* B) are wffs.
been specified for 1, 2, 3, =, and ~. Formal proofs for two of the four
theorems would be
AXIOMS:
Proof 1: 1 = 2 ~ (1 = 2 ~ 2 = 3) AI: aAa
Proof 2: 1 = 2 ~ (1 = 2 --* 2 = 3) A2: ala
1=2 A3: mAa A bAm --* bAa
I=2~2=3 A4: mAa A mlb --* bla

Exercises
RULES OF INFERENCE: .
Rule 1: If A is a well-formed formula, then we may umformly
1. Why is a proof with an infinite number of steps not permitted in a formal
substitute a, b, or m for a, b, or m.
axiomatic system?
Rule 2: Modus Ponens. If I- A --* B and I- A, then I- B.
2. In the natural sciences, what controls whether a well-formed formula is
accepta ble? This axiom system will make use of sentential logic by ~sing.tautol-
3. Why is a formal system not affected when you drain its terms of meaning, . . the derivations and by using statement connectIves m con-
ogles m ' . I
that is, obviate any interpretations for the primitive terms? structing wffs as indicated in th~ ~or~ation :ules. How the tauto~oglCa
4. Give the proofs for the remaining theorems of the formal system set down at formulas may enter into proofs IS mdlCated m the last rule of CS.
the end of this section.
Rule 3: Tautology introduction rule. Any tautological wff may
5. Many of Euclid's proofs relied on features of diagrams. Was Euclid's system be entered in a proof.
a formal system?
6. Discuss how one would go about making the tree method into a formal system. DEFI,NITIONS:
7. Why are questions of soundness and completeness questions relating syntax 1. bEa = df .......,bla
and semantics? 2. bOa = df ",-,bAa

Answers
A proof in CS is a sequence of one or more wffs, each of w~ic~ either
1. It would not be possible to know when one had a theorem. is one of the axioms, an instance of a tautology, a theorem, or IS mferred
3. A machine can be programmed to register 'yes' if a sequence is a proof and from preceding wffs in the sequence by modus ponens or Rule l. !"-
'no' if it is not a proof. theorem is the last line of a derivation. In writing out a proof ;ve ~Ill
take the liberty of numbering each line and indicating to the rIght ItS
justification.
8.3 System CS
TI bEa --* ",-,bla
l. ",-,bla --* ",-,bla Identity, ",-,bla/P, R3
It will prove useful to present a second, somewhat more complicated 2. bEa --* ",-,bla df
formal axiom system. In order to make absolutely clear that this is a
formal system we will first lay down the primitive basis of the system, T2 ",-,bla --* bEa
which will be called CS, and derive a few theorems without considering l. ",-,bla --* ",-,bla Identity, ",-,bla/ P
what interpretation is intended for the primitive symbols. 2. T2 df
 'Identity' is the name for the tautology P -+ P If a tautol .
£ . . ogy IS needed interpretation of OS, i.e., the intended interpretation for the primitive
or whICh we do not have a familiar name" we will list
. it Qt th b . .
e eglnnln
of the proof, as we do in the next proof. g symbols. . .. . .
OS is an axwmatlzatwn of SyllOgIStICS. The system presented here,
T3 bla -+ ,,-,bEa with certain modifications, is taken from the Polish logician Jan
(i) (P -+ "-'Q) -+ (Q -+ ",-,P) ..:t:;ukasiewicz' work, Aristotle's Syllogistic. A and I are syllogistic quanti-
1. (bEa -+ "-'bla) -+ (bla -+ ,,-,bEa) fiers, and a, b, and m are general name variables. A general name is an
(i) bEa/P, bla/Q
2. bEa -+ ,,-,bla TI expression that determines a class. Examples of general names would
3. T3 1,2, MP be: man, five-cent hamburgers, unicorns, and class. Al reads "All a is a."
T4 ,,-,bEa -+ bla
Instances of Al would be: 'All men are men', 'all roaring mice are roaring
T5 bOa -+ ,,-,bAa
T6 ",-,bAa -+ bOa mice', and 'all positive integers are positive integers'. A2 tells us that
T7 bAa -+ ",-,bOa there is something which is a-a is not an empty term. It reads: "Some
T8 ,,-,bOa -+ bAa a is a." A3 is the valid formula corresponding to the valid syllogism
T9 bla -+ alb called Barbara and A4 is the valid formula corresponding to the valid
1. (bAb A bla -+ alb) -+ syllogism called Datasi. The system is strong enough to derive the con-
(bAb -+ (bla -+ alb)) Exportation, bAb/P, ditionals corresponding to the 24 argument sequences thought to be
bla/Q, alb/R valid in syllogistics and, in addition, all the laws of immediate in-
2. rnAa A rnlb -+ bla A4
3. bAb A bla -+ alb ference (conditionals with simple components as antecedents and
b/rn, bfa, alb (RI)
4. bAb -+ (bla -+ alb) 3,1, MP consequences), omitting inferences involving negative terms. The
5. aAa Al formation rules can be seen to be designed to give us categorical prop-
6. bAb b/a ositions (see Section 5.2) and categorical propositions linked by
7. T9 6,4, MP statement connectives.
TlO bAa -+ bla
Tll mEa A bIrn -+ bOa Exercises
(i) (P A Q -+ R) -+ ("-'R A Q -+ ,,-,P) 1. Construct proofs for the unproven theorems.
1. rnAa A rnlb -+ bla A4 2. Give some instances of the axioms and theorems.
2. (1) -+ (3) (i) rnAa/P, rnlb/Q,
bla/R 3. Provide a second interpretation for CS.
3. ,,-,bla A rnlb -+ ,,-,rnAa 1,2, MP 4. What is the logical status of a definition in a formal system, for example,
4. bEa A rnlb -+ rnOa df the definitions in CS?
5. Tll rn/b, b/rn
5. State some theorems of SC other than those given above.
TI2 rnAa A blrn -+ bla
6. Distinguish between giving an interpretation for a system and giving an
TI3 rnAa A bArn -+ bla
interpretation for a formula.
1. rnAa A bArn -+ bAa A3
2. bAa-+ bla TlO
3. TI3 Answers
1, 2, Trans, substituting,
andMP 2. A3: If all babies are illogical and all philosophers are babies, then all
philosophers are illogical.
TI4 mEa A bArn -+ bEa
3. Let a, b, and rn stand for classes, and let xAy be x c y (reads: "class x is
included in class y"), and let xly be x n y i= A (reads: "the intersection of x
. This exercise should have convinced the reader that one can have an and y is not an empty set").
aXIOm system without an interpretation for the primitive symbols of
the sy~tem: In OS, theorems are derived without supplying any inter- 4. Logically speaking, a definition in a formal system introduces no new
content; it is a decision or a convention useful for purposes of abbreviation
pretatIOns for a, b, m, A, and 1. Let us now supply the principal and familiarity.
 . . -""orma! Axiom Syste 195
Metalogical Properties of CS
6, To gIve an mterpretation for a s stem IS
. .
of the system; to give an iute y t t. to mterpret the primitive symbols rules are applied to true lines, the result must be true lines. For if
variables of the formula. rpre a IOn of a formula is to interpret the
the rules are valid, then if the axioms are true, all the theorems must be
true. In other words, given this interpretation for as, which results in
each axiom being a true mathematical statement, as is consistent if the
8.4 Meta/ogica/ properties of CS rules are valid. It is easy to prove that the rules of as are valid.
A second way to prove consistency of the axioms of a system is to
There are many reasons for develo in .
respect to syllogistics one such p ? a formal aXIOm system. With interpret the system in terms of another system. That is, if one can show
, reason IS system t' II that the axioms of system one can be interpreted so as to be theorems of
the logical truths in syllogistics from a fe ,..a lCa y ~o deduce all of
rules. Another reason is to sh ' w pnnCIples (axIOms) and a few system two, that system two has the same inference rules as system
, . ow preCIsely the relat' b one, then system one must be consistent if system two is consistent.
gIStlCS and sentential logic A thO d . IOn etween syIIo-
h . 11' reason IS to . th This second method was needed to establish the consistency of non-
wether the system has certain m t I . I ~aIse e question of
are: consistency soundness co e at oglCa propertIes. These properties Euclidean geometry since the axioms were initially regarded as plainly
.' ,mp l e eness, and independ W false of any familiar domain of things. Interestingly, each axiom of
now;o a dIscussion of these properties and as. ence. e turn
elliptical geometry can be interpreted so as to become a theorem of
system, whether an axiom system 0
is negation consistent when t h ' l' a natural deduction system
Euclidean geometry ('plane' is made to signify the surface of a Euclidean
e aXIOms and/or l' I d ' sphere, 'point' a point on this surface, 'straight line' an arc of a great
wff and its negation to be deduced as. . u es 0 not allow both a
cannot both be derived "T . IS c~nslStent, then, if A and ",A circle on this surface, and so on). On this interpretation, the replacement
. 0 say a system IS d . for the parallel postulate reads: "Through a point on the surface of
theorems are acceptable. To say r I sou~ IS to say that all its
. 't' a ~orma system IS compl t . th a sphere, no arc of a great circle can be drawn parallel to a given arc of a
pnmi Ive sense of 'complete" t e e m e most
formed formulas of the th IS °h say that all the acceptable well- great circle." A numerical interpretation of both Euclidean and non-
'd eory are t eorems of the s t A Euclidean geometry can also be provided. Thus if number theory is
sal to be independent when n . ys em. system is
0 aXIOm or rule can be d . d f consistent, so are these geometries.
th
o er axioms or rules of th t enve rom the
. d e sys em. as can easily b A third way to prove consistency of an axiom system is to find a
m ependent and consistent Th I e proven to be
problematic matter. . e comp eteness and soundness of as is a property such that (i) the axioms have the property, (ii) the rules pass
on the property, and (iii) two formulas of the form A and ,....,.,A cannot
How does one prove that any axiom s . .
Obviously one cannot prove that an a . ystem IS ~egatIO~ consistent? both have the property. If each axiom has the property and the rules
fact that the theorems alread d d dXIOm system IS conSIStent by the pass it on, then all the theses of the system must have the property.
e
it is possible that undeducelth uce do no~ contradict each other, for If A and ,....,.,A cannot both have the property, then the system must be
A . eorems may mtroduce t d' simply consistent. The property in question in such a proof may be an
n mconsistent system cannot be i . . a con ra lCtion.
the axioms and theorem . t g ven an mterpretatIOn such that all interpreted or uninterpreted property. We should note that a variation
s are rue (a model) Th' . of this method was used to prove SO and PO consistent. 'True' is of
prove the consistency of as T thO . IS provIdes one way to
. . 0 IS end let us I th course such a property; thus to use a model to prove consistency is an
mterpretation for the prim't' b ' supp y e following
lIve sym ols of as. Let example of the third way in which the property is an interpreted
xAy be x can be put in t property.
xly be x is included in ya one- o-one correspondence with y Let us use this method again to sketch a second proof of the negation
a be the set of positive integers consistency of as. The formulas for categorical propositions in the
m be positive integers divisible by 2 axioms of as can each be rewritten by following this procedure:
b be the set of even positive integers

On this interpretation each of the fou . (i) Replace the small letters with sentential variables.
statement. To prove SO consist ..1' aXIOms of as becomes a true (ii) Drop the A and put in a ~.
ent, It IS now sufficient to show that if (iii) Drop the I and put in a ~.
 196
Metalogical Properties of US 197
Following this procedure, Al to A4 become
is compIete , in what is sometimes called the relative sense of complete- h
P+--+ P h en all the acceptable formulas of the theory are theorems of t e
nesS,
. mW s stem. Supposing the field of knowle d ge h as b een .lorma .c 1·Ize d ,
P+--+ P
(Q+--+ P) II (R+--+ Q) --+ (R+--+ P) axlOmay Yay
we s a system is complete if it is sufficiently powerful to generate
(Q+--+ P) II (Q+--+ R) --+ (R+--+ P) all its acceptable well-formed formulas. The completeness of
~e~ t 0 f . .
.
OS raIses problems beyond the scope of thIS book. OS may . .be SaId
. to . be
As we can see ~y following this procedure, each axiom becomes a in another sense in that all the truths of syllOgIStlCS, Ignormg
tauto.logy. We wIl.l call formulas obtained by following this procedure compIet e . . .
·
nega t lve terms , are theses of the system. When EuclId aXIOmatIzed
as~oClated sententIal formulas (asf's). It can thus be said that all the t he tried to select his axioms so that all the truths of geometry
aXIOms possess the property of having valid asf's. It can next be seen geome ry, d· db
could be derived from his axioms-that is, not only those Iscovere y
that the rules pass on this property. If A has a valid asf and one of the Egyptians and Greeks, but all possible truths of geometry. .
ru~es is applied, .then the formula that results has a valid asf. Now if ~~: The axiom systems and natural deduction systems for sententIal
aXIOms have thIS feature and the rules pass it on, all the theses of OS logic and predicate logic are complete in the way Eu.clid hoped his
must have t~is feature. Finally, it cannot be the case that both A and geometry was. Not all formal systems, howeve~, are relatIvely c~mplete.
,......,A have thIs feature, so OS must be consistent. For example, a complete, consistent set of aXIOms cannot be gIven for
To prove an axiom system sound-to prove that all theorems are number theory (deductive system for the acceptable stateme~ts found
acceptable-it is sufficient to prove that each axiom is acceptable und in the arithmetic of positive integers). No matter what aXIOms are
the pri~cipal interpretation of the system, and that the rules of inferen::
se1ec t ed , Kurt Godel in 1931 showed how one could construct . an
are valId. The proof that the rules of OS are valid presents no real acceptable sentence of number theory that would be true Iff ~ot
problem. However whether each of the axioms is acceptable is problem- rovable in that system. Godel's theorem showed that mathematIcal
atic. Under the principal interpretation of OS the axioms are: iruth cannot be identified with derivability from any particular set of
Al All a is a. axioms.
A2 Some a is a. There is a second sense of completeness, called absolute complete-
A3 All m is a II all b is m --+ all b is a. nesS. A system is absolutely complete if and only if a nonderivable wff
A4 All m is a II some m is b --+ some b is a. produces an inconsistency if added to the axioms of the system. Few
systems are absolutely complete. In logic, only the s~ntenti~l calcul~s
where a and b are general name variables. A well-formed formula in (and fragments of it) is absolutely complete. There ~s no dIfficulty III
OS is acceptable or valid if on every interpretation for its variables it is a showing that OS is not absolutely complete. OS IS not absolute~y
true statement. Let us SUppose such an interpretation occurs for a complete since there are underivable formulas that do not result III
formula in OS when (i) a non empty domain of discourse is indicated and contradictions if added to the axioms, e.g.,
(ii) . general names in the domain are assigned to the general ~ame
varIables. If we SUppose a formula in OS is valid if true under all such (I) alb --+ (r-.>aAb --+ bAa)
interpretations, then AI, A3, and A4 are all valid. However, A2 (2) bla --+ bAa
presents us with problems, for we can give an interpretation for A2 that The asf's of (I) and (2) are tautologies, so if (I) and (2) were added to
res.ults in a false statement. For example, let V: natural world, a: the axioms of OS, what would result would not be an inconsistent system.
Ulllcorns. If, however, we modify the definition for an interpretation of a But how does one show that (I) and (2) are not derivable in OS? Again,
OS formula by adding that the domain must be such that there is at mere failure to find a proof is not enough to demonstrate that (I) and
least one thing that has a, b, and m (remember rule one allows us to (2) are not derivable. This brings us to the question of the independence
substitute small letters), then A2 is valid. But now 'valid' has a different of a wff of a system from other wffs of the system, and thus to the
sense in OS than it does in predicate logic.
question of the independence of an axiom system.
Another powerful motive in formalizing a theory is to answer In using the axiomatic method, the aim is often not only to establish a
questions concerning the completeness of the theory. An axiom system system in which all the basic features are made explicit, but to establish
"""/-,
 199
J.l10

a simple or nonredundant system. An unnecessary axiom or rule would The other three axioms become:
violate this latter demand. When all the axioms are necessary, that is
A2: F --+ F
when no axiom can be derived from the other axioms, the axioms ar~ A3: (m 1\ F) 1\ (b 1\ m)--+b 1\ F
said to be independent. Since rules can be derived from the axioms and A4: (m 1\ F) 1\ (m --+ b) --+ (b --+ F)
rules of a system, we can also speak of the rules being independent.
. kl be confirmed that under this interpretation Al has the
Independence is not an essential property of an axiom system; in fact
It can qlUC! 1 ' s the other three axioms have the truth-value T.
in many cases a number of dependent axioms make proofs of theorem~ th value 1./ , W lereac
t,I'U -
.
d . d dent Obviously another or different Illter-
easier. Many system builders, nevertheless, value elegance and con- 1 's thus prove III epen .
A I.. ded to prove the independence of A2, A3, and A4.
ceptual simplicity and thus independence is sought after. ConSistency pretatlOn IS nee
is essential; soundness is needed; completeness is most desirable, though
not always obtainable; independence is obtainable, aesthetically pleasing, Exercises
and can reduce one's labors in metalogic. Prove the independence of A2, A3, and A4.
The question of independence has played an important role in the 1. h t Al plus the other three axioms is a consistent set. What does
2. Show tar-.;
history of thought. As mentioned earlier, the parallel postulate was this prove? . ..
regarded by the Greeks as not being as self-evident an axiom as: The Ab cannot be derived from the axioms and that Its addItIOn
3 Showthat aIb --+ a .'
whole is greater than its parts. Attempts were made to prove the . to tIe
I axIO. ms would not result in an IllconsIstent system.
postulate dependent. All of these attempts failed since the parallel u ose a system is consistent if and only if there is at least one well-forme~
postulate is, as has been proved, independent. The efforts to prove its 4. ~or~ula of the theory that is not a theorem. Show that the system below IS
supposed dependence led, however, to the development of non-Euclidean inconsistent: .
geometry. Since such systems have been made use of in the theory of (1) If P and Q are wffs, then P --+ Q IS a theorem.
relativity, and since these theories demanded clarity concerning the (2) Rule of inference: MP.
requirements for a purely formal system, this question of independence 5. Raise and try to answer the three metalogical questions with respect to the
of an axiom has had a significant role in the history ofthought. It should system at the end of Section 8.2.
be noted that if the parallel postulate had been removed from the 6. How do you prove that a rule of a system is not independent?
Euclidean axioms, as many wanted, some of the theorems of Euclidean
geometry that are regarded as expressing true geometric principles Answers
would be underivable. This follows from the fact that the parallel 4. P --+ Q is a theorem. (P --+ Q) --+ P is a theorem. Thus. by MP, P is a
postUlate is independent in Euclid's system. theorem, and if P is a theorem, every wff of the system IS a theorem, and
An axiom can be proven independent of other axioms if all the hence the system is inconsistent.
axioms can be given an interpretation that results in the latter axioms' 6. One way: show that what can be done with the rule can be done without the
having a property that the first lacks, a property that must be passed on
rule. (Recall exercise 4.7.3)
by the inference rules. The usual procedure to show that an axiom is
independent in this way is to provide an interpretation that will make
it false and the rest of the axioms true.
The proof that Al is independent may proceed along these lines. 8.5 Axiom systems and logic
Let
The advantages of a formally developed axiom system ~or any theory
A be 1\
are many indeed. These would be some of the most l~portant ad-
I be--+
a be F (false) vantages: (1) There are no hidden assumptions that may gIVe .us trouble
since everything is explicitly set out. (2) The basis from whICh .all the
Al becomes on this interpretation: theorems follow is perspicuous and thus the system can be ratlOn~lly
evaluated. (3) Deductions are made rigorously, relying only on logICal
AI: F 1\ F
 2UU 201

form, and no appeal is made to intuition. Intuition in this context ents For example the failure to verify a consequence from a
deveIomp ' , . .
would be the process of making inferences in an informal m tt entails the necessity either to modIfy the aXIOms (or any other
. hi' a er theory . I nonmathematical principles use d'm t h e d enva . t'IOns ) or t 0
WIt out re ymg on formal proofs. But intuition is a relative thin ' non1ogwa, .
. .... g, and
re I ymg on mtUItIOn IS not always a safe procedure for avoiding er late new physical entities and forces to explam the apparent
1'01', and
.
postLt'u t between theory and phenomena-th' b e h aVIOr
. b' .
m ~ formally developed axiom system, it need not be relied on. (4) An conulC .
ell' emg m-
u~mterpreted system opens the possibility of providing an interpretation 'ated into the theory through new aXIOms.
corpol . '
dIfferent from the one the system builder intended. (5) Finall tl In passing, it is worth noting that many pllliosophe~s of SCIence argue
. t t . f . y, 1e
lmpor an questIOns 0 consIStency, soundness, and completeness can · the most advanced scientific theories three mam features can be
tha t m . I D" th .
be asked and answered.
(bs lllg uI'shed'. (1) the formal axiomatIC calcu us. rammg e meanmg
. t' .
Most sciences have axiom systems as their ideal though
. . ' axlO-
' of the nonlogical, nonmathematical t~rms in the axioms of ~ phYSICal
matlzat~on has been ~chieved in a relatively few scientific fields such as system by replacing them with arbItrary marks would dIsplay the
mechamcs, mathematIcs, set theory, and logic. In these fields, the initial abstract calculus of an advanced scientific t~eory. H~re we cou~d spe.ak
development was, of course, nonaxiomatic and intuitional . A n axlO- ' tIl e implicit meaning of such marks, meamng by thIS the relatIOnshIps
matization of a field, normally, can only be attempted when the funda- of . d to them in the axioms Terms such as ,mo IecuI' e an d 'k'me t'IC
aSSlg ne " ' . . .,
ment~l no~ions, pro~erties,. and principles are believed known. The first energy of molecules' in physics can be saId to be ImplICItly defined
. by the
step IS to Judge whICh notIOns are basic, which are to be the primitive . s And to a large extent their meaning is exhausted m these
aXIOm.
terms, and which acceptable statements or formulas are basic th implicit definitions. (2) A set of rules assigning empirica.l contexts to ~he
provl'd'mg th'"e lmtlaI set of axioms. ' ese abstract axiomatic calculus. These rules or laws permIt the deductIO~
from the calculus of empirical laws, laws that can be tested by expen-
. In most
' axiomatizations,
.one usually presupposes certain theo rIes, .
a~IOmatlzed theories previously constructed. Thus the empirical ment. Since the nonlogical, nonmathematical rules are premises in
SCIences presuppose mathematics. Customarily, in the axiomatization deriving testable statements, the disconfirmation of such derived
of an empirical science, mathematical terms are not counted as primitives statements does not imply that the axioms must necessarily be altered
or defined terms, nor are mathematical truths axioms or theorems of the or added to. (3) An interpretation of the axioms that provides a model
system. Nevertheless these terms are found in the theses of the system for the abstract axiomatic calculus. These models usually serve to aid
and mathematical principles are assumptions in derivations. Formal in the understanding of the calculus. Since the primitive terms in an
mathematical theories are constructed by making additions to the advanced theory in physics are not given explicit definitions, models
predicate calculus with identity. Only in axiomatized logic is no system are needed heuristically to picture the theory and thus aid in the
presupposed. development of the theory. It is a mistake, of course, to think one must
In a science such as physics, some laws are deduced from other laws have a model, and it is a mistake to assume that models necessarily
for example, Kepler's laws are derivable from Newton's laws. Also som~ have counterparts in nature.
terms are defined in terms of others. Obviously these deductions and Axiomatization in mathematics brings to mathematics all of the
definitions must stop somewhere. All statements cannot be derived in a desirable features mentioned earlier. A further feature, and a timely
system, nor can all terms be defined. This is one motive for arranging the one, is the application of axiom systems to computers. Computers have
true statements in physics and in an axiom system. Another motive is a fixed set of directives programmed into them. These directives
providing a few basic well-corroborated truths or principles that will be correspond to the explicit rules and axioms of a formalized system.
axioms and from which other statements in the field can be deduced. Godel's proof, by the way, is thought by most to have as one of its
The degree of confirmation obtained by the axioms will thus be passed consequences the impossibility of constructing a supercomputer capable
on, to some degree, to the theorems. Since an infinite number of theorems of solving every problem in mathematics.
can be derived, new laws can be derived and experiments performed to All deductions in science and mathematics, whether formal (theorems
corroborate them. Furthermore, since physical sciences develop partly derived in an axiom system) or informal, are moves in which some
in terms of what nature, so to speak, tells the scientist, these new statement logically follows from other statements. A conclusion or step
experiments will reveal new aspects of nature that may prompt new in a proof logically follows from premises or previous lines only if the
-'/ ..
202 Formal Axiom ~_y~",,",,;_

conditional corresponding to the inference is an instance of a

formula in predicate logic with identity.
For mathematics to be fully axiomatized, it must begin by buildin
on a formal system for predicate logic with identity, just as a natul'a~
science to be fully axiomatized must build on a formal axiom system for
mathematics, including set theory. This being the case, there needs to Any system that is consistent and
be some systematic way to generate all valid formulas in predicate aXioms that has all tautologies and
logic. In a word, there needs to be a formal system for predicate log'IC. only tautologies as theorems-as
The natural deduction system POI given in Part I is a formal for the 'tautology' has been defined in this
system for predicate logic. Our interest in Part II is first to give a formal book-may be called a classical
axiom system for predicate logic and, second, to raise metalogical sentential sentential calculus. (See exercises
9.8.7 for a nonclassical sentential
questions.

Exercises
calculus calculus.) The natural deduction
system SO found in Chapter 4 is one
such sentential calculus. At this
1. Wh~, if a theo~y is presented as a formal system, must the presuppositions time we wish to study an axio-
reqUlred for thIS purpose be understood in their full significance and under. matic system for the classical sen-
stood as true?
tential calculus. To do this in a
2. Why are the terms 'primitive' or 'undefinable' and 'axiom' or 'undemon_ formal way we need to list primitive
strable' to be understood only in a relative sense? symbols, give rules for a well-
3. We could define a system or calculus in terms of theorems or in terms of a formed formula, and give an effec-
particular organization. How are 'sentential calculus' and 'predicate tive criterion for what is to count as
calculus' defined in this book? a proof. In addition, we need to
4. Can disagreement arise over this question: Is P a theorem of formal, non- indicate some tautologies that will
logical theory T? be axioms of the system and some
valid rules of inference that will
Answers allow us to make derivations of
2. A term is undefinable and a statement undemonstrable only within a system theorems. The axioms selected will
constructed in a particular way. be the well-known set found in
Russell and Whitehead's Principia
Mathematica. Later in this chapter
the system will be proven consistent,
sound, and complete. We will refer
to the system as PM, after Principia
M athematica. There are, it should be
noted, other possible sets of axioms
which will allow a consistent, sound,
and complete axiom system for
sentential logic (see Exercises 9.8.4).
The axioms for PM will be finite
in number; there are, in fact, only
five axioms in PM. In Chapter 10
an axiom system that corresponds
to PM and that has an infinite

203
204 Axioms for the Sentential >rellimirmry Discussion of PM 205
number of axioms, will be the above formula, obtaining
How this is possible will be
clear at the beginning of Chapter I

A proof in PM is a finite sequence of wffs such that each of the wfIs

9.1 Preliminary discussion of PM
the sequence either is an axiom or is inferred from a preceding formula
formulas in the sequence by means of one of the rules of inference. We
L~ttle preliminary. discu~sion of. PM is needed. A well-formed formula
regard a definitional transformation of A as just another way of
WIll be defined
. as It was III SC WIth one change. The logical conne ct'rves
~ and V wIll be taken as primitive, undefined connectives and th
.<~'t"""l" A. A theorem is the last line in a proof.
others will be introduced by definition. For example --+ will be intro~
We will continue to follow our conventions for eliminating paren-
duced as follows:
The turnstile, 1-, is again used to indicate that a formula is a thesis
(theorem or axiom) of the system when we have I- A.

This definition allows us to write the first lines of each pair below in
favor of the second. 1. Why is every axiom of PM a theorem?
~(P /\ Q) --+ R ~,--.,PVQ
~. Are the notions of well-formed formula, axiom, and proof in PM effective?
~~(P /\ Q) V R ~P--+Q Can a theorem appear twice in a proof sequence?
4. Are Russell and Whitehead to be censured for having an axiom that can be
However, this definition will not allow us to do the following trans-
derived from their other axioms?
formations:
Can a definitional transformation be regarded as an inference on a par with
~(P /\ Q) --+ R PVQ substitution or MP?
(P /\ Q) V R ~P--+Q
6. Which of the following are not correct definitional transformations?
(a) f"o/P --+ Q
since the same shape is not preserved. /\ and <-+ will be :introduced by PVQ
the following definitions: (b) "'(f"o/P --+ Q)
f"o/(,"" r-JP V Q)
A /\ B =df ~(~A V ~B) (c) ",(P /\ Q)
A<-+ B =df (A --+ B) /\ (B --+ A) f"o/f"o/(~P V ,.....,Q)
(d) P V '""Q
There are five axioms in PM. In Russell and Whitehead's system one ",(,""P /\ Q)
of the axioms, namely, (e) ",(P<-+ Q)
",(P -+ Q) /\ ,.....,(Q -+ P)
(f) ",p -+ ~P
P V (Q V R) --+ Q V (P V R) P V ,""P
(g) f"o/(P V '"" ",-,Q)
can be derived later from the other four axioms of PM. Thus this axiom f"o/P /\ '""Q
is not an independent axiom. However, we will include this dependent (h) "-' ,,-,(~P V "-' ~Q)
axiom in the axiom set for PM. The axioms will be expressed using -+ ,,-,(P /\ ,.....,Q)
for purposes of familiarity. Beside the axioms, there will be two rules,
modus ponens and a rule of substitution. The rule of substitution will
allow us to uniformly substitute any well-formed sentential formula for 1. Every axiom is a theorem because the sequence consisting of a single step-
sentential variables. For example, we can substitute ~P/P and "'QIQ the axiom itself-is a proof, and the last formula in any proof is a theorem.
f·,
 206 Axioms for the Sentential Calculus Development of PM 207
4. If ease of proof construction is valued above other considerations, they
should be applauded.
2. No. It doesn't appear in the list of symbols, primitive or defined.
5. Yes, but it will make the discussion to follow slightly more complicated d
nothing much would be gained by it. ' an 3. Yes, but earlier we agreed not to repeat each time we had formation rules
that parentheses are terms of the theory.
6. (a), (d), (e), (f), and (g).

9.2 System PM 9.3 Development of PM

PRIMITIVE SYMBOLS: \Ve now proceed to prove some theorems of PM. The reader is
l. Sentential variables: P, Q, R, ... urged, as usual, to work carefully through the proofs. It is normally only
2. Statement connectives: """, V in this way that one understands the system and acquires the ability
{;o construct proofs.
FORMATION RULES: The first theorem we will prove is:
l. A variable is a wif.
2. If A is a wif, then """A is a wif.
3. If A and Bare ,,,ffs, then (A V B) is a wif.

AXIOMS: Here is the formal proof of T 1 :

l.PVP-+P
2. P-+Q V P P-+QVP
3. P V Q -+Q V P P-+r-...>Q V P
4. (P -+ Q) -+ (R V P -+ R V Q) P-+(Q-+P)
5. P V (Q V R) -+ Q V (P V R)

RULES OF INFERENCE: No justification for the steps is given since the definition of a proof in
l. Rule of detachment (MP): If I- A and I- A -+ B, then I- B. PM makes no provision for such an analysis. Each step can be effectively
2. Rule of substitution: Let M be a single sentential variable and checked to see if it is an axiom or is inferred from earlier lines by modus
N be any wif and let AM be a formula in which M occurs in A. ponens or by substitution. Checking the above sequence, we confirm
Let AN be the result of replacing every occurrence of M in AM that the first line is an axiom, Axiom 2. The second line comes from the
with N. Then if I- AM then I- AN.
first by using the rule of substitution and making this substitution:
"'-JQ/Q. The last line comes from the second by the definition for -+.
DEFINITIONS:
Since this discussion is intended for human beings rather than com-
l. A -+ B =df """A V B
puters, a justification will be supplied, and it will take the form with
2. A A B =df """("",,A V """B)
3. A+--+ B =df (A -+ B) A (B -+ A) which, by this time, the reader is familiar. The proof and analysis of the
proof of T 1 will be
Exercises
l. P-+QvP A2
l. Set down system PM using as the only logical constant the Sheffer stroke
2. P-+ """Q V P From 1 by thc rule of
(see Section 3.6).
inference 2, making
2. Is I- a symbol of PM? this substitution:
"'-'Q/Q
3. Are ( and) symbols of PM? 3. P -+ (Q ->- P) From 2 by df for-+
-"'/,
 n~"Alor)ment of PM
209

T2 (P -+ Q) -+ ((R -+ P) -+ (R -+ Q» T7 r"-' ,.....,P -+ P

1. (P -+ Q) -+ (R v P -+ R v Q) A4 1. P -+ ,......, ----P T6
,-..,RjR 2. ,.....,P -+ "" r-->,....,P ",-,PjP
2. (P -+ Q) -+ (,......,R V P -+,.....,R V Q)
3. T2 df 3. (P -+ Q) ->- (R v P -+ R v Q) A4
4. (2) ->- (5) ,....,PjP, ,...., "" ,-..,PjQ, PjR
5. P V ........,P -+ P V,.....,,-.., ,.....,p 2, 4, MP
The justification for T2 has been shortened. Since ,-..,R/ R on line 2 6. Pv.......,P T5
indicates a substitution, we did not indicate Rule 2. Since 2 comes from 7. P V ,-..."·....-',....,,P 5,6, MP
the preceding line, we did not indicate 1. Only when we use a rule, such 8. P V Q -+ Q V P A3
9. P V "'-' r--..I "",p -+........, "'-' ,......,p V P "" "'" ""PjQ
as MP, that involves more than one line need we indicate lines in the
10. ""-' r-...J """p V P 7, 9, MP
justification. We omitted the name of the definition on line 3, and to
11. T7 df
reduce writing we indicated the theorem by number on line 3 in accord_
ance with the practice we have been following. T8 (P ->- Q) -+ (.......,Q -+ ,.....,P)
1. (Q -+ ,......, ,....,Q) -+
((P -+ Q) -+ (P -+,....." ""Q» T2, QjP, ,-.., ""QjQ, PjR
T3 (P -+ Q) -+ ((Q -+ R) -+ (P -+ R» (Transitivity of 2. Q ->-........,----Q T6, QjP
1mplication) 3. (P -+ Q) -+ (P -+ "" ,-..,Q) 1,2, MP
T4 P -+ P (Identity) 4. "",-,p V ",-,,......,Q-+........,,.....,Q V """p A3, ""PjP, "" ,.....,QjQ
1. PvP-+P Al 5. (P -+........, ,-..,Q) -+ (........,Q -+ ,-..,P) df
2. P-+Q V P A2 6. (3) -+ ((5) -+ (7» T3, P-+QjP, P-+r-..o,-....,
3. P-+ P V P PjQ QjQ,,.....,Q -+ ~PjR
4. (P -+ Q) -+ ((R -+ P) -+ (R -+ Q» T2 3 t 5, 6, MP (2)
7, T8
5. (1) -+ ((3) -+ (6» P V PjP, PjQ, PjR
6. T4 1,3,5, MP (2) The last proof is shortened by entering needed axioms and theorems
with the desired substitutions.
Line 4 is not an axiom nor does it come from the previous lines in the
sequence by a rule or definition. T2 at the justification place on line 4
T9 C. .·. -'P -+ """Q) -+ (P -+ Q)
TI0 P V (Q V R) ->- (P V Q) V R
means: "At this point insert the proof for T2." Thus we do not have a 1. QvR-+RvQ A3
formal proof but a proof outline. To lessen our proof burdens we will 2. (Q V R -+ R V Q) -+
frequently resort to such outlines. (We are in effect operating with the (P V (Q V R) -+ P V (R V Q» A4
1,2, MP
derived rule theorem introduction of se.) Using line numbers on line 5 3, P V (Q V R) -+ P V (R V Q)
4, P V (R V Q) -+ R V (P V Q) A5
saves time. It also helps to make clear how T2, substitution, and MP T3
5. (3) -+ ((4) -+ (6»
can do the work of a rule corresponding to Trans, namely: If f- A ->- B 6. P V (Q V R) -+ R V (P V Q) 3, 4, 5, MP (2)
and f- a -+ A, then f- 0-+ B. 7. R V (P V Q) -+ (P V Q) V R A3
8. (6) -+ ((7) ----;. (9» T3
T5 P v,-.."p 9. TlO 6,7,8, MP (2)
1.P-+P T4
2. ,-..,Pv P
To shorten the above proof we have omitted indicating the substitutions
df
3. P V Q -+ Q V P A3 made for the axioms and T3.
4 . .......,p V P ->- P v""P ""PjP, PjQ
5. T5 2,4, MP TIl (P V Q) V R -+ P V (Q V R)
T12 P -+ (Q -+ P A Q) (Conjunction)
T6 P-+r"-''''-'P 1. (""P V ........,Q) V r_;(r-"P V r-...;Q) T5
1. P v........,P T5 2. (1) -+ (3) TIl
2. "",P v........,,.....,P ,-..,PjP 3 . .......,p V (,-...,Q V ,-.....,(o--'p V,-.....,Q» 1, 2, MP
3. T6 df 4. T12 df (3)
 Axioms for the Sentential 211

We may derive biconditional theorems using T12 and ha to shoW is that the rule is dispensable. This is shown as follows:
proven A ->- Band B ->- A. This procedure is illustrated in the proof·
of the next theorem. For DR1
Given as an assumption in DRl
LA
Given as an <lssumption in DR1
T13 P+--+,....., I"-'p 2. B
1. P ->- ,....", ,-,..,P T6 3. P -+ (Q -+ P A Q) T12
2. r-..J,,-,p-+ P T7 4. A -+ (B -+ A A B) A/P, B/Q
5. A A B 1,2,4, lVIP (2)
3. (1) -+ ((2) ->- (4)) T12
4. (1) A (2) 1, 2, 3, MP (2)
5. Tl3 df 'j' t tl 's shows is that whenever we move from A, B to A A B, by
\Vw 1 t11 DR.1 we can convert tIns
. '
move lllto a proo f 0 fA A Bb Y USlllg
.

Exercises
11ppeasu~stitution, and modus ponens twice. It will prove helpful to
delllonstrate a second derived rule.
1. Prove the unproven theorems. It may be found convenient to prove some
relatively simpler theorems prior to proving one of the listed theorems. DR2 If f- A -+ B <lnd f- B -+ C, then f- A -+ C
reader is encouraged to do this. 1 (P -+ Q) ->- ((Q -+ B) -+ (P -+ B)) T3
2: A -+ B -+ ((B -+ C) -+ (A -+ C)) AlP, BIQ, CIB
3 A-+B A
4: (B -+ C) -+ (A ->- C) 1,2 MP
5. B-+C A
9.4 Derived rules 6. A ->- C 4, 5, lYIP

The proof of theorems is made considerably easier by using derived rules rf3 is derived from the primitive basis of PM, so the .proof here is .a p~oof
(DRs). A derived rule is merely a substitute for what can be carried out tl t what is done with DR2 can be carried out by USlllg T3, substItutIOn,
by using the axioms and primitive rules. The use of derived ~'ules thus 1a mocl1lS ponens twice . The
nnc next l)roof makes use of this derived rule.
1 .

does not increase the class of provable formulas. They in fact produce
more proof outlines. T14 P A Q ->- P
1. ,.....,P -+ ,.....,Q V ,....,P A2
The justification for derived rules is similar to the justification for A3
2. '"'-'Q V ",-,P -+ ,....,P V ,.....,Q
definitions. A definition is a convenient way to write in shorter form a 3. ,...."P -+,.....,p V ,.....,Q 1,2, DR2
longer expression. Similarly, a derived rule allows us to do in fewer 4. (3) -+ (5) TS
steps exactly what we could do using the primitive rules and axioms of 5. ,.....,(,.....,P V ,.....,Q) ->- o--',......,p 3,4, MP
the system. This means that whenever a derived rule is used in a proof, 6. ,-..,; I"-'P ->- P T7
we could give the unabridged proof if we wished. The proof of a derived 7. r_+__ 'P V I"-'Q) -,.. P 5,6, DR2
S. T14 df
rule is thus, in effect, to set down all the machinery needed to make
unabridged proofs. If one wished to dispense with any use of a derived
TI5 P A Q-+Q
rule in a particular proof and write it out in full, all that would be needed TI6 (P -+ (Q -+ B)) -+ (Q -+ (P ->- B))
is the machinery employed in the proof of the rule. 'DRn' in a proof is to TI7 PvQ+--+,.....,P-+Q
be read: "Followtheproof ofDRn to convert this lineA into a proof of A." TIS (P+--+ Q) -+ (r-..P<.-> r-..Q)
Let us take as our first derived rule Conjunction: TI9 (P -+ Q) -+ (P vB ->- Q V B)
T20 (P -+ (Q -+ B)) -+ (P /\ Q ->- B)
T2I (P -,.. Q) A (P ->- R) -+ (P --+ Q /\ R)
DR1 If f- A and f- B, then f- A A B. (Conjunction)
In some of these proofs we find, for example, the need of commuting
To demonstrate a derived rule, or to show a rule is a derived rule, all we disjuncts. For example, the proof of T19 would be quite simple if we
 212 213

could go from A4 one may conclude that all the positive integers plus ° have the
erty F in the first case and that all the positive integers have F in
prop
(P ---+ Q) ---+ (R V P ---+ R V Q) the second case. ..,
to Proofs using mathematIcal mductlOns often can be used when the
(P ---+ Q) ---+ (P V R ---+ Q V R) tement to be proven involves the natural numbers. Theorems in logic
sa
t
can involve numbers in a variety of ways. For example, a theorem may
Even if we obtained the derived rule: If I- A V B, then I- B V A, this be about 1, 2, 3, ... , n variables in a formula or it may be about n
would not allow the switching of Rand P and Rand Q in A4. It would number of logical connectives. Or a theorem may be about n number of
allow us to go from, say, mptions or about n number of steps in a proof. Frequently,
aSS U
ma,thematical induction may be used to establish such statements.
(P II Q) V R 1'he proof of the equivalence theorem is by induction on the number
to n of statement connectives in AM. We suppose that only the primitive
R V (P II Q) connectives are employed.
Basis: If n = 0, then AM is M. Then AN is N, and (M +-+ N) ---+ (AM +-+
but it would not allow us to alter the inner structure of a wff. A rule AN) by T4.
that would allow us to alter the inner structure of a wff by replacing B
°
with when B +-+ 0, would be useful in lightening our burden. We need
Induction step: Assume the theorem holds for every formula with n or
fewer connectives and consider AM with n + 1 connectives, where n is
a replacement rule for PM. One more additional theorem is needed to greater than 0. The + 1 connective must be either ""-' or v. Thus the
prove this desired derived rule, namely two cases below exhaust the possibilities for AM.
CASE 1: AN is ,......,BM. By our assumption (the theorem holds for every
T22 (P+-+ Q) II (R+-+ S) ---+ (P V R+-+ Q V S) formula with n or fewer connectives):
To prove the needed derived rule we must first prove what is called
OvI +-+ N) ---+ (BM +-+ BN)
the equivalence theorem. To state the theorem let M and N be formulas.
Let AM be a formula in which M occurs. Let N be the result of re-
placing one or more occurrences of M in AM by occurrences of N , then where BN is the result of replacing the occurrences of }JII in B1V] by
occurrences of N. Then
THE EQUIVALENCE THEOREM:
(JY1 +-+ N) ---+ (AM +-+ AN) by TIS

We will make use of (strong) mathematical induction in the proof of CASE 2: AM is BM V OM. Let BN and ON be defined as in case 1. By our
this theorem and some later theorems. Mathematical induction is often assumption:
used to prove theorems in number theory. For example, if one shows
that
(JY1 +-+ N) ---+ (BM +-+ BN)
Basis (i) 0 has property F, and (JY1 +-+ N) ---+ (CM +-+ CN)
Induction step (ii) for any positive integer n if every integer
equal to or less than n has F, then n + 1 '['hen (M +-+ N) ---+ (BM V OM +-+ BN V ON) by T21 and T22
has F,
or if one shows }1'rom the equivalence theorem we may prove the desired replacement
theorem. Let M, N, AM, and AN be as in the equivalence theorem.
Basis (i) 1 has the property F, and
Induction step (ii) for any positive integer n, if every integer
equal to or less than n has F, then n + 1 'l'HE REPLACEMENT THEOREM (R):
has F, If I- 111 +-+ Nand f- AM, then I- AN
214 Axioms for the Sentential Calculus Consistency and Soundness of PM 215
Proof: vVo have already had occasion to review why consistency in the first,
1. JvI +-t N A the negation sense, is an essential property for a sentential calculus
2. AM A
(Section 4.8). If PM is negation inconsistent, then any wff is a theorem.
3. (111 +-t N) --+ (AM +-t AN) Equivalence theorem
4. AM+-t AN 1,3, MP Consistency in the other two senses also prevents a system from
5. AN 2, 4, df, T14, MP being useless. In any formalization of a field only some of the possible
wffs-those regarded as acceptable-are desired theorems. If a system
Exercises were absolutely inconsistent, then every wff would be a theorem and thus
the system would be trivialized. And if a system were inconsistent in
1. Without using replacement, prove the unproven theorems through T22. If
Post's sense, then given the rule of substitution, it would also be the case
necessary, you may prove other theorems in order to prove the numbered
theorems. Prove some of the 55 theorems found in Chapter 4 that have not that every wff would be a theorem.
been proven so far, using replacement if appropriate. However, do not Use There are good reasons why consistency in the second sense is called
a biconditional theorem with replacement unless it has been proven. absolute consistency. First, a formal theory might not have a negation
2. Select some of the proofs' outlines using derived rules, and construct a sign as one of its terms. If so, it could not be negation inconsistent and
formal proof not using any derived rules. thus it would be vacuously negation consistent. In turn, a formal theory
with no sentential variable as a term would be vacuously consistent in
3. A derived rule corresponding to RCP can be proven for PM. It is usually
called the deduction theorem and may be stated as folluws: Post's sense. However, a formal theory cannot be vacuously absolutely
If /'-" P f- Q, then /'-, f- P --+ Q consistent. Second, a system that is vacuously consistent in the negation
This reads: "if there is a deduction of Q from /'-, (a set of formulas which may and Post's sense could be a totally useless system if it were absolutely
be empty) and P, then there is a deduction of P --+ Q from /'-,." A ded'llction inconsistent, since a formal theory with every well-formed formula a
here is a finite sequence of formulas such that for each formula at least one theorem is useless. However, if a system is absolutely consistent, it is
of the following holds: saved from such a fate. Third, if a formal theory is absolutely consis-
(1) It is an axiom or a substitutional instance of an axiom.
(2) It is an assumption. tent, then it must be negation consistent, if P /\ ,,-,P --+ Q is a theorem,
(3) It is inferred from the preceding formulas by MP. and consistent in Post's sense. This can be proven by proving:
Use the deduction theorem in constructing a proof for a few of the 55
theorems found in Chapter 4. (1) If a system is negation inconsistent, then it is not absolutely
consistent.
(2) If a system has a theorem that consists of a sentential variable
then it is not absolutely consistent.

9.5 Consistency and soundness of PM If a system is negation inconsistent, then any wff is a theorem if
p /\ ,-..,p --+ Q is a theorem. Thus (1) is true. If P is a theorem, then any
There are three senses in which an aXIOm system can be said to bo wff is a theorem. Then (2) is true.
consisten t: We shall now proceed to prove three consistency theorems. We wish
to prove that PM is negation consistent, absolutely consistent, and
(1) Consistency with respect to negation, or simple consistency: A consistent in the sense of Post. To prove PM negation consistent, we
system is consistent in this sense if no well-formed formula and first will prove that every theorem of PM is a tautology. For if every
its negation are both theorems of the system.
theorem is tautological, then A /\ """A cannot both be theorems.
(2) Absolute consistency: A system is absolutely consistent if not
every wff of the system is a theorem. As previously indicated, a system is sound when all its theorems are
(3) Post's sense of consistency:2 A system is consistent in this sense acceptable. Since tautologies are the acceptable formulas of PM, to
if there is no theorem in the system consisting of a sentential prove that every theorem of PM is a tautology is to prove the soundness
variable alone. of PM.
To show that every theorem in PM is tautologous, we interpret
2 The American logician E. L. Post introduced this sense in 1921. P, Q, R, ... as sentential variables having only two values true or false,
-,.
216 Axioms for the Sentential Calculus Independence of the Axioms of PM 217
and we interpret '"'-' and V as statement connectives defined by the Theorem 1 establishes the soundness of PM. Thus:
standard truth-tables. On this interpretation, every axiom of PM is a
tautology and the rules of inference applied to tautologies give a Theorem 2: PM is negation consistent.
tautology.
Lemma 13: Every axiom of PM is a tautology.
Theorem 3: PM is absolutely consistent.
Proof: Standard truth-tables show that each axiom takes the value T
for every assignment of truth-values to its components. Proof: Select any axiom, say AI. ,.....,AI is not a tautology by Lemma
Lemma 2: M Od11S ponens applied to two tautologies gives a tautology. 1. Thus ,......,Al is not a theorem of PM by Theorem 1.

Proof: Suppose A -+ B and A take the value T for every assignment to Theorem 4: PM is consistent in the sense of Post.
the components A and B. Then B cannot have the value F since if
Proof: Let A be any wff consisting of a single sentential variable. Then
A-+B
A is not a tautology, and by Theorem 1 is not a theorem of PM.
TT
occurs, B must take the value T. Exercises
Lemma 3: The rule of substitution applied to a tautology gives a
tautology. 1. Show that if
P-+Q
were added to PM, the resulting system, PM', would be inconsistent in all
These lemmas establish:
three senses of inconsistency.
Theorem 1: Every theorem of PM is a tautology. 2. Show that a sentential calculus is negation consistent iff it is absolutely con-
sistent.
Proof: If every axiom is a tautology and the rules applied to tauto-
3. In what sense is a system without ,-...J consistent?
logies can only give tautologies, then every theorem is a tautology since
a theorem is the last line of a sequence of which each line either is an 4. Show that if a wff A were added to the axioms of PM and the resulting set
axiom or results from applying a rule. Using mathematical induction we alone were inconsistent, then ""A would be a theorem of PM.
can precisely set out this argument:
Proof: Let Av ... , An be a proof. We prove by induction on n that
every step in a proof is a tautology. 9.6 Independence of the axioms of PM
Basis: If n = 1, then Al is an axiom. By Lemma 1 every axiom is a
tautology. An axiom A of a theory T is independent iff A is not provable in T
Induction step: Suppose that for every proof with n steps, every step from the other axioms of T. A dependent axiom does no harm; it is
is a tautology and consider a proof At> . . . ,An' A)/+l with n + 1 merely not needed.
steps. By the induction hypothesis At> . .. ,A)/ are all tautologies. Earlier it was pointed out that A5 of PM is not an independent
An+! is an axiom or is inferred by modus ponens or substitution. If axiom. A5 can thus be proven as a theorem using the other four axioms.
A)/+l is an axiom, then An+l is a tautology by Lemma 1. If An+l is To prove the independence of the first four axioms of PM, we cannot rely
inferred from Ai and Aj by modus ponens, then An+! is a tautology by on our failure to derive each axiom from the remaining ones, for such
Lemma 2, and if A"+l is inferred from A.I by substitution'then A 1 is a failure may be due to lack of ingenuity or luck. We must use another
n+
tautology by Lemma 3. method.
Up to this time, to give an interpretation for a formula in the
3 A lemma is a preliminary or auxiliary theorem demonstrated or accepted for sentential language has meant to assign truth-values to its components.
use in_-l·,a demonstration of a main theorem. vVe can also speak of giving an interpretation for the primitive symbols
 Independence of the Axioms of PM 219
218
Given the above interpretation, the value of the axioms can now be
of PM. In fact, the following is the pTincipal interpretation oj P)J!I:
determined in a way analogous to determining the truth-value of a
(1) We understand P, Q, R, ... as having the values true formula on the standard truth-tables. First we employ the definitions
or false. rewrite the axioms in terms of "'-' and V as follows:
(2) We give a standard truth-table interpretation for the
two primitive constants "-' and V.
Al ",-,(P V P) V P
A2 <"""--'P V (Q V P)
:Under this. interpretation. e~~h axiom is tautologous. However, other ..-.-;(P V Q) V (Q V P)
A3
mterpretatlOns of. the .prImItIVe symbols are possible . Now sa Y tllat r---'("'-'P V Q) V (~(R V P) V (R V Q))
A4
under such an . mterpretatlOn we find that axioms A2 ' A3 , n a d A4
possess a Cel'tam feature. And let us also suppose that the primitive We next determine the value of Al as follows:
rules pass on this feature. This being the case, all the theorems derived
from A2, A3, and A4 must have this feature. But suppose Al lacks the "'-' (P V P) V P
feature. This would mean that Al is not derivable from A2, A3, and A4
in this system and thus is an independent axiom in the system . Tl·. - lIS I 0 0 0 0 0
then, is the method which will be used to prove the independence of th~ 0 1 1 I 0 I
1 2 0 2 2 2
first four axioms of PM. W·e cannot, of course, use the principal inter-
pretation of PM. We need a different interpretation. Such an inter-
pretation will look strange, but it can be as strange as we like. Under the main connective of Al we find 0, 0 and 2. A2's value IS
The independence of Al from the other three will be proved by means determined by this table:
of an arithmetical interpretation. The variables will take the values 0 1
P V Q V P)
and 2, and the signs "-' and V are interpreted by these tables: "

,...,.,p I 0 0 0 0 0
P P Q PvQ 1
0 1 0 0 0
2 2 0 0 0 2
0 I 0 0 0
I 0 0 I 0 0
I 0 I 0 0
0 I 0 I I 1
2 2 2 0 0
2 2 0 I 2 2
0 I 0
I 0 0 2 0 0
I I I
0 I 0 2 2 I
2 I 2
2 2 0 2 0 2
0 2 0
1 2 2
2 2 0 As we can see, under the main connective we have all O's. A2's value is
thus O. The reader, as an exercise, can confirm that A3 and A4 are also
which can be more conveniently written as: O-formulas. This property of having a O-value is passed on by the two
rules just as having a T-value is passed on. Thus all the formulas derived
V 0 I 2 fi'om A2, A3, and A4 have the same value. Consequently, Al is inde-
pendent of the other three axioms. In the exercises immediately to

-ciT
I
2
0
2
0
1
2
0
0
0
0
I
2
0
2
0
follow, values and matrices to define ~ and V are to be devised that
prove the independence of A2, A3, and A4, and thus complete the proof
-of the theorem below.
These tables provide an interpretation for ,-...., and V in the same manner
Theorem 5: The first four axioms of PM are independent.
as the standard truth-tables do this for these connectives.
 220 Axioms for the Sentential
Completeness of PM 221
Exercises
every formula in 6 is a theorem-where 6 is the class of acceptable
1. Prove the independence of the remaining axioms. Let us continue to call this completeness relative completeness.
2. Show that the following is not a thesis of PM: Since, under the principal interpretations, we wish all tautological form-
(P -+ Q) A Q -+ P ulas to be theses, PM is complete in the relative sense if and only if:
3. Dependent axioms do no harm' they are merely superfluous S
.. '
h owe vel', a deCISIOn . orne t'Imes
whether an axiom is independent can have' , Theorem 6: All tautologous wffs of PM are theorems of PM.
. Important
consequences. How does the brief account of the history of geometr" .
Chapter 8 illustrate this point? y gIven In
If a system is relatively complete, then the system is sufficient to
4. Why ca~not the principal interpretation be used to prove the independen generate the set of all its acceptable wffs. This suggests that we might
of an aXIOm of PM? ce distinguish a further sense of complete and say that a system is com-
5. Derive A5 of PM from the other four axioms . Use no theorems 0 l' d enved
. plete iff an inconsistency will result if it is made more powerful. In other
ruI es t h at h ave been proven by using A5. words, if any underivable wff is added to the axioms, an inconsistent
system will result. This is called absolute completeness. Absolute
6. Try to provide an interpretation for the primitive symbols of PM that
results in the axioms being (a) true statements and (b) names. completeness is not a relative notion since a system is consistent or
inconsistent independent of an interpretation. PM is absolutely com-
Answers plete if and only if an inconsistency arises when there is added to the
1. We can demonstrate the independence of A2 from Al and A3 and A4 by . axioms a formula not provable. Since PM can be inconsistent with
the following interpretation: usmg respect to negation, absolutely or in the sense of Post, there are three
corresponding senses of absolute completeness for PM.
v o 1 2 3 There are several ways to construct a proof of relative completeness.
One way rests on our ability to construct truth-tables for well-formed
o 3 o o o o o formulas in PM; another rests on our ability to transform any well-
1 2 1 o 3 o 3
formed formula in PM into a certain kind of form. We will use this
2 1 2 o o 3 3
3 o 3 o 3 3 3 second method. To prove PM complete in the second way we must first
show that every well-formed formula can be transformed into con-
6. (~l Let P, Q, R, ... denote 1, 2, 3, ... respectively, V be
A +, and r-...J A be junctive normal form (CNF). Every wff of PM can be put into CNF by
vA. using certain combinations of the following: PM definitions, rule of
replacement, double negation law, distribution laws, commutation laws,
and association laws. All of this is available in PM. We are able to
9.7 Completeness of PM establish the lemma below:

In most general terms an axiomatic system is complete if it is suffi-

Lemma 1: If A is a wff in PM, then there is a wff in PM, A I , in CNF
ciently powerful to make possible the derivation of all the acceptable
such that r A +-+ A'.
well-formed formulas. The point of PM is to provide a system in which all
tautologies are theorems. The point of having PM is to have a con-
sistent system in which we set down a few valid formulas and a few However, we will ignore the proof of this lemma. A second lemma is
rules, and from which all valid formulas can be generated. Let us now needed, namely:
consider the question of the completeness of PM.
There are several definitions for completeness. To sayan axiom Lemma 2: Every tautologous wff of PM in CNF is a theorem of PM.
system is complete, as just indicated, can be taken to mean that, given the
principal interpretation, the system is sufficiently powerful to generate This is the_fundamental lemma to proving the relative completeness of
all its acceptable wffs. A theory is complete relative to D. if and only if PM.
-,.
 i<n<..ml,~terless of PM 223
222
a conjunction B containing no mutually contradictory com-
All tautologous ONF formulas will contain in every conjunct
Let us suppose B is
mutually contradictory disjuncts. If a formula in ONF is tau
each conjunct must be valid, and this can be possible only if in P V r--'Q V R V r--..JS
conjunct we have a variable appearing as one disjunct along with
negation as another disjunct. we substitute r"-'P for all the negated variables and a P for the non-
If we can show that PM allows the derivation of ccny t variables, then we obtain:
.,,,,llUAc'U

formula in ONF, Lemma 2 will be established. The following LVLLeUbl(Y P V r--' ,.....,P V P V ,......, r--'P
indicate that any formula in ONF that is a tautology is provable in
DN and Al we obtain from this the wff P. Notice that if B were a
contained a sentential variable both negated and unnegated,
(1) We can derive P V ,......,P, and by substitution we can replace we could not make this kincl of substitution.
by any wff. It is now a simple matter to show that PM + A is inconsistent in all
(2) By using A2, P -+ Q V P, or the DR: If I- A, then I- A V senses. It is negation inconsistent since for P we can substitute
we can add any formula to P V,......,P. /I r-'P. It is absolutely inconsistent since by substitution for P every
(3) By repeated use of (1) and (2) and conjunction, we can
would be a theorem. And it is inconsistent in Post's sense since P
any tautologous ONF. We then, if necessary, can Use O·HOIII''''c>.,
ment, association, and commutation to insure a perfect match, a theorem.
"Ve may note that one can also show that

Let us consider an example to illustrate these steps: PM minus Axiom 1

minus anyone of the other independent axioms) is both relatively

Derive: (,......,P V P) /\ ((R V Q) V,......,Q) absolutely incomplete. We have already established that Axiom 1
1. P V,......,p theorem
an independent axiom. Thus Al cannot be derived from PM minus AI.
2. Q V "-'Q Q/P
3. (Q V ,......,Q) V R addition Al is a tautology, it follows that PM - Al is not relatively
4. (P V ,......,P) /\ ((Q V ,......,Q) V R) 1, 3, Oonj complete. Since Al cannot be derived from PM - AI, its addition as a
5. (,......,p V P) /\ ((R V Q) V,......,Q) Comm, Assoc, replacement would not produce a contradiction in any of the three senses.
the addition of Al would not result in an inconsistency, PM - Al
Theorem 6 follows Lemmas 1 and 2. If for any wff I- A +--> A '(Lemma not absolutely complete.
and if all tautologous A' are derivable (Lemma 2), then all
wffs are derivable in PM, for I- A follows from I- A +--> A' and I- A'.
Let us now see that PM is complete in the absolute sense, that Show that
inconsistency arises when there is added to PM a nonderivable formula. P-+(Q-+p/\Q)
Since we have distinguished three senses of consistency, what we is a thesis of PM by;
now to show is: (a) transforming it into CNF,
(b) proving each conjunct and then using Conj to get the CNF as a theorem,
(0) reversing the transformation of CNF to obtain it.
Theorem 7: PM is absolutely complete with respect to negation
Show that PM plus
consistency, absolute consistency, and in Post's
(P-+Q)-+ (Q-+P)
of consistency. is inconsistent 9Y using CNF and substitution.
Proof: Let A be any formula that is not provable in PM. Let A' be
If A is added to PM and PM + A is inconsistent, is ,......,A a theorem of PM?
ONF of A. Since A' is not provable, it is not a tautologous formula, An absolutely complete system has been described as a balloon filled to
capacity that will explode if any attempt is made to enlarge it. Why is the
in proving metatheorem 6 we have proven that all tautologous metaphor apt?
formulas of PM are theorems of PM. Since A' is not tautologous, it
 225
224 Axioms for the Sentential Calculus 9.8 Completeness of SC
e go about proving the soundness, consistency, and complete-
5. Prove that PM is absolutely complete if there is no wff of PM that is inde_ S HowwOUld on . 1
pendent of the axioms of PM. . neSS 0 f tIle tree-method for the sententIal calculus.
w that these axioms for the sentential calculus make up
6. Oan a theory be vacuously relatively complete~ 4. HoW wou ld you sho
a complete set:
Answers Al P -+ (Q -+ P)
A2 (P -+ (Q -+ R)) -+ ((P -+ Q) -+ (P -+ R))
3. Not necessarily; e.g., consider adding P to the axioms of PM. A3 (~P -+ ~Q) -+ (Q -+ P)
4. An inconsistent system is like a broken balloon-both are useless.
With MP and the rule of substitution added to the axioms in (4) derive a few
5.
theorems.
9.8 Completeness of SC oU show that A3 in (4) could or could not be replaced by
6 How would Y
• (~P -+ ~Q) -+ ((~P -+ Q) -+ P)
One of the matters that have been postponed from Part I is the proof of without altering the class of theorems~ Oan it replace A3~
the completeness of the natural deduction for sentential logic SO. We
7. The following is a set of axioms for intuitionistic sentential calculus.
are now in position to prove that SO is both relatively and absolutely
1. P-+(Q-+P)
complete. 2. (P -+ (Q -+ R)) -+ ((P -+ Q) -+ (P -+ R))
The simplest way now at hand to prove SO relatively and absolutely 3. P /\ Q -+ P
complete is to prove that all the theorems of PM are theorems of SO. 4. P /\ Q-+Q
This can be quickly disposed of. All the theorems of PM are theorems 5. P-+(Q-+P/\Q)
of SO since all the axioms of PM can be proven as theorems in SO; the 6. P-+ P V Q
rule of inference in PM, MP, is a rule in SO; and, finally, the rule of 7. Q-+PVQ
8. (P -+ R) -+ ((P -+ R) -+ (P V Q -+ R))
substitution in PM is in effect presupposed in SO since the rules are 9. (p-+Q)-+ ((P-+~Q)-+~P)
formulated in terms of symbols that are variables for any wff of senten- 10. ~P-+ (P-+Q) .
tiallogic. (a) Show that the two theorems below are not theorems of thIS system.
The direct way to prove the relative and absolute completeness of PV~P
SO can proceed along lines strictly analogous to the completeness proof ,......,P-+P
(b)
How would one show that the intuitional calculus is a subsystem of PM-
for PM. A review of the theorems of SO proved in Ohapter 4 reveals that all the theorems of intuitional calculus are theorems of PM, but not the
Lemma I holds for SO if it holds for PM. The proof of Lemma 2 for SO other way around ~
is the same as the proof for PM since the three steps outlined in the last
section can be followed in SO. Finally, the proof that SO is absolutely
complete again can proceed along the same course as the proof that PM
is absolutely complete.

Exercises
1. Fill in the details of a direct proof of the completeness of SO.
2. Use mathematical induction in constructing a proof of the soundness of SO.
Let AI' ... , An be a proof in SO. Prove by mathematical induction on the
number of sentences in a proof. Show that the first sentence is a tautological
consequence of its assumptions, and that if every sentence appearing on
lines through n is a tautological consequence of its assumptions, then the
sentence line n +
I is a tautological consequence of its assumptions. If aJl
sentences appearing on a line are tautological consequences of their assump-
tions, then a theorem must be a tautology since it is a tautological conse-
quence of no assumptions.
~'h
 10.1 Preliminary Discussion of LPC

1
Two axiom schemes corresponding
to identity elimination and intro-
duction will be added after a partial
development of the predicate cal-
culus. The result is an axiomatic
Any system that is consistent and system for predicate logic with
aXioms that has all tautologies and valid identity.
predicate formulas and only tau-
for the tologies and valid predicate formulas
a~ theorems may be called a pre- 10.1 Preliminary discussion of LPC
predicate dICate calculus. The natural de-
duction system PC found in Chapt In the axiom system for the sentential calculus, the first axiom is
6 . er
calculus IS such a calculus. In this chapter
PVP~P
we wish to formulate and partially
with identity develop the predicate calculus as an
axiomatic system. To do this we Using A t.o stand for any well-formed formula in sentential logic, the
shall add t~ PM a r~le corresponding axiom scheme corresponding to this axiom is
to the ulllversal mtroduction and
two new axioms, one related to AvA~A
changing the scope of a universal
quantifier and the other corre- This scheme determines an infinite number of axioms, namely all those
sponding to universal elimination. sentential formulas that are instances of A V A ~ A. For example,
Later in ~his chapter the resulting instances of this scheme would be
system WIll be proven consistent
and an outline of the proof for
PVP~P
soundness and completeness will be QvQ~Q
given. (P /\ Q) V (P /\ Q) ~ P /\ Q
It will prove convenient to make
some other changes in PM. The
and each will count as an axiom of a system that has A V A --+ A as an
rule of substitution will be elim-
axiom scheme. Thus in setting down the basis of a system if A V A --+ A
inated. This will not affect the
derivation of theorems since the is given as an axiom scheme, then this would mean that the system had
axioms of PM will be restated in an infinite number of axioms, namely all instances of the scheme.
terms of what are called axiom A V A ~ A will in fact be one of the axiom schemes of LPC. In LPC,
schemes. The two new axioms will however, A, B, and G will be metavariables for formulas in both sentential
also be stated as axiom schemes. In logic and predicate logic. Of course other axiom schemes besides
effect, this change will result in a A V A --+ A are needed to achieve a complete sentential calculus. But
system that has an infinite number each such scheme will determine an infinite number of axioms. One
of axioms rather than a finite reason why axiom schemes are preferred for the predicate calculus is that
number. How this is achieved and having schemes does away with the need for a rule of substitution that is
why this is desirable is discussed cumbersome to state for predicate logic.
below.
Formulas-in LPC will be made up of sentential variables, statement
The system for the predicate
calculus will be given the name connectives, predicate variables, individual variables, and quantifiers.
LPG (after lower p1'edicate calculus). The formation rules for a well-formed formula will be almost the same as
those set down in the last section of Chapter 5. We will not need
226
 Axioms for the Predicate Calculus with Identity 229
10.1 Preliminary Discussion of LPC

individual constants. As will be indicated in a moment, forgoing the Use We are now in position to add our restriction to (1) so that it will
of individual constants will make necessary provisos so that individual become a valid scheme. Below is the needed valid scheme:
variables do not become accidentally bound. Term now is exclusively used
to indicate an individual variable. (4) 'IvA -+ A(tfv) provided t is free for v in A
We will use some of the notation used in Chapter 6. Let A be a wff,
va variable, and t a term. Then A (t/v) is the formula that results when (4) is in fact one of the axiom schemes we will use, and now neither (2)
each free occurrence of v in A is replaced by an occurrence of t (note the nor (3) is an instance of (4). .
addition of 'free'). Suppose we are given this scheme which corresponds Another axiom scheme of LPC wIll be
to UE in PC:
(1) 'IvA -+ A (tfv) (5) 'v'v(A -+ B) -+ (A -+ 'v'vB) provided v is not free in A

The universal quantifier 'Iv, appearing in the antecedent, has been

Instances of (5) are
dropped in the consequent and a term is substituted for v in A. Instances
of this scheme are 'v'y(Fx -+ Gy) -+ (Fx -+ 'v'yGy)
'v'xFx -+ Fx 'v'x(P -+ Fx) -+ (P -+ 'v'xFx)
(2) 'v'x3yFxy -+ 3yFyy 'v'y(Fx -+ 'v'z(Gxy -+ Hy)) -+ (Fx -+ 'v'y'v'z(Gxy -+ Hy))

Scheme (1) as it stands is not a valid scheme. This means that some Note that P here is either Tor F, so there can be no free variables in P.
instances of (1) are not valid formulas, e.g., (2) above. The invalidity of The next formula is not an instance of (5) since v is free in A:
(2) is demonstrated given this interpretation: V: positive integers, Fxy:
x < y. Another instance of (1) that is invalid is:
'v'x(Fx -+ Fx) -+ (Fx -+ 'v'xFx)
(3) 'Ix I " - ' 'v'yFxy -+ ,......,'vyFyy
The invalidity of this last formula is demonstrated with the following
If we suppose V: {I, 2} and Fxy: x is identical to y, (3) expresses a false interpretation, supposing that Anderson is a poor student: V: human
statement. To rule out such invalid instances we need to add a proviso beings, x: Anderson, Fx: x is poor.
to (1) to make it a valid predicate scheme. The definition for a theorem is the same as in PM.
In these last two examples, when the term is introduced for v it The universal quantifier will be taken as primitive and the existential
becomes bound by a quantifier in A. In PC such accidental binding of quantifier will be introduced by definition. The axi~m schemes for the
individual variables could not occur since in the quantifier rules use is propositional calculus will be the schemes correspondmg to the first four
made of individual constants; t in PC is a variable for individual con- axioms of PM.
stants. Now t is a variable and only a variable for individual variables, Though we will rule out as well-formed formulas, vacuously quanti-
and accidental binding becomes a menace. fied formulas we will allow formulas such as 'v'x(Fx -+ 'v'xFx). One
To introduce the proviso necessary so that individual variables do reason these ~ere excluded from the class of well-formed formulas in PC
not become accidentally bound we first must introduce the notion of a is that their exclusion helped to simplify the rules of inference of PC.
term being free for some variable. We can say y is free for some variable, Their exclusion will not now help to simplify the rules of LPC, and
say x in A, if and only if no free occurrence of x in A falls within the since the only fault of these formulas is that they contain idle quantifiers,
scope of a quantifier containing y. In other words, y can be substituted they will be allowed.
for x in A without y coming under the scope of a quantifier in A. For The symbols "'-', V, A, -+, +-+, 'I, 3, and = will be used in a self-
example, x is free for x in (3), but y is not free for x. In this formula referring manner in the formation rules (as they have been used all
along) and in the axiom and theorem schemes. They ,:ill be us~d to
'v'x(Fx V Gxy) V Fz
refer to object-language symbols of the same type. That IS, they WIll be
z is free for y, x is not free for y, x is free for z, and y is free for y. metaconstants in a way analogous to A, B, and 0, which have been used
 230 Axioms for the Predicate Calculus with Identity
231

as metavariables for wffs. The object-language in this case is, of course DEFINITIONS:
the predicate language, and the metalanguage is what we use to t n' A -+ B = df (,.....,A V B)
about the object-language in the formation rules and what we use at~ A II. B = df
Af-tB = df (A-+B)
,.....,C.....,A V ,.....,B)
II. (B-+A)
refer to the object-language in axiom and theorem schemes. 3vA = df~\lv~A

Exercises AXIOM SCHEMES:

AvA-+A
I. Give. s~me more instances of (I), (4), and (5), being sure to observe an A-+BvA
restrIctIOns. Are all your instances of (I) valid? y AvB-+BvA
(A -+ B) -+ (C V A -+ C V B)
2. Let \IvA be \lx3yFxy. (i) In which example below is t free for v in A? Al \lv(A -+ B) -+ (A -+ \lvB) provided v is not free in A
(ii) Which is A(t/v), A(x/y), and A(x/x)? A2 \IvA -+ A (t/v) provided t is free for v in A
(a) 3yFxy
(b) 3yFyy
RULES OF INFERENCE:
(c) 3yFxx
Rule of detachment (MP): If I- A, I- A -+ B, then I- B.
(d) 3yFyx Rule of universal introduction (UI): If I- A, then I- \I vA.
3. Let \IvA be \lx\ly\lz(Fxy /\ Fyz -+ Fxz). Which of the following are true?
(a) x is free for x in A Exercises
(b) Y is free for x in A 1. Make the changes necessary to the primitive base so that 3 becomes a
(c) z is free for x in A primitive symbol and \I a defined symbol.
(d) w is free for x in A
2. How many theorems does universal introduction provide?
Answers 3. The number of axioms in LPC is infinite. How can the notion of an axiom be
2. (i), (a) and (c); (ii), (a) and (b), (c) and (d), and (a) and (c). effective?
4. What needs to be discarded so as to have only the sentential calculus?
3. (a) and (d) are true.
5. Since the number of axioms is infinite, could a new axiom be added?
6. Why isn't a rule of substitution found in LPC?
7. Which of the following are definitional transformations?
10.2 System of LPC'
(a) ~~\lx~Fx
~3xFx
PRIMITIVE SYMBOLS:
(b) 3x~Fx
1. Sentential and predicate variables: F, G, H, ... ~\lxFx
2. Individual variables: x, y, z, ... (c) ~\lx~Fx
3. Quantifier: \I ~,.....,3x~~Fx
4. Statement connectives: ,....., and V
(d) 3x~Fx
~\lx~,-...JFx
FORMATION RULES: (e) ~3x~~Fx
1. All atomic formulas are wffs. \Ix ~ Fx
2. If A is a wff, then ,.....,A is a wff. (f) r-;"\fx r--.J Fx
3. If A and Bare wffs, then (A V B) is a wff. 3xFx
4. If A is a wff which contains a free individual variable v then
\IvA is a wff. ' Answers/ __
I. Needless to say, one thing needed is a definition such as: \IvA = df ~3v"""" A.
1 The calculus described in this section is an adaptation of FI in Alonzo Church
Introduction to 1Ylathematical Logic, Princeton, N.J.: Princeton University Press: 2. Infinite number. UI can be used on each axiom provided by the axiom
1956. schemes.
 232 Axioms for the Predicate Calculus with Identity 233

3. A is an axiom iff it has one of six recognizable forms. liberty of putting (i) in a nonscheme form. From now on, in using
tautologies we will omit their introduction in the proof and omit the use
5. Yes, e.g., P. Any axiom that does not have one of the six recognizable forms.
ofMP.
7. Only (a), (d), and (f) are correct definitional transformations.
T3 A(tjv) --+ 3vA provided t is free for v in A
(i) (P --+ ",-,Q) --+ (Q --+ ",-,P)
1. 'iv",-,A--+",-,A(tjv) A2 provided t is free for v in A
10.3 Development of LPC 2. A(tjv) --+ ",-,'iv "'-/ A (i)
3. T3 df
Our unnumbered axioms plus MP constitute a complete sentential
calculus (axiom schemes making the use of a rule of substitution un- The proviso of A2 needs to be carried on to the statement of T3. Note
necessary). In the interest of brevity no theorem schemes ofthe sentential that substitution of a t for a v in the antecedent is involved in T3. It is a
calculus will be proved. If a theorem is needed from the sentential useful exercise to provide nonvalid wffs of LPC that would be instances
calculus, it will be entered without proof. Standard truth-table pro- of T3 if the restriction were removed.
cedures can verify for us if a formula is a tautology.
T4 'ivA --+ 3vA
TI ,-..;'iv "'-' A+-+ 3vA
1. ",-,'iv",-,A+-+",-,'iv",-,A Identity We will now provide ourselves with some useful derived rules.
2. ~'iv "'-' A+-+ 3vA df
DRI If I- A --+ E, then I- A --+ 'ivB provided v is not free in A
TI is a metatheorem. It is the statement: 'For any A, "'-''iv "'-/ A +-+ 1.A--+E A
3vA is a theorem'. The proof of TI is in the metalanguage. It is a 2. 'iv(A --+ E) DI
blueprint for any proofof any instance of T1. A proof of a theorem 3. 'iv(A --+ E) --+ (A --+ 'ivE) Al provided v is not free in A
scheme gives a-uniform and effective procedure for constructing a formal 4. A --+ 'ivE 3, 2, lVIP
proof in LPC for any instance of the theorem scheme. For the record
let us derive a theorem using the proof pattern above, thus obtaining ~ This proof indicates that whenever DRI is used it may be deleted and a
proof of a theorem in LPC which is an instance of Tl. formal proof constructed using DI, AI, and MP. DRI is a rule involving
the changing of the scope of a quantifier, and the Al proviso that deals
1. ",-,'ix",-, Fx +-+ ",-,'ix "'-' Fx Identity with scope change needs to be carried over. We note that no proviso
2. ",-,'ix",-, Fx +-+ 3xFx df
needs to be given concerning accidental binding since in the proof of
The formula on line 2 is a theorem of LPC. The proofs to follow will be DRI no axiom or rule having this proviso is used. Also note that A --+ B
in the metalanguage. In effect, in constructing proofs in the meta- in line 1 is assumed to be a theorem. We must have I-A --+ B to use DI.
language, A, B, and 0 in the axioms and rules take as instances well- As in PM, the use of previously proven theorems and derived rules
formed formulas that are metaformulas. What counts as a well-formed will indicate proof outlines and not formal proofs.
metaformula we can tell directly from the formation rules.
DR2 If I- A --+ E, then I- 'ivA --+ B
T2 'iv(A --+ B) --+ ('ivA --+ B)
(i) ((P --+ (Q --+ E)) /\ (S --+ Q)) --+ (P --+ (S --+ E)) T5 'iv(A --+ E) --+ ('ivA --+ 'ivE)
1. 'iv(A --+ B) --+ (A --+ B) A2 (t is free for v in A --+ B) 1. 'iv(A --+ E) --+ ('ivA --+ E) T2
2. 'ivA --+ A A2 (t is free for v in A) 2. 'iv(A --+ E) --+ 'iv('ivA --+ E) DRI (v is not free in
3. (1) /\ (2) --+ (4) (i) 'iv(A --+ E))
4. T2 1, 2, 3, lVIP (2) 3. 'iv('ivA --+ E) --+ ('ivA --+ 'ivE) Al (v is not free in 'ivA)
-4. T5 2,3, Trans
In 'ivA --+ A, A is always A (v/v). Thus in both uses of A2 above, t is free
for v in A since t is v and v is always free for v. We can easily see that DR3 If I- A --+ E, then I- 'ivA --+ 'ivE
lines 1 and 2 are metainstances of A2. Line 3 is obtained by employing l.A--+B A
the t~utology from the sentential calculus indicated at the beginning of 2. 'iv(A --+ E) DI
the proof and labelled (i). For the sake offamiliarity, we have taken the 3. 'ivA --+ 'ivB T5, lVIP
234 Axioms for the Predicate Calculus with Development of LPC

T6 \1'vA -+ \1'uA(ujv) provided u is free for v in A and u is not TIl \1'v(A -+ B) -+ (3vA -+ B) provided v is not free in B
free in \1'vA (i) ""p -+ ""Q+-7 Q -+ P
1. \1'vA -+ A (ujv) A2 provided u is free for v in A (ii) "" P -+ Q +-7 "-'Q -+ P
2. T6 DRI provided u is not free in \1'vA 1. \1' v(,,-,B -+ ,,-,A) -+ (""B -+ \1' v ,,-,A) Al provided v is not
free in B
2. \1'v(A -+ B) -+ (""B -+ Vv ,,-,A) (i), R
T7 \1'u\1'vA -+ \1'v\1'uA
1. 3. Vv(A -+ B) -+ (""Vv ""A -+ B) (ii), R
\1'vA -+ A A2 (t is free for v in A)
2. \1'u(\1'vA -+ A) VI 4. TIl df
3. \1'1t\1'vA -+ \1'uA T5, MP
4. \1'v(\1'u\1'vA -+ \1'uA) VI It should be noted that the replacement theorem of PM could be carried
5. (4) -+ (6) Al (v is not free in \1'u\1'vA) over into LPC, but then replacement could only be done with wfl's of
6. T7 5,4, MP PM. It would sanction line 3. We wish to have a replacement rule
that applies to quantified formulas. It is the replacement theorem (tbove
T8 ",,3vA -+ \1'v "" A that sanctions line 2.
(i) "" ""P +-7 P
1. "" ",,\1' v "-' A +-7 \1' v "" A (i)
TI2 Vv(A -+ B) -+ (3vA -+ 3vB)
2. T8 df TI3 Vv(A A B) +-7 VvA A B provided v is not free in B
TI4 3v(A A B) +-7 3vA A B provided v is not free in B
T9 \1'v"" A +-7 ,.....,,3vA
The statement of the metatheorems corresponding to the other major
theorems of the predicate calculus (Section 6.6) and their proofs will be
left for exercises. Note that none of these metatheorems carry provisos
The equivalence theorem for LPC can make use of all the notation or need to carry them, since such provisos would be idle and would not
of the equivalence theorem in PM with a few additions. Let AM be a do any work; if sentential variables do occur in theorems, then the
formula in which M occurs in A. Let AN be the result of replacing corresponding metatheorem may need the addition of a restriction. To
one or more occurrences of M in AM by occurrences of N. Let every illustrate, the scheme corresponding to T32 in Section 6.6 (Vx(P A Fx) +-7
variable that is free in M or N and bound in AM be in the list VI> ••• , v,., P A VxFx) is
then
Vv(A A B) +-7 A A VvB

THE EQUIVALENCE THEOREM:

Since A and B above can take open sentences as instances, v might be
\1'Vl'" \1'vn (M+-7N)-+ (AM +-7 AN)
free in A and the result be an invalid formula. To insure that the above
is a valid scheme, we need but add the restriction that A contain no free
The proof is by induction on the number n of symbols in AM, counting occurrences of the individual variable v.
each occurrence of "", v, or \1' as a symbol. The cases are the same as
those in Section 9.4, with the added case where AM is \1'vBM. T5 from Exercises
LPC is needed in this case. We leave the details as an exercise. 1. Prove the unproven derived rule and theorems through TI4, found in this
Let M, N, AM, and AN be as in the equivalence theorem, then section. Some of the proofs may be aided by first proving relatively simpler
theorems.
THE REPLACEMENT THEOREM (R): 2. Write down some actual theorems of LPC.
If f- M +-7 Nand f- AM, then f- AN -- ------
3. A proof of a theorem scheme gives a uniform and effective procedure for
constructing a formal proof of any instance of the theorem. Vsing some
Some uses of the replacement theorem are illustrated in the next scheme's proof, construct a proof of a theorem of LPC.
proof. 4. Prove schemes corresponding to the numbered theorems found in Chapter 6.
236 LPC with Identity 237
5. Give the proof by induction for the equivalence theorem. T3 V'xV'yV'z(x = y -+ (y = z -+ x = z))
6. Show that V'v(A A B)~ A A V'vB can have an invalid instance if v is free (i) (P --+ Q) -+ ((R -+ P) --+ (E -+ Q))
1. Y = x -+ (y = z -+ x = z) 12
in A.
2. x = Y --+ Y = X T2, A2 (t is free for v
in A)
3. (1) -+ ((2) -+ (4)) (i)
4. T3 1,2,3, MP (2), DI (3)
10.4 LPC with identity

T4 V'xV'y(Fx A x = Y -+ Fy)
We can obtain a first order calculus with identity by adding to LPC a (i) (P -+ (Q -+ E)) -+ (Q A P --+ R)
new two-place predicate constant Ixy that may be read, as before, "x is 1. x = y-+ (Fx-~ Fy) 12
identical with y". With the addition of this predicate, the formation 2. Fx A x = Y -+ Fy (i)
rules will be those in LPC and 3. T4 DI (2)

I followed by two individual variables is a wff. T5 V'x(Fx~ 3y(x = Y A Fy))

T6 V'x(Fx~ V'y(x = y -+ Fy))
T7 V'x3y(x = y)
To obtain significant theorems employing I we need two additional T8 V'xV'yV'z(y = x A z = x --+ y = z)
axiom schemes:

II Ivv
Some useful definitions may be introduced, and theorems employing
12 Ivu -+ (A -+ A (u/v)) provided u is free for v in A these new terms may be derived. As we saw in Chapter 7

Here I now acts as a metaconstant. The first axiom, II, established the There is exactly one x such that Fx.
reflexivity of I, and the second axiom scheme established the sub-
stitutivity of equivalents. The more familiar notations = and 0/= may can be translated as
be introduced by definition as follows:
(1) 3xFx A V'xV'y(Fx A Fy -+ x = y).
v = u =dfIvu
v 0/= u =df ,....,Ivu
(1) may be abbreviated as
The resulting system is a lower predicate calculus with identity, and it
will be denoted by LPGI. (2) 3!xFx
We can prove the schemes corresponding to theorems but the
provisos would be tedious to state. So we prove theorems rather than (2) is called the numerically definite quantifier.
metatheorems. If (2) =df (1) is introduced into LPC with identity, then this allows
us to prove a theorem such as the next theorem.
T1 V'x(x = x)
1.x=x II
2. T1 DI T9 V'x3!y(y = x)
1.x=x II
T2 V'xV'y(x = y -+ y = x) 2. 3y(y = x) T3 (t is free for v in A)
(i) (P -+ (Q -+ E)) -+ (Q -+ (P -+ E)) 3. V'xV'yV'z(y = x A z = x -+ y = z) T8
1. x = y -+ (x = x -+ y = x) 12 4. V'yV'z(y = x A z = x -+ y = z) A2 (t is free for v in A)
2. x = x II 5. (2) A (4) 2,4, Conj
3. (1) -+ ((2) -+ (4)) (i) 6. 3!y(y = x) df
4. T2 1, 2, 3, MP (2), DI (2) 7. T9 DI
238 Axioms for the Predicate Calculus with Identity 10.6 Soundness of LPC with identity 239
Exercises Since A is its own transform and not a tautology, LPO is absolutely
1. Give some instances of II and 12. consistent. Oonsistency in Post's sense is assured by the same fact.
2. Prove the unproven theorems in this section and the numbered theorems Exercises
found in 7.3.
1. Show that P is not a theorem of LPCI and thus prove that LPCI is negation
3. Prove consistent.
(a) Vx(Fx~ Ox) --+ (3!xFx~ 3!xOx)
(b) 3!xFx~ 3xVy(x = y~ Fy) 2. Use induction to prove every theorem of LPCI has a tautological trans-
(c) 3!xFx~ 3x(Fx A Vy(Fy --+ x = y)) form.
3. Do the same (as in 2) for PC.
4. Let T be an axiom system that has only contingent sentential formulas as
10.5 Consistency of LPC with identity theorems. (a) Is this theory negation consistent~ (b) Is it absolutely
consistent ~
To prove the predicate calculus with identity consistent, we do some-
Answers
thing analogous to transforming formulas into associated sentential
formulas (see Sections 6.7 and 8.4). 1. LPCI is negation consistent iff it is absolutely consistent.
In this case, each formula will be transformed as follows: First, every 4. (a) No, since P and ,.....,p are theorems; (b) Yes, since P V ,.....,p is not a
occurrence of Vv for every variable v will be deleted. Second, each theorem.
occurrence of every individual variable will be replaced by x. Third,
each occurrence of x = x will be replaced by an occurrence of x =
x --+ x = x. A and B are regarded as transforms ofthemselves. Following 10.6 Soundness of LPC with identity
this procedure, the transforms of the axioms of LPOI are:
To show LPOI is sound is to show that every theorem is valid. To
AvA--+A speak of a wff in LPOI as valid is to say that it is true under any inter-
A--+BvA
pretation. As described in earlier sections, one obtains an interpretation
AvB--+BvA
(A --+ B) --+ (0 V A --+0 V B) of a formula in LPOI when he:
(A --+ B) --+ (A --+ B)
A--+A 1. Indicates a nonempty domain
x=x--+x=x 2. Assigns to each sentential variable the value Tor F
(x = X --+ X = x) --+ (Ax --+ Ax) 3. Assigns to each free individual variable a name for an element in
the domain
The transform of every axiom is a tautology. 4. Assigns to each n-place predicate a predicate in the domain
5. Assigns to = the identity predicate in the domain.
Modus ponens applied to two tautologies always gives a tautology.
If UI is applied to a formula whose transform is a tautology, then the To illustrate, consider
transform of the resulting line is a tautology.
Since a proof consists of axioms or lines derived by MP or UI, all (1) Fx A Vy(Fy --+ P) A x = Y
the lines of a proof have tautologies as transforms, and thus do all the
metatheorems. and let the domain be show biz people, Fx: x is fat, P be T, x be Oliver
Finally, if a metatheorem has a tautological transform, then all its Hardy, y be the partner of Stan Laurel. In this interpretation (1) is true.
instances have tautological transforms since they have the same form; Af~rmula in LPOI is valid if and only if it is true on every inter-

thus all theorems of LPO with identity have tautological transforms. pretation. We again note that if an interpretation of a formula A
Since for any formula A, A A ,.....,A cannot have a tautological transform, yields a true proposition, this interpretation is called a model for A.
we obtain the result that LPO with identity is negation consistent. Thus if a formula in LPOI has models for any interpretation, then it is a
valid formula.
240 Axioms for the Predicate Calculus with Identity 10.6 Soundness of LPC with identity 241

We now begin a somewhat intuitive commentary concerning the then 'ixFx is always true; thus the consequent of 'ix(Fy -)- Fx)-+
proof that every theorem of LPOI is valid. (Fy -)- 'ixFx) will always be Fy -)- T, which is true, and thus the axiom
Truth-tables confirm that the first four axiom schemes only have is true.
valid formulas as instances. OASE 2: If Fx has only false instances in any domain of discourse,
To prove that the axioms from Al and A2 of LPOI are valid, We then 'ixFx only has false instances. If'ixFx has only false instances, then
suppose that what is true of one axiom from the scheme is true of all its Fy can only be false. If Fy is false in 'ix(Fy -)- Fx) -)- (Fy -)- \;fxFx) ,
axioms since form is preserved. We will take as our instance of axiom both the antecedent and the consequent are true; thus the axiom is true.
scheme one OASE 3: If Fx has only true instances in a domain, case 1 applies,
and if it has only false instances, case 2 applies.
'ix(Fy -)- Fx) -)- (Fy -)- 'ixFx) OASE 4: If Fx is sometimes true in a domain, then in the case in
which Fy is true when a name is assigned, the antecedent will be false.
and as our instance of axiom scheme two In the case in which Fy is false when a name is assigned, the ante-
cedent is true; but then the consequent Fy -)- 'ixFx is also true. So
'ixFx -)- Fx the axiom is true.
We leave the account of the validity of the identity axioms as an
Both of these formulas are made up of one and only one predicate, exercise.
the one-place predicate Fx. There are no other predicat~s and there are Rl: Modus ponens is a valid rule.
no sentential variables. Interpreting these two formulas is thus re- R2: If 'If ~ A, then I- 'ivA' is a valid rule, then when it is applied
stricted to (1), (3), and (4) above. to a valid line, it only permits the inference of a valid line. If Fx is valid,
A one-place predicate such as Fx can on interpretation (l) give only then it immediately follows that 'ixFx is valid.
true statements for any domain, (2) give only false statements for any Since the axioms of LPOI are all valid, and since Rl and R2 only
domain, (3) give only true statements in some domains and only false allow the derivation of valid lines from valid lines, every theorem of
statements in other domains, or (4) give some true and some false LPOI is valid.
statements in a domain. If in each of these cases the above axioms are LPOI can again be proven negation consistent. For no A can both A
true, then the axioms are valid. and ,.....,.,A be a theorem, for A and ,.....,.,A cannot both be valid.
Let us then consider each axiom and these four cases, beginning with
Exercises
the second axiom.
1. Provide a model for each of the following:
OASE 1: The predicate represented by Fx has only true instances
(a) 'ix(Fx+--t 3y(y = x /\ Fy))
in any domain of discourse. If this is so, 'ixFx -)- Fx must be true since (b) 'ix'iy3zFxyz
Fx can never have a sentence expressing a false statement as an instance. (c) 3y'ix(Fx+--t x = y)
OASE 2: The predicate represented by Fx has only false instances in (d) 3xFx /\ (Fa -)- 3x3y(x *- y /\ (Fx /\ Fy)))
any V. If this is so, then 'ixFx has only sentences expressing false 2. Demonstrate the validity or invalidity of each of the following:
statements as instances; thus 'ixFx -)- Fx is always true. (a) 'iv(A V A) -)- 'ivA V 'ivA
OASE 3: If Fx has only true instances in a domain, the considerations (b) 'ivA -)- A
(c) A -)- 'ivA
in case 1 apply, and if it has only false instances in a domain, case 2
(d) 3v'iuA -)- 'iv3uA
applies. (e) A -)- 3vA
OASE 4: Fx has in a domain some instances that are true and some (f) ('ivA -)- 'ivB) -)- 'iv(A -)- B)
that are false. In this case 'ixFx can have only sentences expressing 3. Prove that the identity axioms are valid.
false statements as instances, in which case 'ixFx -)- Fx is true.
Turning next to the first axiom and the four cases, we again consider
4. WhYIs DI of LPC valid in LPC but invalid in PC?

each. Answers
O~SE 1: If Fx has only true instances in any domain of discourse, 1. (b) V: positive integers, Fxyz: x +y= z.
242 Axioms for the Predicate Calculus with Identity 10.7 Completeness of Predicate Logic Theories 243

10.7 Completeness of predicate logic theories From Theorem 1 and

The sentential calculus is complete in both the relative sense (all tauto_ Theorem 2: If LPC has a model, then LPC is consistent.
logical formulas are derivable) and in the absolute sense (a non-
derivable wff conjoined to the axioms results in an inconsistent axiom Theorem 3 below follows:
set). The predicate calculus, however, is not absolutely complete but is
relatively complete. The first is easily shown. There are nonderivable Theorem 3: LPC is consistent if and only if it has a model.
formulas that will not create inconsistent axioms if added to the axioms Theorem 4: Let A be a well-formed predicate formula of LPC.
of the predicate calculus LPC. Examples of such formulas are: Let LPC' be the result of conjoining A to LPC. If
LPC' is inconsistent, then "",A is a theorem of LPC.
3xFx -+ 'ixFx, 3xFx A 3xGx -+ 3x(Fx A Gx), Fx-+ 'ixFx
Proof: If LPC' is inconsistent, then there is a proof in LPC that
Their transforms can be seen to be tautologies. Consequently, the ~ A -+ "",A. Since ~ (A -+ "",A) -+ "",A, "",A is a theorem.
addition of these to the axioms of LPCI would thus not alter the con-
sistency of LPCI. The above predicate formulas are not valid. Since Theorem 5: Let B be a predicate formula of LPC. If B is true in
every thesis of LPCI is valid, these formulas are not derivable. Hence every model for LPC, then B is a theorem of LPC.
LPCI is not absolutely complete.
The proof of the relative completeness of the predicate calculus is Proof: Suppose that B is true in every model for LPC. Let LPC' be
complicated and requires quite a few preliminaries. We will not the result of conjoining "",B to LPC. If M is a model of LPC', then both
attempt to give the proof in this book. Many feel that the demarcation Band ,--.,B are true in M, which is impossible by Theorem 3. Therefore
point between introductory logic and advanced logic is the completeness LPC' has no model. Then LPC' is inconsistent by Theorem 3 (and 1).
proof for the predicate calculus. Then"", "",B is a theorem of LPC by Theorem 4. Hence B is a theorem
The completeness proof for the predicate calculus was first given by ofLPC by DN.
Kurt Godel in the early thirties and makes use of a normal form for well-
Theorem 6: Every valid predicate formula is a theorem of LPC.
formed formulas in predicate logic. The point is to show that machinery
is available in the predicate calculus to grind out all the valid formulas in
Proof: Suppose A is valid. Then A is true in every model for LPC.
the normal form.
Hence, by Theorem 5, A is a theorem.
The completeness proof in wide use today is Leon Henkin's 1949
There is an effective decision procedure to determine validity for
proof. The important theorem of this proof is the theorem that every
sentential formulas and for certain classes of predicate formulas (see
consistent first order theory, of which the predicate calculus is one, has a
Section 5.6).3 There is, however, no effective decision procedure that
countable modeJ.2 We have by this time some experience in providing
suffices to determine for an arbitrarily selected wff of LPC whether or not
models for formulas-interpretations that give true statements. There
it is valid and thus a theorem. The American logician Alonzo Church
can also be a model for sets of formulas and a model for the set of
proved that such a procedure could not be obtained for the first-order
formulas that make up the predicate calculus. In general M is a model
predicate calculus. His proof was based on a 1931 paper by Kurt Godel
for a theory T iff M is a model for the set of theorems of T. A model is
that showed that no matter what consistent axiom system we select for
countable when its elements can be enumerated. Let us set out, then,
the arithmetic of positive integers, we can formulate an acceptable
the crucial theorem:
statement of arithmetic that is not provable in the system-in other
Theorem 1: Every consistent first order theory, LPC being a first words, axiom systems for number theory cannot be relatively complete.
order theory, has a countable model. However, even though there is no effective decision procedure for the
va1i.sJ:ity of the formulas in LPC, there are mechanical procedures for
2 This outline is adopted from Section 3.27 of Angelo Margaris, Fil'St Order

1I1athematical Logic (Waltham, Mass.: Blaisdell Publishing Co., 1967). 3 See Church, Introduction, Section 46.
244 Axioms for the Predicate Calculus with Identity

constructing proofs for any valid predicate formula (see, for example,
Quine's Methods and Jeffrey's Formal Logic).
Earlier the question was raised whether the natural deduction
system for predicate logic set down in Chapter 6 is complete-whether all
valid wffs of PC are theorems. Here again we can prove this in two ways.
The following list contains a few of
First, prove that all the theorems of LPC are theorems of PC by showing
that all instances of the axiom schemes of LPC can be proven by using
further the books that will prove useful to
those who wish to pursue some
PC and that the rules of LPC are to be found in PC. Since LPC is com-
plete, PC must be complete. Second, produce a proof on lines analogous
readings aspects of logic set forth in this
text. Books of the introductory
to the completeness proof for LPC. type that provide lucid alternative
treatment of the topics in this book
Exercises are Copi and Mates. Kalish and
1. Prove from the theorems given that if LPC has a model, then it has a Montague, Lemmon, and Suppes
countable model. give alternative, up to date, intro-
ductory treatment of natural de-
2. Prove theorem 2.
duction systems and include topics
3. Prove that if any theory has a model, then it is consistent if exactly one of not considered in this book. Hughes
the pair (P, ,--,P) can be true in M. and Londey provides an intro-
4. If A is true in every model for the predicate calculus, then is A valid 1 ductory treatment of finite logistic
systems, including syllogistics.
5. A system is negation complete if and only if for every wff A either A or ,--,A is Jeffrey gives a detailed introductory
a theorem. Prove that PM and LPC are not negation complete.
treatment of the tree method and
6. A statement is a logical truth (or is analytic) if and only if the way that '--', even has a chapter on Godel's
V, 'if, and = enter into its construction makes it true regardless of how its incompleteness theorem. Quine's
other terms are understood. (a) Can the notion of a logical truth be an Methods also provides an intro-
effective notion 1 (b) Is it possible to tell in all cases whether a statement is a ductory account for natural de-
logical truth 1 ductions, along with truth-table
7. Why would we not want the predicate calculus to be absolutely complete1 methods, normal forms and dis-
cussion of some philosophical issues.
Advanced books that provide
amplification of some of the topics
treated here are Church, Kleene,
Margaris, Mendelson, and Quine's
Mathematical Logic. Church pro-
vides the most comprehensive treat-
ment of sentential and predicate
axiom systems. Kleene, Margaris,
and Mendelson cover many topics
outside the range of this book, for
example, the formal development of
number theory.
Hilbert and Ackermann is a
classical compact account ofaxio-
matic logical systems. Keene's book

245
246 Further Readings 247

IS a condensed exposition ofaxio_ Quine, W. V. 0., Mathematical Logic. Cambridge, Mass.: Harvard University
matic predicate calculus. Kneale Press, 1951.
and Kneale covers the history of Quine, W. V. 0., Methods of Logic. New York: Holt, Rinehart & Winston, Inc.,
logic, while-Bukasiewicz's book pro- 1950.
vides the pioneer treatment of Robison, Gerson B., An Introduction to Mathematical Logic. Englewood Cliffs,
axiomatic syllogistics. Beth's book N.J.: Prentice-Hall, Inc., 1969.
is an advanced and condensed Suppes, P., Introduction to Logic. Princeton, N.J.: D. Van Nostrand Co., Inc.,
Bxposition of the topics in this book 1957.
including the tree method. It als~
covers topics not found in this book
for example, machines which prov~
theorems in logic. Robison is an
easily understood introductory book
that concentrates on the application
of logic to mathematical theories.

Beth, E. W., Formal Method. New York: Gordon & Breach Science Publishers
Inc., 1962. '
Church, A., Introduction to Mathematical Logic. Princeton, N.J.: Princeton
University Press, 1956.
Copi, 1. M., Symbolic Logic. New York: The Macmillan Company, 1967.
Hilbert, David, and W. Ackermann, Principles of Mathematical Logic. New
York: Chelsea Publishing Co., 1950.
Hughes, G. E., and D. G. Londey, The Elements of Formal Logic. London:
Methuen & Co. Ltd., 1965.
Jeffrey, R., Formal Logic: Its Scope and Limits. New York: McGraw-Hill Book
Company, 1967.
Kalish, Donald, and Richard Montague, Logic: Techniques of Formal Reasoning.
New York: Harcourt, Brace & World, Inc., 1964.
Keene, Geoffrey, First-Order Functional Calculus. New York: Dover Publi-
cations, Inc., 1964.
Kleene, S. C., Mathematical Logic. New York: John Wiley & Sons, Inc., 1967.
Kneale, W., and M. Kneale, The Development of Logic. Oxford: The Clarendon
Press, 1962.
Lemmon, E. J., Beginning Logic. London: Nelson, 1965 .
.l:;ukasiewicz, J., Aristotle's Syllogistic from the Standpoint of Modern Formal
Logic. Oxford: The Clarendon Press, 1957.
Margaris, A., First Order Mathematical Logic. Waltham, Mass.: Blaisdell
Publishing Co., 1967. \
Mates, B., Elementary Logic. New York: Oxford University Press, Inc., 1965.
Mendelson, E., Introduction to Mathematical Logic. Princeton, N.J.: D. Van
Nostrand Co., Inc., 1964.
 Index 249
Contradiction, 80 Elimination, universal (UE), 137
Contradiction law, 95 Empty term, 114
Contraposition (Con), 48
I Corresponding conditional, 7
Equivalence (Equiv), 48
logical, 47
Counterexample, 75 Excluded middle law, 90
abstract and concrete, 117 Exclusive use of 'or', 25
index CS
associated sentential formula for,
Existential elimination (EE), 140
Existential quantifiers, 107
196 Export-import law (EI), 95
completeness of, 196-197
consistency of, 195-196
soundness of, 196 Formal language, 13
Absolute completeness, 197, 221 Church, Alonzo, 157, 230n, 243 syllogistics system, 190-193 Formal system, 188
Absolute consistency, 99, 214 Circuits, simplifying, 61 Formation rules, 32
Absorption (Abs), 48, 95 Commutation (Com), 48 with individual constants, 143
Accidental binding, 228 Completeness Decision procedure, 35 for LPC, 230
Addition rule (Add), 74 absolute, 197,221 normal form, 57 for PC, 150-151
Analytic statements, 12, 131, 168 relative, 197, 220-221 predicate formula, 116-117 for predicate language, 130
Antecedent, 7 Compound statement, 20 truth table method, 36 for sentential language, 32
sentential, 44 Compound statement former, 20 Deduction, see system Formulas, 9
valid, 5 Conclusion, 3 Deduction theorem, 214 atomic, 130
valid patterns of, 41 Conditional formula, 20 Definite descriptions, 175-178 basic components of, 135
Argument pattern, 12 Conditional proof rule (RCP), 78 Definitions, 19 bi-conditional, 20
Aristotelian logic, 107, 190-193 Conditional sentence, 7 De Morgan, A_, 48 conditional, 20
Associated sentential formula (asf) Conjunct, 16 theorem (DM), 48 contingent, 38
for CS, 196 Conjunction, 16 Derived rules, 91, 210 inconsistent, 38
for LPCI, 238 rule, 73 Dilemma, constructive, 95 interpreting, 113, 134, 166, 239
for PC, 155 Conjunctive normal form, 57 Dilemma, destructive, 95 in logic, 131, 168
Association (Assoc), 48 Connectives Discharged assumption, 77 model for a predicate, 115
Assumption main, 37 Discourse, universe of (V), 105 sentential, 10, 50
discharged, 77 ranking, 33 non-empty, 114 valid, 10, 38, 113
rule (A), 71 scope of, 23 Disjunction, 19 well-formed (wff) , 32
scope of, 78 truth-functional, 21 rule (Dis), 79 Free individual variable, 101
Atomic formula, 130 Consequence, tautological, 68 Disjunctive normal form, 59
Atomic statement, 20 Consequent, 7 Disjunctive syllogism (DS), 91
Axiomatic method, 186 Oonsequentia mimbilis, 94 Disjuncts, 19 General name, 193
Axiomatic theory, 188 Consistency Distribution (Dist), 48 Godel, Kurt, 197,242,243
Axiom scheme, 227 absolute, 99, 214 Domain of discourse (V), 105 Group theory, 179
Axiom systems, 205 of CS, 97 Double negation (DN), 48, 73
of LPCI, 238 Dropping inconsistencies (DI), 52
Basic components of a formula, 135 negation, 97, 214 Dropping tautologies (DT), 52 Henkin, Leon, 242
Basic truth-table, 17 o~PM, 216-217 Duns Scotus' law, 92
Biconditional formula, 20 Post's sense of, 99, 214
Bound individual variable, 110 of SC, 97
Idempotent (Idem), 48
Constants, individual, 134
Effective notion Identity, 160-161
Calculus, predicate, 6, 180 Constants, sentential, 23
for a proof, 188 axioms for, in LPC, 236
see also predicate calculus Contingent
for a tautology, 40 elimination (IE), 167
Calculus, sentential, 67 formula, 38
for a theorem, 188 introduction (II), 167
Categorical propositions, 107 statement, 6
for a well-formed formula (wff), 32 rules for, in PC, 167

\ 248
250 Index Index 251
Implication (Imp), 48 interpreting a formula in, 239 Polyadic predicates, 120 Rules of inference, 69, 70
transitivity of, 90 predicate calculus system, 236 Post's sense of consistency, 99, 214 valid, 70
Inclusive use of 'or', 25 soundness of, 239-241 Predicate, monadic, 120 Russell, Bertrand, 179, 203
Inconsistencies, dropping (DI), 52 ..B"ukasiewicz, Jan, 193 Predicate calculus
Independence system LPC, 230-231
of CS, 194 Satisfying a predicate, III
system LPC!, 236
of PM, 217-219 Main connective, 37 SC
system PC, 150-151
of SC rules, 95 Mathematical induction, 212 completeness of, 97, 224
system PC!, 167
Individual constants, 134 Mention, see use/mention distinction consistency of, 97
Predicate formula, 116-117
Individual variables, 102 Metalogical questions, 97 replacement rule in, 94
model for a, 115
Induction, 212 Metalogical variables, 32 rules, independence of, 95
Predicate language, see formation
Inference, rules of, 69-70 Model for a predicate formula, 1I5 soundness of, 97
rules
Interpretation Modus ponens (MP), 69 system, 67, 86-88
Predicate logic, 58
of identity sign, 161 Modus tollens (MT), 74 turnstile sign in, 89
with identity, 180
principal, 192-193, 218 Monadic predicate, 120 Scope
Predicate term, 104
of quantifiers, III of a connective, 23
n-place, 104
of statement connective, 16-20 of a quantifier, 108
Predicate variables, 128
Interpreting a formula, 113 Name, general, 193 Semantics of a theory, 188
Premise, 3
with identity, 166 Natural deduction systems, 67, 189 Sentence, singular, 110
Principal interpretation, 192-193, 218
with individual constants, 134 Necessary condition, 28 Sentential argument, 44
Principia Mathematica, 203
in LPC!, 239 Necessary statements, 6 Sentential calculus, 67
Proof
Introduction, existential (EI), 138 Negation, 17 Sentential constant, 23
by cases, 95
Introduction, universal (UI), 142 consistency, 97, 214 Sentential formulas, 10
by contradiction, 95
Intuitional logic, 225 double (DN), 48, 73 transformation of, 50
effective notion for, 188
Normal forms, 57 Sentential language, 16
see also system
conjunctive, 57 see also formation rules
Propositions, categorical, 107
Kleene, S. C., 28n disjunctive, 59 Sentential logic, 99
Numerically definite quantifier, 237 Sheffer stroke, 55
Simplification (S), 69
Language Quantification, vacuous, 129 Simplifying circuits, 61
formal, 13 Object language, 32, 229-230 Quantifiers, 106 Singular sentence, 110
meta, 32, 229-230 '01", inclusive use of, 25 existential, 107 Singular term, 103
object, 32, 229-230 exclusive use of, 25 interpretation of, 110 Soundness
predicate, 130 Paraphrase, principle for correct, 42 numerically definite, 237 of OS, 196
Leibniz, G. W., 6 PC scope of, 108 of LPC!, 239-241
Limited term, 140, 145 completeness of, 155-157, 244 universal, 106 of PO, 155-157
Logic, see intuitional, predicate, sen- predicate calculus system, 150-151 of PM, 215-217
tentiallogic replacement rule in, 153 of SC, 97
Logical equivalence, 47 rules for identity in, 167 Ranking connectives, 33 Statement variables, 31
Logically imply, 11, 131, 168 soundness of, 155-157 Reductio ad absurdum Substitution, rule of, 49
Lower/higher predicate calculus, 180 system, 134, 150-151 rule (RAA), 80 in PM, 206
LPC PC!, 167 truth-table test, 45 Sufficient condition, 20
axioms for identity in, 236 PM Relations, properties of, 162-165 Syllogism, disjunctive (DS), 91
axiom system, 230-231 axiom system, 206 Relative completeness, 196-197, 220- Syllogistics, system OS, 190-193
completeness of, 242-243 completeness of, 220-223 221 Syntax of a theory, 188
predicate calculus system, 230-231 consistency of, 216-217 Replacement rule, 49 System
replacement rule in, 234 independence of, 217-219 in LPC, 234 axiom system LPO, 230-231
LPCI replacement rule in, 213 in PC, 153 axiom system LPCI, 236
axiom system, 236 rule of substitution in, 206 in PM, 213 axiom system PM, 206
consistency of, 238 soundness of, 215-217 in SC, 94 OS, 191
"",.,
252 Index
System-(cont.) Truth-table, 17
PO, 134, 150-151 basic, 17
POI,167 method fo~ one-place predicate
SO, 67, 86-88 formulas, 119
test for reductio ad absurdum, 45
Truth-value, 6
Tautological consequence, 68 Turnstile sign, 70
Tautology, 39, 135 in axiom systems, 205
dropping (DT), 52 in SO, 89
effective notation for, 40
Term
empty, 114 Universal elimination (UE), 137
limited, 140, 145 Universal introduction (UI), 142
Theorem Universal quantifiers, 106
of OS, 191 Universe of discourse (V), 105
deduction, 214 non empty, 114
effective notion for, 188 Use/mention distinction, 14
of LPO, 229
of PO, 151
of PM, 205 Vacuous quantification, 129
of SO, 88 Valid argument, 5
Theory, 187 Valid argument patterns, 41
Transformation of sentential for- Valid formula, 10, 38, 113
mulas, 50 Variable
Transitivity of implication (Trans), free, 101
90 individual, 110
Tree (or tableau) method metalogical, 32
for predicate logic, 158 statement, 31
for predicate logic with identity,
180
for sentential logic, 99 Well-formed formula (wff), 32
Truth-functional connectives, 21 see also formation rules

-~

~ ..