INTRODUCTORY MODAL LOGIC

Introductory
Modal Logic
KENNETH KONYNDYK

University of Notre Drone Press

Notre Drone, Indiana 46556
 University of Notre Dame Press
Notre Dame, Indiana, 46556
www.undpress.nd.edu
All Rights Reserved

Published in the United States of America

Copyright © 1986 by University of Notre Dame

Reprinted in 1993, 1996, 2002, 2008

Library of Congress Cataloging-in-Publication Data

Konyndyk, Kenneth.
Introductory modal logic.

Bibliography: p.
1. Modality (Logic) I. Title.
BC199.M6K66 1986 160 85-41007
ISBN 13: 978-0-268-01159-8 (pbk.)
ISBN 13: 978-0-268-02407-9 (hardback)

∞ This book is printed on acid-free paper.

 CONTENTS

Preface ix
Introduction 1
1 Propositional Logic 3
1.1 Symbolism 4
1.2 Rule of Replacement and Equivalences 5
1.3 Rules of Inference 6
1.4 Proofs 9
1.5 Exercises 9
2 Propositional Modal Logic 11
2.1 The Modal Concepts 12
2.1.a An Explanation of Necessity 12
"Ungiueupable" 14
"Unable to be rationally rejected" 15
"Self-evident" 15
A priori 15
2.1.b The Possible Worlds Picture 16
2.1.c A Map of the Modal Concepts 17
2.1.d Symbols and Definitions 18
2.1.e Exercises 19
2.1.f Symbolizing Conditionals 20
2.1.g The Necessity of the Consequence and the
Necessity of the Consequent 21
2.1.h Exercises 23
2.2 The Logic of Modality 23
2.2.a The Need for a Logic of Modality 23
2.2.b Logical Relationships 25
2.2.c Systems of Propositional Modal Logic 30
2.3 The System T 31
Exercises 32

V
vi CONTENTS

2.3.a Rules 32
The necessity elimination rule 33
Necessity introduction subproofs 34
The necessity introduction subproof rule 34
The necessity introduction rule 35
The T-reiteration rule 36
2.3.b Hints 36
2.3.c Exercises 38
2.3.d More Rules 39
Possibility introduction 39
Possibility elimination 39
Modal modus ponens 40
Exercises 40
2.3.e Decision Procedure 41
2.3.f Counterexamples 41
2.3.g Exercises 44
2.3.h Iterated Modalities and Reduction of
Modalities 46
2.4 The System S4 49
2.4.a Rules 49
2.4.b Exercises 50
2.4.c Reduction of Modalities in S4 50
2.5 The System S5 51
2.5.a Rules for S5 52
2.5.b Exercises 53
2.5.c Reduction Laws 53
2.5.d Exercises 55
2.6 Philosophical Matters 55
2.6.a Differences between Systems 55
2.6.b Some Assumptions Implicit in Our
Presentation 60
2.6.c Is There Only One Correct System of
Modal Logic? 64
3 Quantification 69
3.1 Symbols 69
3.2 Interpretation 70
3.3 Equivalences 71
3.4 Rules 71
Universal Instantiation 72
Universal Generalization 72
Existential Generalization 72
 CONTENTS vii

Existential Instantiation 72
Exercises 75
4 Quantified Modal Logic 77
4.1 Motivation 77
4.2 'l\vo distinctions 78
4.3 Symbolization 84
4.4 Some Objections to Quantified Modal Logic 85
4.5 Essential Properties and Possible Worlds 88
4.6 Considerations Bearing on Quantified Modal Logic 90
4. 7 The Systems 92
4.8 A Kripke System 92
Exercises 94
4.9 Objections to This System 95
4.10 Kripke's Revised S5 97
The necessity introduction rule 98
Possibility elimination 99
4.11 Reflections on Kripke's Revision 101
4.12 The Actualistic Version of Possible Worlds 103
4.13 Actualism and Quantification 105
Exercises 108
4.14 Actualistic Quantified Modal Logic 108
4.14.a Rules 109
4.14.b Exercises 112
4.15 Existence and Modality 113
4.16 The Problem of Transworld Identity 115
4.1 7 Superiority of Actualism 117
Appendix I 119
Appendix II 121
Appendix III 125
Notes 127
Bibliography 131
 PREFACE

This text is designed as an introduction to modal logic for a stu-

dent who has studied elementary symbolic logic. It presents the sym-
bolism, natural deduction proof procedures, and problems of modal
logic, using the systems which have come to be known as T (or M),
S4, and S5.
Since elementary symbolic logic texts differ with respect to sym-
bolism, proof techniques, and rules, this book begins with a brief
specification of symbols, rules, and proof technique to be used here.
Propositional modal logic and quantified modal logic are devel-
oped separately. The chapter treating propositional modal logic can
be studied immediately after ordinary propositional logic. Although
it introduces some metatheoretical considerations, it does not pre-
suppose any prior acquaintance with metatheory. I consider some
of the philosophical motivation for modal logic and do not try to pre-
tend futilely that modal logic is or should be philosophically neutral,
preferring instead to point out areas of philosophical controversy over
modal logic.
Quantified modal logic raises even more philosophical questions.
Which system is preferable depends quite directly on how certain
metaphysical questions are answered. Indeed, the point of quantified
modal logic is to represent the logical structure of certain correct
metaphysical views. Accordingly, I develop the chapter on quanti-
fied modal logic so as to make clear where metaphysical choices have
to be made, and I indicate what I think are the correct choices. There
are plenty of references to discussions of the philosophical issues at
stake. But I develop this chapter so a student can get a good grasp
of quantified modal logic without being overwhelmed by the philo-
sophical difficulties.
My presentation owes a great deal to the work of Saul Kripke and
to the interpretations of Alvin Plantinga and Arthur N. Prior, but
I am most deeply indebted to my colleagues Alvin Plantinga and

ix
X PREFACE

Thomas Jager. I am grateful to my colleagues in the Calvin College

Philosophy Department for discussing this book in our weekly col-
loquium. They have suggested innumerable corrections and improve-
ments. I also owe a great deal to my logic students who have served
with varying degrees of willingness as the subjects of my experiments
with this text.
I wish to express special thanks to Donna Kruithof, who exhibited
patience and fortitude in typing and retyping the manuscript.
I began preliminary work on this project while I held an NEH
Fellowship in Residence for College Tuachers in 1975-1976, and I con-
tinued working on it during a sabbatical from Calvin College in the
fall of 1977. I gratefully acknowledge the support of the NEH and
of my home institution.
 INTRODUCTION

Modal logic studies inferences that involve modalities of proposi-

tions. Here we will confine our attention to alethic modalities.
Propositions are true or false. However, we sometimes have occa-
sion to note that a proposition is not merely true but that it is nec-
essarily true, while other propositions, though false, are possibly true.
So necessity and possibility are regarded as modes or modifications
of truth (aletheia), and have come to be called alethic modalities.
Aristotle knew, and perhaps discovered, that these notions of pos-
sibility and necessity can each be expressed in terms of the other.
He worked out some of this in De Interpretatione. Aristotle also did
some of the first recorded work on modal inferences within the basic
framework of the syllogism.
Like ordinary propositional logic, modal logic employs a notion
of implication or consequence. In modal logic the notion of implica-
tion is strengthened by adding necessity to produce a concept labelled
necessary (or "strict") implication, necessary consequence, or some-
times, entailment. This concept furnishes a more intuitive and accu-
rate representation of many "if ... , then .. .'' sentences in ordinary
usage, especially philosophical and theological usage, than classical
propositional logic provides. One of the interesting things for us to
note as we develop modal logic is which tautologies or theorems true
of the classical implication relation (' ::>') have or lack counterparts
for necessary implication.
Modal logic is usually developed as an extension of classical propo-
sitional logic and first-order quantification theory. That is the way
it will be done here. Since most of the textbooks for learning first-
order logic differ slightly from each other in symbolism, choice of
rules, and formulation of rules and proof techniques, we will begin
our treatment of modal logic by briefly specifying the formulation
of propositional logic we will use. Once this is done, we turn to propo-

1
2 INTRODUCTION

sitional modal logic, considering some of the motivation for it and

different interpretations of it, as well as developing the formal ap-
paratus. Then, in chapters 3 and 4, we do the same for quantified
modal logic.
 1
PROPOSITIONAL LOGIC

This book uses a natural deduction approach to validating modal

inferences. We evaluate an inference by symbolizing it and deducing
the symbolized conclusion from the symbolized premises, via simple
and immediately obvious inference rules. For this approach we need
rules for introducing and eliminating the various connectives in a
proof. We will also use a rule of replacement, which enables us to re-
place equivalent expressions whenever we want to. We will consider
an inference valid just in case there is a way of deriving the conclu-
sion from the premises according to these rules.
The alternatives are the axiomatic and semantic approaches. In
an axiomatic approach, we specify a set of axioms and simple rules
for deriving theorems. We then evaluate an inference by extracting
its logical form and determining whether or not that form is deduc-
ible from the axioms. This is a workable but a difficult and less in-
tuitive way to proceed. The semantic approach is illustrated by truth
tables in propositional logic. There we could evaluate an inference
by extracting its logical form and then determining via a truth ta-
ble, truth tree, or some other semantic device, whether or not that
form was tautologous. These methods evaluate statement forms by
assigning an interpretation, usually a truth value, to the variables,
treating the connectives as mechanical ways of combining these val-
ues, and calculating the value of the statement form. This method
can be very cumbersome and nonintuitive, but it has the virtue of
providing us with a decisive method of showing argument forms
valid and invalid. These semantic tests are examples of what are usu-
ally called effective procedures - they will produce a yes or no answer
to our question about validity in a finite number of steps by a me-
chanical method.
Any one of these three approaches can be used to determine the
validity of inferences in propositional logic. In choosing among them,

3
4 PROPOSITIONAL Lome

therefore, we base our choice on factors like naturalness and ease of

use. We use the natural deduction approach because the reasoning
in symbols parallels the way we might argue in English for the va-
lidity of an inference. Usually it is the easiest way as well. Later on
there will be occasion to look at axiomatizations and semantics.
We want a set of rules, then, that permit easy, straightforward,
and fairly natural derivations. If simplicity and elegance were our
only considerations, we would only need rules for introducing and
eliminating our primitive operators. But simplicity and elegance take
a back seat when our objectives are naturalness and convenience.
Consequently some of our rules and equivalences will be redundant
-we could get along without them.
We now review propositional logic in order to specify the symbol-
ism and the rules to be used in this book.
The system of propositional logic presented here is the one found
in most logic books. Different authors use different symbols or dif-
ferent rules, but each set of rules and each symbol system yields the
same results when applied to an argument. This system of logic will
be the base on which we build our modal logic.
What is presented here under the rubric of propositional logic is
frequently labelled sentential logic or statement logic in other text-
books. Part of the reason is metaphysical controversy about what
things are really the bearers of truth and falsity. 1 Here the word
'proposition' will ordinarily be used to refer to what are, properly
speaking, the sorts of things that are true or false. This usage, how-
ever, is intended not to beg any metaphysical questions. Those who
prefer 'sentence' or 'statement' should be able to substitute their choice
salve veritate.
Since logic is customarily developed as a symbolic language, we
will here refer to the symbolic representations of propositions as sen-
tences of our symbolic logical language. These sentences will be spo-
ken of as having truth values. Likewise, we will often refer to propo-
sitions (without trying to prejudice the issue of exactly what sorts
of things they are) by means of sentences that are ordinarily used
to express or assert them.

1.1 Symbolism

We use the following symbols for the operators:

 RULE OF REPLACEMENT AND EQUIVALENCES 5

OTHER COMMONLY
SYMBOL OPERATION EXAMPLE USED SYMBOLS

negation ~A ,A, -A, A

:) implication A:)B A-B
& conjunction A&B AAB, A.B, AB
V disjunction AvB
(or alternation)
= material equivalence A= B A-B

Roman capitals from the beginning of the alphabet represent

statements which are either true or false: A, B, C D . ... Lower-case
letters from the latter part of the alphabet represent well-formed for-
mulas: p, q, r, .... Rules of inference and equivalences are stated in
terms of p, q, r, etc., since the rules or equivalences apply to any for-
mula exhibiting the structure under consideration.
We adopt the usual semantic, i.e., truth table, definitions of the
operators or connectives.

p,q pvq p,q p&q

T T T T T T
T F T T F F
F T T F T F
F F F F F F

p, q p:)q p' q P=Q

TT T TT T
T F F T F F
FT T FT F
F F T F F T

1.2 Rule of Replacement and Equivalences

The rule of replacement enables us to replace an expression with

an equivalent expression. The expression being replaced may consti-
tute either a part or the whole of a sentence or a well-formed formula.
Such a replacement does not alter the truth value of the sentence
within which it is made. The following is a list of familiar equivalences
which we will use.
6 PROPOSITIONAL Lome

1. De Morgan's ~(p&q) = (~pv ~q)

Theorems ~(pvq) = (~p& ~q)
2. Commutativity (p&q) = (q&p)
(p V q) = (q V p)
3. Associativity (p&(q&r)) = ((p&q)&r)
(p V (q V r)) = ((p V q) V r)
4. Distributivity (p & (q vr)) = ((p &q)v(p & r))
(pv(q&r)) = ((pvq)&(pvr))
5. Double Negation p = ~~p
6. 'fransposition (p :Jq) = (~q:) ~p)
7. Material (p:Jq) = (~pvq)
Implication (p:Jq) = ~(p & ~q)
8. Material (p=q) = ((p:Jq)&(q:Jp))
Equivalence (p=q) = ((p&q)v(~p& ~q))
9. Negation of Mate- ~(p=q) = (~p=q)
rial Equivalence ~(p =q) = (p = ~q)
10. Exportation ((p & q) ::, r) = (p ::, (q::, r))
11. Absorption (p:Jq) = (p:J(p&q))
12. Tautology p = (pvp)
p = (p&p)
Obviously, this list could be extended if we so desired.

1.3 Rules of Inference

Rules of inference permit us to obtain new formulas - "new" in the

sense that they need not be logically equivalent to any prior formula
occurring in a proof. Unlike the rules for replacement of equivalences,
rules of inference may not be applied within formulas or to parts of
formulas.
Given the rule of replacement and our list of equivalences, the
only rules of inference we require are modus ponens and implication
introduction. However, as we have already said, elegance (using as
little apparatus and as few rules as possible) is not our objective here.
We adopt the following rules because they correspond to basic
and familiar inference patterns in everyday life, and they are the most
convenient for us to use in our proofs:

1. Modus ponens 2. Modus tollens

(Implication elimination) p :Jq
p:Jq ~q
p :. ~p
:.q
 RULES OF INFERENCE 7

3. Transitivity 4. Disjunctive introduction

p-::Jq (Addition)
q -:Jr p
:.p-::Jr :.pvq
5. Disjunction elimination 6. Conjunction introduction
pvq p
~p q
:.q :.p&q
7. Conjunction elimination
(Simplification)
p&q
:.p

To these rules we add two more which will help simplify our proofs.
These rules also have the virtue of being "natural" -we often use them
in ordinary deductive reasoning. The first is the rule of implication
introduction, often called "conditional proof." This rule provides a
way of introducing implication statements into a proof.

8. Implication introduction (Conditional proof)

p At any line in a proof, we may enter any for-
mula as a line with the justification: assump-
tion. When we enter such a line we indicate
the assumption by drawing a horizontal line
q
beneath it. We indicate the scope of the as-
sumption by drawing a vertical line to its left,
and continuing the line downward beside all
r
subsequent lines until the assumption is dis-
charged or terminated. An assumption is dis-
charged or terminated by ending the vertical
s
line and entering as the next line (or lines)
of the proof a conditional (or conditionals),
whose antecedent is the assumption and
P -::Jq impl intro whose consequent is any line within the scope
of the opening assumption that is not within
the scope of any subsequent assumption.
P -::Jr impl intro More than one conditional may be entered
on the basis of a single assumption. The jus-
tification for such lines is implication intro-
p -::Js impl intro duction.
8 PROPOSITIONAL Lome
All the usual rules apply within the scope of an assumption, but
no lines within the scope of a discharged or terminated assumption
may be used to derive subsequent lines in a proof. No assumption
may be discharged or terminated unless every assumption within its
scope has been discharged or terminated.
An assumption also may be terminated by ending the vertical line
indicating the scope of the assumption, entering no conclusions on
the basis of the assumption, and proceeding with the proof as though
the assumption had not been made. In such a case, we will say that
the assumption has been terminated although it has not been dis-
charged. This may be done when it becomes apparent that the as-
sumption is not useful in proceeding with the proof.
Here are two ways of proving the same formula illustrating impli-
cation introduction:
A--.:JB A--.:JB
C--.:JD :.(A&C)--.:J(B&D) C--.:JD :.(A&C)--.:J(B&D)
A&C asp A asp
A conj elim B MP
B MP µ,;;. asp
C conj elim
D MP j~&D ~:j intro
B&D conj intro C--.:J(B&D) impl intro
(A&C)--.:J(B&D)impl intro impl intro
A--.:J(C--.:J(B&D))
(B&D) exp
(A&C)--.:J
Note that, by the statement This illustrates a use of impl
of rule of impl intro, we intro within the scope of an-
could legitimately infer, e.g., other usage of impl intro.
(A&C) --.:JC,(A&C)--.:JB,etc.,
although in this problem
there is no reason to do so.
9. Reductio ad absurdum (Negation introduction)
This rule may be considered an abbreviation of a special appli-
cation of implication introduction. We introduce it for conve-
nience and because it parallels an ordinary and commonly used
pattern of deductive inference. It is exactly like the rule of im-

any formulas q and ~ q, the assumption

may be discharged and ~ p entered as the
r:
plication introduction, except that it gives different instructions
for discharging the assumption. If within the scope of an as-
sumption p, there occur as lines in the proof
q
~q
next line of the proof. ~p RAA
 PROOFS (OR DERIVATIONS) 9

Note than when a line within the scope of an assumption contra-

dicts a line occurring outside, providing that the line outside the scope
of the assumption does not occur within the scope of some other ter-
minated or discharged assumption, equivalence (5) (double negation)
enables us to bring the contradictory line within the scope of the as-
sumption. Although "officially" required, this step may be eliminated
in writing proofs.
The rule allows us to conclude that whatever implies a contradic-
tion is false.

1.4 Proofs (or Derivations)

We have already been speaking of proofs, for the point of intro-

ducing these rules is the construction of proofs. A proof or deriva-
tion of a formula p is an ordered finite sequence of formulas, every
member (or line) of which is either an assumption, a premise, or de-
rived from previous lines by the rules, and whose last member (or
line) is p, where this occurrence of p is not within the scope of any
assumption.
An argument form is valid just in case there is a proof (or deriva-
tion) of its conclusion from its premises. And a formula p is a theo-
rem of propositional logic if and only if there is a proof of p which
uses no premises.

Example: 1. Ip ::, q assumption

2. Iq::) r assumption
3. p::) r 1,2 Transitivity (Rule 3)
4. (q ::Jr) ::J(p ::Jr) impl intro
5. (p ::Jq) ::J((q ::Jr) ::J(p ::Jr)) impl intro
Line 5 is a theorem.

1.5 Exercises

(There is no need for exercises of the usual sort here. Those given
are related to remarks made in the text.)
1. Show that the rule modus tollens is redundant (i.e., give a
derivation of ~ p from p ::Jq and ~ q that does not use the
rule modus tollens). Do the same for:
rule 3- Transitivity
rule 4- Disjunction elimination
rule 9- Reductio ad absurdum
10 PROPOSITIONAL Lome

2. Show that the following formulas are theorems:

(a) ( ~ P =>~ q) =>((~ P =>q) => p)
(b) (p :> (q :> r)):) ((p :> q) :> (p :> r))
(c) = =
(p (q :> r)) (((p & q) :> r) & ((q :> r) :>p))
(d) p:) (q V ~q)
3. Note that among our rules of inference, we have rules for
introducing and eliminating each of our logical connectives,
except for negation. Why is there no negation elimination
rule among our rules of inference?
 2
PROPOSITIONAL MODAL LOGIC

Propositions, we have said, are either true or false. We have ob·

served that there are propositions that, though true, could have been
false, and they are distinct from those true propositions that could
not have been false. Likewise, among false propositions there are those
that could have been true and those that could not. Some truths are
contingently true while others are necessarily true; some falsehoods
are contingently false while others are necessarily false or impossible.
Aristotle and the medievals thought of necessity, possibility, or con·
tingency as modes or ways propositions may be related to truth or
falsehood.
This list of modal notions fails to highlight an important subclass
of necessary truths - those that are statements of entailment or nec-
essary implication. In making assertions to the effect that one claim
entails another, philosophers (and others as well) are claiming that
some conditional is necessarily true. The antecedent entails the con·
sequent; the statement that the antecedent implies the consequent
is a necessary truth. Not all conditional sentences express this kind
of conditional, as we shall see in the section on symbolization, but
claims of this kind are common in philosophy and theology. Interest
in the conditionals of this kind has provided a strong impetus to the
development of modal logic in this century. Some of the basic argu·
ments, however, apparently originated among the Stoics and were
carried on by the Medievals.
Some grasp of these modal concepts is essential not only for under·
standing modal logic, but also for understanding the concept of va-
lidity, which is the central concept of logic. The usual definition of
validity goes something like this: an argument form is valid just in
case it is not possible for an argument of that form to have true prem·
ises and a false conclusion, or a form is valid if the conclusion nee·
essarily follows from the premises. If we were to say instead that an
argument form is valid if and only if there is (and has been) no argu·

11
12 PROPOSITIONAL MODAL LOGIC

ment of that form having true premises and a false conclusion, then
we might be committed to accepting as valid an argument form that
is patently invalid, just because no instance of it actually having
true premises and a false conclusion has ever been presented or
thought of.
Clearly the validity of an argument form must not depend solely
on what arguments may or may not have been given in that form.
The validity of an argument form is supposed to guarantee that we
cannot use it to go from truth to falsehood. But then the modal term
has been introduced into the definition, and we cannot get rid of it
and still have what we want. Some primitive understanding of modal
terms is essential to understanding the central concept of logic.
This chapter is concerned first with clarifying our grasp of the
modal concepts and with the idea of a modal logic. The bulk of the
chapter, however, is devoted to explaining three closely related sys-
tems of modal logic and developing facility in using them. The chap-
ter concludes with remarks about some metaphysical assumptions
implicit in our treatment and with some consideration of whether
there is a "correct" system of modal logic.

2.1 The Modal Concepts

2.1.a An Explanation of Necessity

Here we want to explain something of what we mean when we call

propositions necessary, possible, or contingent. Since these notions
are interdefinable, we concentrate on just one: necessity.
Most philosophers have thought there are some necessary truths.
But there the widespread agreement ends. Which truths are neces-
sary? Why are necessary truths necessary? How can necessary truths
be characterized? The theories proceed in every direction.
Furthermore, there are different things one might legitimately
mean by necessity. Th call something a necessary truth is to imply
that it somehow has to be true. But we sometimes say a claim is
necessary when it is the consequence of certain antecedent conditions
that are not within our (or anyone's) power to change. Perhaps inevi-
table is a more accurate word to describe this usage. But this usage
is not exactly what we have in mind here, for it is possible that the
antecedent conditions should have been different than they were, or
that the laws relating these conditions to the claim under considera-
tion could have been different. For example, that which has already
 TuE MODAL CONCEPTS 13

happened in the past is often described as necessary in this sense

of being inalterable, but the truth of a statement about a past event
will not be necessary in the sense we have in mind. Likewise, physi-
cal and causal laws, sometimes said to be necessary, could be other-
wise, and are therefore not necessary in the sense we are trying to
describe. The same goes for such claims as
(1) No human being can pole vault 1000 feet (under normal
conditions with a standard pole, etc.).
Here we want to think of necessity as logical necessity, or as Al-
vin Plantinga describes it, "broadly logical necessity." Broadly logi-
cally necessary truths are those whose denials are self-inconsistent.
Included in this group are truths of logic, for example, truths prov-
able in first-order logic, such as,
(2) For any statements p and q, if p is true and p implies q,
then q is true.
and (3) If all men are mortal and Socrates is a man, then Socrates
is mortal.
So are truths of mathematics and set theory, such as
(4) 7+5=12,
and (5) The union of a set A with a set B is the same set as the
union of B with A.
There is another group of truths, difficult to characterize, often la-
belled "analytic." Leibniz tries to characterize them by saying that
the concept of the subject contains the concept of the predicate, e.g.,
(6) All brothers are siblings,
and (7) All bodies are extended.
Leibniz thinks that all necessary truths are analytic, but we will not
make that rash claim here. We leave open the question of whether
or not truths of logic and mathematics are analytic.
There is another broad class of necessary truths that is hard to
characterize and whose members are sometimes difficult to pick out.
They are not truths of logic or of mathematics, nor are they analytic.
Yet their denials seem to lead directly to contradictions. Consider
the assertion that some propositions are true. Its denial is
(8) All propositions are false.
But that implies that proposition (8) is false. So if it is true, it is
false, and of course if it is false, it is false. Therefore it is false, and
14 PROPOSITIONAL MODAL LOGIC

its denial, some propositions are true, is true and necessarily so. Still
other examples seem equally necessary, but their denials do not lead
to such obvious arguments.
(9) Nothing is both red all over and green all over at the same
time;
(10) Nothing weighs more than itself;
(11) No numbers are human beings.
Still other candidates for this title of broadly logically necessary truth
are hotly disputed, both as to their truth as well as to their neces-
sity. Here are some examples:
(12) Man is a rational animal;
(13) There is no private language;
(14) For everything, there is a reason why it is so and not other-
wise;
(15) Nothing happens without a cause;
(16) In an infinite stretch of time all possibilities (potentiali-
ties) are realized.
Although this concept of a logically necessary truth cannot be
defined precisely and simply, the above examples help to communi-
cate the concept. It is also helpful to look at some nearby concepts
which should not be confused with it. Broadly logical necessity has
already been distinguished from physical or causal necessity. But
there are several more concepts to be distinguished from logical pos-
sibility. Once again our discussion draws heavily on Plantinga (The
Nature of Necessity, chap. 1).

"Ungiveupable"

Our sense of necessity does not mean "ungiveupable." W. V. 0.

Quine, a leading contemporary philosopher, has urged philosophers
to give up the analytic-synthetic (or necessary-contingent) distinc-
tion in favor of a notion of relative revisability. There are propositions
that we are extremely reluctant to believe false, while we would quite
readily reject others, given certain appropriate evidence. For exam-
ple, a person may find a truth of logic almost impossible to give up,
while she might be willing to give up the belief that there is a full
cup of coffee on the desk, if she looks at it and it appears empty,
even though she was sure she had just refilled it. But, Quine main-
tains, no propositions are in principle "ungiveupable," provided we
are willing to adjust the rest of our beliefs.
 THE MODAL CONCEPTS 15

Clearly we cannot equate relative "ungiveupability" with neces-

sity, even though it may be a fairly reliable clue to necessary truth.
For a belief may be ungiveupable for reasons that have nothing to do
with its truth, and are unrelated to the necessity of its truth. A belief
may be a guiding belief for a person's life to such an extent that the
person may be psychologically incapable of giving it up. Or I may
find the belief that I exist ungiveupable, but its "ungiveupability"
has nothing to do with the necessity of its truth. Conversely a propo-
sition may be a necessary truth, yet not have been thought of by any-
one, and so not have achieved the status of being "ungiveupable."

"Unable to be rationally rejected"

Is a necessary truth one that cannot be rationally rejected? No,

for we can describe circumstances that make it clear that these con-
cepts are distinct. I cannot rationally reject the proposition that I
am thinking, but it is not a necessary truth. And given that true
mathematical propositions are necessary, suppose that someone
whom I know to be a notorious liar, has just told me that Zorn's
Lemma is logically equivalent to the axiom of choice. Suppose fur-
ther that I am totally ignorant of set theory and I know he has some
reason for wanting to deceive me on this occasion. Here I could ra-
tionally reject what is in fact a necessary truth.

"Self-evident"

Even if we had a sense of "self-evident" that did not also include

contingent propositions, we still could not say that all necessary
truths are self-evident. The example of complicated theorems of higher
mathematics suffices to show this. Could we say that a proposition
is necessary just in case it is either self-evident itself or deducible
from self-evident propositions by means of self-evident rules of infer-
ence? Consider Goldbach's conjecture-that every even number is the
sum of two primes. It is, if true, necessarily true, and if false, neces-
sarily false. But which is it? There is no self-evident answer and none
is self-evidently deducible from self-evident truths, so far as we know
right now.

A priori

Is this hoary category the clue to picking out necessary truths?

One of the chief difficulties with this category of the a priori is under-
16 PROPOSITIONAL MODAL LOGIC

standing it. If we mean by it those truths that are known prior to

sense experience, we can quickly answer our question No, since plenty
of necessary truths are not known. Suppose then we say an a priori
proposition is one knowable by someone independently of experience.
Now we have a much more difficult question on our hands. Is God
one of the persons whose knowing powers are under consideration
here? If He is, then presumably all truths are a priori. Let us limit
ourselves to human knowers. The philosopher Leibniz thinks that
while God can know everything about every individual a priori, hu-
man knowers, due to the finitude of their knowledge, cannot. Human
knowers, however, can know that they think and that they exist inde-
pendently of sense experience. But these are not necessary truths.

All of these would-be explanations of necessity that we have re-

jected, except that of causal necessity, are epistemic notions. But the
concept of necessity is an ontological one, not relative to anyone's
actual knowledge or ability to know. Our illustrations have taken ad-
vantage of that fact.

2.1.b The Possible Worlds Picture

The great modern rationalist Leibniz suggests another way of

thinking of necessary truth. Leibniz introduces possible worlds as
different ways God could have created the universe, or, to put it a
little differently, possible worlds are alternative universes God could
have created. Leibniz observes that God could have created different
individuals from the ones that presently populate this world, he could
have changed the natural laws, and he could have decided not to cre-
ate at all. So a possible world is a total way things could be or could
have been, and a necessary truth is a proposition that is true in every
possible world.
Plantinga takes Leibniz's idea and tries to make it a bit more pre-
cise. He calls a possible world a maximal or complete possible state
of affairs. He calls a possible state of affairs, S, maximal or com-
plete just in case for every state of affairs, S either includes it or pre-
cludes it.
This Leibniz-Plantinga account does not provide an alternative
way of explaining necessary truth, for it presupposes that anyone
who uses it already understands the notion of possibility in the
broadly logical sense. But if we have some grasp of that, we can un-
derstand what necessity in the broadly logical sense is, and we do
not need to introduce worlds and maximal states of affairs. Of course
 THE MODAL CONCEPTS 17

Plantinga does not give this account as an additional or independent

way of explaining necessity, but as a help in understanding it.
If we have some idea of what is meant by calling a proposition
a necessary truth, we can understand the other modal concepts by
defining them in terms of necessity. A proposition is possible (or pos-
sibly true) just in case it is not necessary that it be false. Obviously
then, a proposition is impossible just in case it is necessary that it
be false. A contingent proposition is one which is both possibly true
and possibly false. One proposition entails another if and only if it
is not logically possible for the former to be true and the latter false.
A pair of propositions is compatible or consistent if and only if their
conjunction is possible, and incompatible or inconsistent just in case
their conjunction is impossible. A pair is contradictory just in case
both cannot be true and both cannot be false.

2.1.c A Map of the Modal Concepts

Let us try to picture some of the relationships among these con-

cepts. We begin by dividing propositions into two groups - true and
false. Each of these groups can be divided again into necessary and
contingent, yielding necessary truths and contingent truths, and nec-
essary falsehoods and contingent falsehoods. The necessary falsehoods
are the impossible propositions; the necessary truths together with
the contingent truths and contingent falsehoods comprise the pos-
sible propositions. 'lhte propositions expressing entailments are nec-
essary truths, although propositions standing in the relationship of
entailment to each other need not be necessary, nor need they be true.

TRUE FALSE

NECESSARY

________ I
"'°oo>-+----------
Impossible

o>..,6
(0
CONTrGENT
18 PROPOSITIONAL MODAL LOGIC

It may be helpful to have a few examples of the various categories:

contingently true:
The average annual snowfall in Grand Rapids is 76
inches;
The Detroit Tigers won the 1984 World Series;
contingently false:
The Houston Oilers won Super Bowl XVII;
The earth is not more than 40 million miles from the
sun;
impossible:
7+5=13;
There is a barber in Seville who shaves all and only
those persons in Seville who do not shave them-
selves;
entailment:
That I am wearing a blue sweater entails that I am
wearing a sweater;
That I walk entails that I exist.

2.1.d Symbols and Definitions

We introduce the symbol D as a sentence-forming operator on sen-

tences. What falls within the scope of a D must be a sentence of our
logical language, and the result of prefacing a sentence with a D is an-
other sentence. D p means that p expresses a necessary truth. ◊ is
the possibility operator; ◊p means that p expresses a possible truth.
Now we can express symbolically the definitions given at the end
of 2.1.b:
p is possible ◊p=df~ □ ~p
p is necessary □ p=df~◊~p
p is impossible ~◊p=df □ ~p
p is contingent =df (◊p& ◊ ~p)
p entails q p- q =df~◊(p& ~q)
=df □ (p::)q)
p and q are compatible =df ◊(p&q)
p and q are incompatible =df ~◊(p&q)
p and q are contradictory =df ~◊(p&q)&~◊(~p&~q)
or (p- ~q) & (~p-q)
Aristotle is the first philosopher we know of to discuss these modal
relationships. In De Interpretatione, chapters 12 and 13, Aristotle
 THE MODAL CONCEPTS 19

struggles with the interrelationships among the modal notions and

their negations. His account is marred by his failure always to
distinguish clearly between possibility and contingency, but he even-
tually arrives at most of the interrelationships we have presented in
our definitions and our previous discussion. For example, Aristotle
discovered the equivalences (definitions)
~◊p=df □ ~p,
~◊ ~p=df □ p,

and inferences
□ p, l:.◊p,
◊p, l:.~~◊p,
and p, !:. ◊p.
What interests us here is not how Aristotle knew these, nor whether
he understood "possibility" and "necessity" in exactly the same way
we use these terms, but rather that he saw logical relationships to
be worked out.

2.1.e Exercises (adapted from Aristotle, De Interpretatione,

chap. 13)

A. Symbolize the following and show which groups contain state-

ments that are all equivalent to each other:
1. It is possible that 'lbm is rational.
It is not impossible that 'lbm is rational.
It is not necessary that 'lbm is rational.
2. It is not possible that 'lbm is a mathematician.
It is impossible that 'lbm is a mathematician.
It is necessarily false that 'lbm is a mathematician.
3. It is possibly false that 'lbm is a cyclist.
It is not impossible that it is false that 'lbm is a cyclist.
It is not necessarily false that 'lbm is a cyclist.
4. It is not possibly false that 'lbm is bipedal.
It is impossible that it be false that 'lbm is bipedal.
It is not necessary that 'lbm is bipedal.
B. Give the contradictories of:
1. It is necessarily false that 'lbm is a mathematician.
2. It is impossible that it be false that 'lbm is bipedal.
3. It is not possible that 'lbm is a mathematician.
20 PROPOSITIONAL MODAL LOGIC

C. Are the following pairs compatible, incompatible, or contradic-

tory?
1. pis true.
p is possibly true.
2. pis true.
p is possibly false.
3. p is necessarily true.
p is possibly true.
4. p is necessarily true.
p is contingently true.
5. p is necessarily true.
pis false.
6. p is necessarily true.
p is necessarily false.
7. p is possibly true.
p is possibly false.
8. p is possibly true.
p is necessarily false.

2.1.f Symbolizing Conditionals

The expressions "if... then ... ," "... implies ... ," and "... en-
tails ... " are often used, especially in philosophical and theological
literature, in the sense we represent by our - . However, we should
take note of certain important exceptions, so that we do not repre-
sent as modal arguments arguments that were never so intended by
their authors. Sometimes conditionals may express causal relation-
ships:
If hydrogen and oxygen are mixed in the presence of a
flame, then there will be an explosion;
If you plant beans, then you will not get broccoli.
Sometimes they express subjunctive conditionals:
If the Papacy had remained in Avignon, then today's Pope
would be a Frenchman;
If Bach had written the 1812 Overture, he would have
known the melody of the Marseillaise.
And there are other kinds of conditionals:
If Jager buys a bicycle, it will be either a Schwinn or a Fuji;
If you sit in the front of the plane, then you may not smoke;
 THE MODAL CONCEPTS 21

If you took the $100 from Jones' drawer, then you have vio-
lated the law;
If you expect your share of the inheritance, then you must
stop your carousing.
None of these expresses a conditional that is a necessary truth; in
each case it is possible for the antecedent to be true and the conse-
quent to be false. Yet each of these may, in the proper context, ex-
press a true conditional.

2.1.g The Necessity of the Consequence

and the Necessity of the Consequent

During the Medieval period, the ability to distinguish the neces-

sity of the consequence from the necessity of the consequent became
a required item in the philosopher's repertoire. St. Thomas Aquinas
makes a typical use of this distinction to resolve a puzzle about God's
foreknowledge and human free will in his Summa Contra Gentiles,
part I, chapter 67. But even St. Augustine and Boethius show an
awareness of this distinction, though less clear than St. Thomas's,
in their treatments of the same puzzle many centuries earlier. 2
Basically, the distinction is between two different ways of taking
the scope of necessity in a conditional. The occurrence of the word
'necessarily' in a conditional may signal the necessity of the connec-
tion between the antecedent and the consequent of that conditional,
that is to say, the necessity of the consequence. Or 'necessarily' may
indicate the necessity of the consequent of the conditional. Repre-
senting the situation symbolically, we are distinguishing
□ (p ::)q) necessity of the consequence
from
(p :::>□ q) necessity of the consequent.
The symbolization above makes it immediately clear that there is a
distinction to be made and makes clear what the distinction comes
to. Many earlier writers seem to think there are two different kinds
of necessity involved. Once we see the distinction represented sym-
bolically, it may become difficult for us to see how astute philoso-
phers have managed to confuse these two readings.
This confusion is often engendered by the ambiguous ways of ex-
pressing necessity in conditionals in natural languages. For exam-
ple, consider the following claim put in the mouth of Evodius by St.
Augustine in the latter's On the Free Choice of the Will:
22 PROPOSITIONAL MooAL Lome

"Since God foreknew that man would sin, that which God foreknew
must necessarily come to pass."
What is the scope of "necessarily" here? Is Evodius claiming that
If God foreknew that man would sin, then it follows neces-
sarily that man would sin (or, equivalently, It is necessary
that if God foreknew that man would sin, then man would
sin),
or that
If God foreknew that man would sin, then it is (a) neces-
sary (truth) that man would sin?
The philosophical question being discussed in the passage is whether
God's foreknowledge necessitates what he knows. Which reading we
choose could make all the difference in how we answer that question.
However, since the question of God's knowledge gives rise to addi-
tional complications and confusions, it might be better to pursue the
question of interpretation with the help of a different exampl~.
Most philosophers have held that it is impossible to know, actu-
ally know, something false. In other words, if I know p, it necessarily
follows that p is true, i.e., □ (S knows p ::>p is true). I might, on some
occasion, express this truth by saying that if I know there is coffee
in my cup, then necessarily there is coffee in my cup. However, this
way of expressing myself might lead someone to think what I said
expresses the necessity of the consequent:
((I know there's coffee in my cup)::> □ (there's coffee in my
cup)).
But a brief moment's reflection is enough to lead us to reject that
as a serious interpretation, assuming that I was trying to say some-
thing true. For if this reading were correct, I would generate a quick
but specious argument for the necessary .truth of everything I know
or anyone knows.
In spite of what may now seem obvious, there is a strong and re-
curring temptation in philosophy and theology to transfer the neces-
sity of the connection in a conditional to the consequent of the con-
ditional. It is important to resist this temptation. Indeed, it is a good
rule of thumb in translating from natural language into symbols to
suppose that necessity expressed in a conditional should be taken
as expressing necessity of the consequence, unless there is overwhelm-
ing specific evidence to the contrary.
 1'iIE Loa IC OF MODALITY 23

Remember that to assert the necessity of the consequence is to

assert that it is broadly logically impossible for the antecedent of
the conditional to be true while the consequence is false. Asserting
the necessity of the consequent, on the other hand, is to claim that
the antecedent implies the necessary truth of the consequent.

2.1.h Exercises

A. Which of the following express necessary conditionals?

1. If you study the history of music, then you cannot omit Bach.
2. If Gacy killed those young men intentionally, then he is guilty
of murder.
3. If a triangle is equilateral, then it is necessarily equiangular.
4. If Hitler had conquered Russia, then Khrushchev would not
have come to power.
5. If the millage does not pass, the Superintendent will have to
resign.
B. Symbolize the following:
1. If God does not exist, then he cannot come into existence.
2. If God does not exist, then his existence is impossible.
3. If God exists, his existence is necessary.

2.2 The Logic of Modality

2.2.a The Need for a Logic of Modality

Earlier we said that modal logic was the logic of these notions of
possibility and necessity. Now that the concepts are before us, what
should a logic of these concepts be like and what should it tell us?
Clearly, a modal logic should provide us with a way of exhibiting
the logical structure of those inferences that use modal concepts in
a way which affects their validity. There is no need to introduce more
apparatus than we have in ordinary propositional logic to validate
inferences of the form
□p:::) □ q ◊p:::)◊q
□p or ◊p
:. □ q :.◊q
since both are simply instances of modus ponens.
There are, however, some inferences in which these modal concepts
obviously play an important role, and the validity of those infer-
24 PROPOSITIONAL MODAL LOGIC

ences cannot be shown in ordinary propositional logic. Consider the

inference
It is necessary that all black cats are black
:.(It is true that) All black cats are black,
and its counterpart for possibility,
(It is true that) Jager is a bicyclist
: . It is possible that Jager is a bicyclist.
These inferences are obviously valid. Yet we cannot show that they
are, given only the resources of propositional logic. In each case we
would have to represent the structure of the argument as
p
:.q
which is an invalid inference.
Although the above way of symbolizing the inference in proposi-
tional logic fails to show its validity, perhaps there is some other way
to represent this inference in terms of truth-functional connectives
alone that does show the validity of this inference. Can Op (i.e., p
is a necessary truth) be represented truth functionally? Our set of
connectives is functionally complete, that is, they are adequate to
represent all truth functions.
If Op can be adequately represented truth functionally, there will
have to be some sentence using only p and the usual truth-functional
connectives that is equivalent to Op and that makes valid the pat-
ently valid inferences involving Op. There are only four possible ways
to express Op truth functionally, represented here by the four truth
functions, {1-f4.
TRUTH FUNCTIONS

T T F T F
F T F F T
1. Suggestion {1 is a truth-function that makes Op true
no matter what p is. Hence it makes Op equivalent to
(p V ~p);
2. Suggestion {2 is a truth-function that makes Op false
no matter what p is. Hence it makes Op equivalent to
(p& ~p);
 THE LOG IC OF MODALITY 25

3. Suggestion fa is a truth-function that makes D p have the

same truth value asp. Hence it makes D p equivalent top;
4. Suggestion f4 is a truth-function that makes D p have the
opposite truth value of p. Hence it makes D p equivalent
to ~p.
Obviously each of these is unsatisfactory as a representation of D p.
1. Suggestion f1 fails to make valid D p ::,p, which is surely
valid. (p v ~p) :J p, for example, is obviously invalid.
2. Suggestion f2 makes valid D p ::,p, because (p & ~p)::, p
is valid. But it also makes valid the obviously invalid
□ p ::, ~p, incorrectly making it equivalent to (p & ~p)::, ~p,
a valid formula.
3. Suggestion fa makes D p equivalent to p; while this vali-
dates D p ::,p, it also would validate p::, D p, which is pat-
ently invalid.
4. Suggestion f4 makes D p equivalent to ~p. This fails to
make Op :J p valid.
The failure of all four possible suggestions shows that the logic
of necessity (of our D) is not truth functional. Valid modal infer-
ences cannot be systematized using only truth-functional proposi-
tional logic. Some additional rules for modal inferences will have to
be developed.

2.2. b Logical Relationships

We mentioned earlier that Aristotle saw that statements involv-

ing modal concepts had logical relationships to each other, and that
he began to work this out. One of the things a modal logic should
provide is an account, perhaps in the form of definitions, of the di-
rect relationships between the modal concepts.
The medievals spelled out these relationships in their squares of
opposition and rules of equipollence (something like logical equiva-
lence). On the following page is a loosely translated presentation of
a scheme that was standard by the twelfth century. 3 Note that con-
tingency is still not distinguished from possibility here. Each of the
groups of sentences beside the "corners" of the square consists of
equipollent or equivalent expressions. The groups at the ends of the
diagonals are contradictory, the upper groups are contraries (they
cannot both be true), the lower groups are subcontraries (they can-
not both be false), while the sides represent subalternation (the up-
26 PROPOSITIONAL MODAL LOGIC

per group on a side implies the lower). The reader can easily verify
that this works out, except for contingency.
Modern work in modal logic, and much of the modal logic in this
book, owes its beginning to the work of C. I. Lewis. Lewis was un-
happy with material or logical deducibility. Russell and Whitehead,
when they introduced the '::>'in Principia Mathematica, called the
relationship "material implication" and this connective the "material
conditional" precisely because they recognized it differed from what
is often called implication. The main problem with it is that it seems
to allow too much: propositions turn out to "materially imply" lots
of propositions we would not want to say that they "imply." This ob-
jection can be summarized formally in the "paradoxes of material
implication":
p ::>(q :::>
p), i.e., a true proposition is implied by any propo-
sition
and ~p ::>(p :::>
q ), i.e., a false proposition implies any proposition.
That this understanding of implication led to these paradoxes was
not something Lewis discovered. The ancient Megaric philosopher

It is not possibly not the case It is not possibly the case

It is not contingently not the case It is not contingently the case
It is impossibly not the case It is impossibly the case
It is necessarily the case It is necessarily not the case

"'
Ill
contraries .,, "'
Ill
• I!,
;::, ;::,
·~ C' -<" ·~
.... o?
,!' •
'l,o
C,
....
<::! <::!
..
i:::
'I,-<
,..Q ~"
o,~· ..
i:::

-
Ill Ill
.... o"'
C',!'
o,..,,.
....
C,
<::! p <::!
..c d' ..c
;:! ;:!

"' sub-contraries "'

It is possibly the case It is possibly not the case

It is contingently the case It is contingently not the case
It is not impossibly the case It is not impossibly not the case
It is not necessarily not the case It is not necessarily the case.
 THE Loa 1c OF MODALITY 27

Philo proposed defining the conditional in just the way material im-
plication is defined-false when the antecedent is true and the con-
sequent false, and true otherwise. This was quickly criticized by his
fellow philosophers, including his teacher Diodorus Cronos. Further-
more, these ancients discussed the understanding of the conditional
that Lewis favors - that a conditional is true precisely when the nega-
tion of the consequent is incompatible with the antecedent. 4
Lewis calls his definition of implication, which we adopt, "strict
implication"; it is also referred to as "entailment":
p-qiff ~◊(p&~q), which is equivalent to D(p:::)q).
However, this sense of implication is not without apparent paradoxes
of its own. Although the paradoxes of material implication do not
hold for it, there are analogues of those paradoxes:
□ p:::) (q:::)p) a necessary truth is strictly implied by
any proposition;
~ ◊ p:::) (p -q) an impossible proposition strictly implies
any proposition.
Lewis intended his sense of strict implication to represent the re-
lationship of logical deducibility. That is, if p entails q, q should fol-
low logically from p. Yet, as one can see from the analogues to the
paradoxes of material implication, Lewis's strict implication is not
completely faithful to the relationship of logical deducibility. It per-
mits the deduction of a necessary truth, such as 2 + 2 = 4, from any
proposition.
A. R. Anderson and Nuel Belnap, Jr., have criticized Lewis's no-
tion of strict implication, while concurring with his objection to ma-
terial implication. They claim that in addition to the requirement that
there be a necessary connection between a proposition and what it
implies, the premises or antecedent must be relevant to what is im-
plied. As we have just seen, the paradoxes of strict implication, like
the paradoxes of material implication, expose the possibility of hav-
ing a conclusion strictly implied by irrelevant premises. Anderson
and Belnap have done extensive work on a system of logic designed
to remedy these defects. 5
In the face of these objections, why do we adopt the Lewis account
of entailment? Why do we accept these paradoxes?
The "paradoxes," as Lewis likes to point out, are the consequences
of straightforward and obviously acceptable rules. Thus our options
are either to accept the paradoxes or to give up some obviously cor-
rect rule. Consider this demonstration of the claim that from a con-
tradiction anything follows.
28 PROPOSITIONAL MODAL Lome
A&~A prem
A conj elim
AvB disj intro
~A conj elim
:.B disj elim
One of the rules used in this proof will have to be declared incorrect
and forbidden in a logic of "logical deducibility," if Anderson and Bel-
nap are right. Which rule should we give up? Which rule is such that
its conclusion is not "logically deducible" from its premise or premises?
We could make the point in a slightly different way. It is plausible
to regard the following as expressing properties of the entailment re-
lationship and therefore theorems of any adequate system represent-
ing entailment:
1. ((p-q)&(q-r))-(p-r) transitivity
2. (p&q)-((pvr)&q) additivity
3. ((pvq)&~p)-q disjoinability
But any system that has the above as theorems (and uses modus po-
nens and conjunction as rules of inference) yields the alleged para-
dox as a theorem:
1'. (p& ~p)-((pvq)&~p) substitution instance of (2)
above
2'. ((pvq)& ~p)-q (3) above
3'. (((p& ~p)-((p V q)& ~p)) substitution instance of (1)
&(((p V q)& ~p)-q))- above
((p&~p)-q)
4'. (p& ~p)-((p V q)& ~p) 1', 2', conjunction
& (((p V q)& ~p)-q)
5'. (p&~p)-q steps 2~4' MP
The price of eliminating the paradoxes of strict implication is high:
we must give up a rule and the corresponding theorem from the proofs
just given. Lewis found it less strange to accept the paradoxical for-
mulas than to reject any of these rules or theorems. Anderson and
Belnap find the formulas so paradoxical that they choose to give up
a rule and corresponding theorem.
They argue that the rules of disjunction introduction and disjunc-
tion elimination, and the theorems labelled "additivity" and "disjoin-
ability" are invalid as they stand. The theorems involve a logically
disreputable and unfit sense of disjunction, if one's aim is (as Lewis's
is) to come up with a calculus of entailment that accurately repre-
sents the properties of logical deducibility. In effect their criticism
 THE Loa IC OF MODALITY 29

goes like this: although the rules in question are valid for the truth-
functional 'v' introduced in the classical propositional calculus, they
permit fallacies of relevance. In any respectable sense of deducibility
the premises have to be relevant to the conclusion, but the rules in
question permit us to deduce conclusions from premises which are
patently irrelevant. E.g.,
Either Bach wrote the Coffee Cantata or there is a largest
prime number;
There is no largest prime number;
Therefore, Bach wrote the Coffee Cantata.
If we introduce restricted versions of these rules, restrictions that
forbid the deduction of irrelevancies, then we will find that we can
no longer deduce q from (p& ~p).
Here we favor Lewis's conclusion, retaining the rules Anderson
and Belnap reject and accepting the so-called paradoxes of strict im-
plication. Two of the three modal systems presented here are ones
formalized by Lewis, though none is the system Lewis himself favors.
Lewis refuses to adopt "officially" his stronger systems S4 and S5,
developed later in this chapter, because they include as a theorem
(p-q)-((q-r)-(p-r)).

Lewis's favored system, S2, includes transitivity,

((p-q) &(q-r))-(p-r),

but not the "exported" form above. Lewis is unable to believe that
(q -r)-(p -r) is logically deducible from p -q.
In moving from truth-functional propositional logic to modal logic,
we leave behind the genuinely puzzling paradoxes of material impli·
cation and obtain a system that better represents entailment or logi-
cal deducibility. Our goal, however, is not to represent strictly by
means of a formal system what "implies" or "is logically deducible
from" means in English. Rather it is to represent the logic of neces-
sity, where necessity is interpreted as ''broadly logical necessity."
We thus elect to use a system that sometimes fails to preserve
relevance. However, we may be surprised at the relevance a clever
person can find. Russell is reputed to have been challenged to prove
that the necessarily false hypothesis 2 + 2 = 5 implied that he was
identical with the Pope. Russell replied, "You admit that 2 + 2 = 5,
but I can prove that 2 + 2 = 4; therefore 5 = 4. 'laking 2 away from
both sides, we have 3 = 2; taking 1, 2 = 1. But you will admit that
I and the Pope are 2. Therefore, I and the Pope are one. Q.E.D."
30 PROPOSITIONAL MODAL Lome

2.2.c Systems of Propositional Modal Logic

As I already have suggested, there are many different systems of

modal logic. These are different not merely in the sense that they
have different rules or axioms. For different axiom sets can still yield
the same system (i.e., the same set of theorems) and different sets
of rules can still validate exactly the same inferences. Here we mean
that there are different systems in the sense of having different the-
orems and validating different inferences.
Which system a philosopher chooses depends on his philosophical
commitments and his interpretation of the modal operators. Well then,
is there not one correct system of modal logic? As is often the case,
the answer is Yes and No; Yes in one sense and No in another. 'lb see
why, we must look at the question more closely.
There are different ways to consider a logical system. We can think
about it as a system of symbols for which we give rules. We have
rules for picking out certain approved strings of symbols and rules
for manipulating these strings. If we work in this fashion, we can
study a system without any reference to its intended use. There may
be several different uses for a system, obtained by interpreting the
symbols of the system in different ways. For example, what you
learned as the calculus of propositions can also be interpreted as a
calculus of classes.
As an exercise, look at this interpretation. The letters now denote
classes: '~' becomes complementation, •-•; '&' becomes intersection,
'n'; disjunction 'v' becomes union, 'U'; ':J' becomes inclusion, 'c'; and
'=' becomes equality of classes, '='. Look back at the equivalences
in the first chapter and rewrite them as statements about classes,
paying attention to what they say under this new interpretation.
If we are speaking of a formal system and want to ask whether
it is a correct system in a more than merely formal sense, we must
know what the intended interpretation is -what the system is sup-
posed to represent. Then a system can be said to represent that cor-
rectly or incorrectly. We would determine this by looking at the theo-
rems (interpreted, of course) and considering whether or not they are
true, and whether all the truths we are interested in are theorems.
We have just seen that systems may have several interpretations.
The uninterpreted formal system for the propositional calculus pro-
vides a correct system under more than one interpretation. But it
can be incorrect under other interpretations. When we ask about the
correctness of a formal logical system and, in the present case, a modal
system, we must have in mind some intended interpretation of the
formal system.
 'IiIE SYSTEM T 31

This question is complicated in the case of modal logic because

there are more than ten formal systems which have been proposed
and studied, and there are several interpretations of the modalities.
Here we have presented an interpretation of modalities; our primi-
tive (undefined within the system) modal operator represents "broadly
logical necessity." Is there a system that is correct under that inter-
pretation? Our tentative answer is that there is one and that it is
the third of the three presented later in this chapter. For a different
sense of "necessity," one of the other systems might be correct.
Why present three systems here? Partly because of the way they
fit together. The first one, T, is built on and thus contains proposi-
tional logic. The second, S4, contains T and more. S5 contains S4
and more. Thus, if S5 is correct, none of these systems will make valid
any invalid inferences. But the latter systems validate more infer-
ences. As we go along, we will point out the differences and their
philosophical significance. Our view is that S5 is superior (more cor-
rect) because it includes as theorems truths which the others lack.
More conservative philosophers may choose one of the weaker sys-
tems. It is also very easy to give natural deduction rules for these
three systems. We can change from a weaker system to a stronger
one (more theorems) with a simple modification of one rule.

2.3 The System T

This system is the only one of the three we study which was not
proposed by C. I. Lewis. It fits so neatly and naturally into the "fam-
ily" of systems proposed by Lewis that it is regarded as a Lewis modal
system. 6 A version of it was studied by von Wright in 1951 and called
M. However, it had been proposed earlier in a different form by Rob-
ert Feys, who called it T. These two forms were subsequently shown
to be equivalent. Our presentation will differ from both of theirs.
First, recall the symbols already introduced:
1. □, an operator attaching to sentences and forming sentences,
read as "it is logically necessary that," or "necessarily." This
is attachable to any well-formed formula or sentence of propo-
sitional logic.
2. ◊, the so-called weak modal operator, also attaching to sen-
tences and read as "it is logically possible that," or "possibly."
3. -, representing a logically necessary conditional, and read as
"necessarily if ... , then ... ," or also as "entails" or "(strictly)
implies."
32 PROPOSITIONAL MonAL Lome

We will not introduce another symbol for logical contingency. When

we want to indicate that a proposition, p, is contingent, we will write
◊p&◊~p.
The rule of replacement of equivalents is retained and the follow-
ing definitional equivalences added:
□ p- ~◊~p
◊p- ~ □ ~p
(p-q)- ~◊(p& ~q)
(p-q)- □ (p::>q)
(p-q)-((p-q)&(q-p))
These formulas may replace each other wherever they occur, with
the justification definition (def).

Exercises: substitute equivalents until both sides are the same:

1. □ ((p&q)::>r)= ~◊(q&(p& ~r))

2. ~◊(Op& ~ □ q)=(~◊q-◊~p)
3. □ (p & (q v r)) = ((~p v ~q)-(p & r))

2.3.a Rules

What are the right rules for modal logic? How should you proceed
if you had to devise rules by yourself? Perhaps the best thing to do
is to look carefully at the concept you want to represent, trying to
discover on an intuitive level what its implications are. Next you
would propose some formal rules grounded in what seem to be basic
inference patterns and investigate their consequences.
When we think about broadly logical necessity this way, one ob-
vious implication, so obvious we might overlook it, is that whatever
is logically necessary is true. We represent that in our symbolism
like this:
(1) □ p:::>p.

Another characteristic of logically necessary propositions is that any

proposition that necessarily follows from a logically necessary propo-
sition is itself logically necessary. We may represent that as follows:
(2) (Dp&O(p::>q))::> □ q.

In writings on modal logic this is usually put symbolically in the

equivalent form
(3) □ (p::>q)::>( □ p::> □ q).
 THE SYSTEM T 33

Furthermore, if we reflect on what it means to be a truth-table

tautology (or theorem of ordinary propositional logic) it seems clear
that all tautologies are necessary truths. Hence, we may adopt as
a rule
(R) If p is a theorem of propositional logic or a tautology, then
□ p is a theorem of modal logic. ( I- R, /:. I- □ R).

As the clever reader no doubt suspects, we are developing the sys-

tem T of modal logic. Indeed we have developed it. Any set of axioms
and rules for the propositional calculus with (1) and (3) added as axi-
oms and (R) added as a rule constitutes an axiomatization of T. Thus
we can see that T is a comparatively fundamental system of modal
logic, inasmuch as it is based on some universal but fundamental
observations about the logical behavior of the concept of broadly
logical necessity.
Our aim in this book is to present natural deduction rules for the
systems we consider, so let us turn to that task. Since we have in-
troduced □ as our primitive (undefined) operator, we should give a
rule for introducing it and a rule for eliminating it. These rules will
be counterparts of (1) and (3),which represent our observations about
the behavior of the concept of broadly logical necessity.

The necessity elimination rule

The rule of necessity elimination (nee elim) is straightforward and

obvious, following (1):
□p
: . p (nee elim)
Stating a rule of necessity introduction will take a little more ma-
chinery. The machinery is not absolutely necessary, but once we have
it, it will facilitate giving proofs and make it easier to switch to the
other systems of propositional modal logic. We might take, our cue
from (2) above and use as our rule
□ (p::>q)
□p
:. □ q

But this rule would not be sufficient without a way of introducing

the theorems of propositional logic as necessary truths. That is, we
would use a rule like (R). And we would need to know and recall a
wide range of these tautologies or theorems.
34 PROPOSITIONAL MODAL LOGIC

Necessity introduction subproofs

Instead, we will provide a process of deducing necessarily true con-

clusions from necessarily true premises by introducing necessity in-
troduction subproofs and providing reiteration rules, to be specified
shortly. The subproof rule used here is adapted from Frederick B.
Fitch, Symbolic Logic (Ronald Press), 1952.
A necessity introduction subproof is a subordinate proof, analo-
gous to the implication introduction rule. Into this subproof, how-
ever, we may bring only lines that are necessary, i.e., lines of the form
□ p. That way we are assured that anything we deduce within such
a subproof is also necessary. The guiding insight is that whatever
follows from what is necessary is itself necessary. If we permit only
formulas that are necessary to be brought into such a subproof, then
we may conclude that anything we deduce from such formulas within
the subproof is likewise necessary. When we terminate such a sub-
proof, we may write the last line or any other line within the sub-
proof (provided it is not within the scope of a further subproof within
the necessity introduction subproof) outside the subproof and prefix
it with the □. We are permitted to bring only lines of the form □ p
into a necessity introduction subproof, a process we call reiteration.
In T, when we enter such a line within the subproof, we remove the □.
The subproof rule thus reflects the idea that anything deduced from
premises that are necessary is itself necessary.

The necessity introduction subproof rule

A nee intro subproof is opened by entering a formula, p, as a line,

provided that either
(i) entering p is justified by the ap- (i) □p
propriate rule of reiteration; □ p T-reit

q
□q nee intro
or (ii) p is the assumption for an impl
intro subproof within the scope
of the nee intro subproof. (It
(ii)
Dr
may be helpful to recall that re-
ductio ad absurdum proofs are
cases of impl intro subproofs.) r
□r
 THE SYSTEM T 35

When a nee intro subproof is opened, we will draw a vertical line with
a □ to the left of it alongside the subproof. The line is continued un-
til the subproof is discharged and indicates the scope of the subproof,
as the above schematic examples illustrate. Any subsequent line
within a nee intro subproof must be either
(i) the result of using the appropriate reiteration rule;
or (ii) an assumption for an impl intro subproof within the scope
of the nee intro subproof;
or (iii) the result of applying our rules to lines within the scope
of the nee intro subproof;
or (iv) the opening line of another nee intro subproof within the
scope of the original.

The necessity introduction rule

A nee intro subproof may be terminated at □:

any point, and any line, p, occurring within the p
subproof and which is not within the scope of an
undischarged assumption within the subproof
may be entered as a line prefixed by a □, with q
the justification "nee intro." □ q nee intro
The termination of the nee intro subproof is indicated by discon-
tinuing the vertical line marking the scope of the subproof.
The complicated-sounding restrictions are due to our desire to
maintain maximum flexibility while avoiding invalid inferences. This
is the only rule of the propositional modal system T that may take
some effort to learn. But that effort will pay off in the great ease
with which you can learn S4 and S5. 'lb obtain these systems, we
merely need to modify our reiteration rule.
We can best see how to use the nee intro rule and the companion
subproof rule by considering an example:
Prove ((p-q)& ~q)-~p
1. □ (p-q)& ~q impl intro asp
2. p-q &elim
3. □ (p::>q) def
4. p ::>q nee elim
5. ~q &elim
6. ~p MT
7. ((p-q)& ~q)::>~p impl intro
8. □ [((p-q)& ~q)::> ~p] nee intro
9. ((p-q)& ~q)-~p def
36 PROPOSITIONAL MODAL LOGIC

The T-reiteration rule

As our statement of the necessity introduction subproof rule makes

clear, the only way to bring "outside" formulas into the subproof with-
out making assumptions is by the rule of reiteration. Now we for-
mally introduce such a rule for T.
Where □ p occurs as a previous line in a proof, p may be
entered as a line within a nee intro subproof with the jus-
tification "T-reit," unless that occurrence of □ p lies within
the scope of a discharged assumption or within the scope
of a terminated nee intro subproof.
Notice that we may reiterate only formulas of the form, □ p, and when
we do, we must remove the □. It may seem that what is happening
is more like nee elim than reiteration, but after we consider S4 and
S5, it will be clearer why we choose to call it "reiteration." For the
moment, it may be sufficient to observe that when we take formulas
"out" of the nee intro subproof, we do put the □ back on them. Thus
it gives us a way to introduce the □ (necessity) onto formulas that
did not have D's on them earlier in the proof.
Why do we remove the □ ? Intuitively, what we are doing in a nee
intro subproof where T-reit is the only allowable reiteration rule is
to bring a sentence that is necessary into the subproof, deduce con-
sequences from it, and conclude that those consequences are neces-
sary. If we did not remove the □, we would in effect be deducing con-
sequences from □ p rather than from p. Furthermore, the essence of
T is that we do not assume that if p is necessary, then it is neces-
sary that p is necessary (□ p :::,□ □ p ). But without the stipulation
that when we T-reiterate we must remove the □, this formula would
be easy to prove.
Here is an example illustrating the use of necessity introduction
subproofs and T-reiteration.
1. p-q prem
2. □p prem/ :. □ q
3. □ (p:Jq) 1, def
4. □ p:Jq 3, T-reit
5. p 2, T-reit
6. q 4,5 MP
7. □q 6, nee intro

2.3.b Hints

Here are a few hints that will be quite helpful in using these rules:
 THE SYSTEM T 37

(1) Remember that in T, the only way to bring something into

a nee intro subproof is by means of T-reit. Any line brought
in must be of the form Op, and the O must be removed.
(2) Since any line and only lines of the form Op may be en-
tered with the use of T-reit, keep in mind that
(p-q)-O(p::Jq)
~◊p-O~p.

(3) You may use impl intro within a nee intro subproof. See
the earlier example. As the rule is stated, it does not re-
quire that you discharge any assumption introduced within
a nee intro subproof. If you make an assumption and then
see that it does not lead to anything useful, you may sim-
ply terminate it, and the proof you produce will still be
"official." Remember, though, that you may not apply nee
intro to a formula from an interior subproof unless it is
first removed from the scope of the assumption.
(4) Observe that by using impl intro as illustrated in the ear-
lier example, you can prove formulas of the form Op with-
out being given any premises. All formulas derivable from
our rules without the help of premises are theorems.
(5) 'fypically, you will use the rule nee intro by constructing
a nee intro subproof and applying nee intro to its last line.
However, you are not restricted to the last line. You may
find it possible, in the course of a proof, to derive several
lines within a single nee intro subproof to which you want
to apply nee intro. The present statement of the rule per-
mits you to do this without constructing a separate nee
intro subproof for each. Let us illustrate this using the ex-
ample of a conjunction that is necessary, where we want
to derive the necessity of each conjunct.
1. 0 (p&q&r&s) /.·. 0 p & 0 q & 0 r & 0 s
2. 0 (p&q&r&s) T-reit
3. p
4. q
5. r 2, conj elim
6. s
7. Op 3 nee intro
8. Oq 4 nee intro
9. Or 5 nee intro
10. Os 6 nee intro
11. Op& Oq& Or& Os 7-10 conj intro
38 PROPOSITIONAL MODAL LOG IC

(6) Note that where you have one nee intro subproof within
the scope of another, you will not be able to T-reit a for-
mula into the innermost subproof unless it has a □ remain-
ing on it in the outer subproof. T-reit only allows you to
cross one vertical line( □ I) at a time. Each one you cross
counts as a separate application of the rule, and so a □
must be removed each time.
□p □□ p
□ p T-reit (correct) □ □p T-reit (correct)

□: □:
p T-reit (forbidden) p T-reit (correct)

2.3.c Exercises

A. Give proofs:

2. p-q 3. □ (p&q)
□p :. □ q
q-r
:. □ r

4. O(pvq) 5. □ (p vq) 6. p-q

□ ~p p-q p- □ r
:. □ q :. □ q q- □ ~r
:. □ ~p

7. □ (pv(qvr)) 8. (p-r) & (q -r) (Note: The first

q-s □ (pvq) premise is not of
(pvr)-t :. □ r the form □ p.)
.'.□ (s Vt)
We typically express the theorems of the various systems in terms
of the lower case letters p, q, r, which are variables ranging over well-
formed formulas of our symbolic language. Such expressions are
schemata; any well-formed formula of the language exhibiting that
logical structure is a theorem. For example, the expression p::) p is
a theorem schema. The substitution of any wff for p yields a theorem:
A::)A
((A vB)=C):) ((A vB)=C).
 THE SYSTEM T 39

Keeping this in mind, we can see two things: (1) the theorems so writ-
ten may be taken as generalizations about all sentences and may be
read that way, e.g., 'p :::>
p' expresses the claim that every sentence im-
plies itself; (2) a proof we give of a theorem schema is a proof schema
for a proof of any substitution instance of the theorem schema.
B. Prove the following theorems. Then try to state in English the
general truth that each one expresses.
1. p-p
2. □□ p- □ p
3. (p-q)-( □ p:::> □ q)
4. ( □ p& □ q)-(p-q)
5. □ p-(q-p)
6. □ ~p-(p-q)
7. (p-(q&~q))- □ ~p
8. ((p-q)&( ~p-q))- □ q

2.3.d More Rules

It will be useful and convenient to have some rules that enable

us to work with the possibility operator directly. Here we formulate
introduction and elimination rules for possibility.

Possibility introduction

p
: .◊p poss intro

Possibility elimination

The name and formulation of this rule are due to Fitch. The rule
is more intuitively viewed as a way of transmitting possibility than
eliminating it. However we continue the pattern of designating the
rules for our operators as "introduction" and "elimination" rules, even
though it is less obviously appropriate here.
This rule enables us to introduce a possibility sign onto a sentence
entailed by a sentence which is itself possible. The basis for the rule
is the theorem

which may be read as saying that whatever is logically entailed by

something possible is itself possible.
40 PROPOSITIONAL MODAL LOGIC

◊p

Note that this is an impl intro subproof within

a nee intro subproof subject to the restrictions
described earlier.
poss elim
Greater fondness for this rule will result from trying to prove the
theorem mentioned above and the theorems illustrated below with-
out the rule. Here are some examples to illustrate proper use of the
rule:
Show ◊(p&q):>(◊p&◊q) Show ( □ p&◊q):>◊(p&q)
◊(p&q) □ p&◊q

□ 1:&q □t
p T-reit
(p&q)
◊p poss elim ◊(p & q) poss elim
◊q poss elim (Op & ◊q) =>◊(p & q)
◊p&◊q
◊(p &q):>(◊p & ◊q)

The rules just introduced are redundant in the sense they do not en-
able us to prove anything that we could not have proven without them.
We can get the effect of these rules with the rules introduced prior
to this section.
Your instructor may wish to add more rules of this sort to make
it easier to give proofs and solve exercises. It is convenient to have
some versions of modal modus ponens:

Modal modus ponens

p-q p-q
p □p
:.q :. □ q

Exercises

1. Prove the redundancy of the rules just introduced in this section.

2. Show the redundancy of Rule R: If p is a theorem of propositional
logic, then □ p is a theorem of propositional modal logic.
 THE SYSTEM T 41

2.3.e Decision Procedure

Our rules for T give us a way of demonstrating the validity of

propositional arguments that involve modal concepts. We symbolize
the argument using our connectives, capital letters to represent in-
dividual statements, and □ and ◊ to indicate the modality of the
statements. Then we see whether or not we can derive the conclusion
from the premises. If we can, the argument has a valid form. But
if we can not, we do not know if this is because the form is invalid
or because we were not sufficiently ingenious in trying to devise a
proof.
In propositional logic, if we were unable to construct a proof, we
could test the argument form by looking at the truth table for its
corresponding conditional (or more quickly, by determining whether
there is any consistent assignment of truth values that makes the
antecedent true and the consequent false). The two-valued truth ta-
bles of propositional logic provide what logicians call an effective pro-
cedure for determining whether or not a given argument form is a
tautology. That means that it is a mechanical procedure which will
give an answer to your question in a finite number of steps. Now the
truth tables were an effective procedure for telling that a formula is
a theorem or a tautology, and they were effective for telling that a
formula is not. In modal logic there is also a truth-table technique,
but it is four-valued instead of two-valued, and though it is effective
for telling that a given formula is not a theorem, the technique is
not effective for telling whether a formula is a theorem of modal
logic. This procedure is tedious and cumbersome, requiring 4n rows
in a table for a formula with n propositional variables. 7 So, for ex-
ample, a formula with 3 different propositional letters would require
a table with 4 3 = 64 rows.
Saul Kripke, a leading contemporary writer on modal logic, has
adapted a technique of E. W. Beth involving semantic tableaus. In
Appendix III, we mention some authors who develop this semantic
tableau technique for T, S4, and S5. While semantic tableaus are less
tedious than quasi truth tables, they are better suited to establish-
ing validity than uncovering invalidity.

2.3.f Counterexamples (S5)

Here we will develop a technique for giving counterexamples. This

technique, like that of devising proofs, depends on our ingenuity.
42 PROPOSITIONAL MonAL Lome

There is no series of steps for which we can give mechanical instruc-

tions and that will always produce a counterexample if the formula
is invalid. So our failure to give a counterexample will not prove va-
lidity. But a counterexample is a decisive proof of invalidity. Although
the technique has these limitations, it is interesting and useful from
a philosopher's standpoint.
One limitation should be noted immediately. The technique to be
described cannot discriminate T or S4 from S5. It can only be used
to give counterexamples to formulas which are not theorems of any
of these systems.
The intuitive idea for giving counterexamples is simple and easy
to grasp. If someone claims that all A's are B's, and we point out an
A that is not a B, then we have refuted the claim by giving a counter-
example. Suppose, for example, someone claims that all the great
philosophers were unmarried. We could give a counterexample by
pointing out that Aristotle was married. Similarly, suppose someone
claims that p necessarily implies q. We could give a counterexample
to this by exhibiting a possible situation in which p is true while q
is false. For example, if someone foolishly suggests that my shirt is
not red entails that my shirt is blue, we can show that my shirt is
not red does not entail that my shirt is blue in the way suggested.
It is possible that I am wearing a green shirt. So it is possible that
my shirt is not red while it is false that my shirt is blue. The basic
idea here is simply to find an instance or type of instance that the
universal generalization under consideration fails to cover.
This technique can be adapted to deal with modal formulas ex-
pressing inferences. Suppose we have before us the conditional for-
mula corresponding to some alleged inference. Our efforts to dem-
onstrate its validity have been unsuccessful, and we suspect it is
invalid. We can prove its invalidity by displaying a false substitution
instance- specific propositions which, when substituted properly into
the argument form, yield a falsehood.
Example:
The formula in question is (◊p & ◊q)-◊(p &q).
Let p = I am exactly six feet tall.
Let q = I am exactly five feet tall.
Each of these is possible and the conjunction of their possibilities
is true, but the conjunction of p and q is clearly impossible. Hence
the formula in this example is not a theorem.
It is also possible to give general counterexamples to modal for-
 THE SYSTEM T 43

mulas such as the one in the previous example. A general counter-

example is one which uses propositional variables rather than specific
propositions, and which in effect describes a large class of specific
counterexamples. In giving general counterexamples we use these gen-
eral assumptions:
a) there is at least one contingent proposition- ◊p & ◊ ~p;
b) some contingent proposition is true-p;
c) there is a pair of logically independent propositions -
~(p-q)& ~(q-p).

Our objective is to choose a general substitution instance of the in-

ference pattern in question that
a) makes the antecedent (or premises)
-either a necessary truth (e.g., some obvious theorem)
- or equivalent to an obvious assumption
(note that this includes the conjunction of an assumption
with a necessary truth);
and b) makes the consequent (or conclusion)
- either an obvious necessary falsehood
- or equivalent to the denial of an assumption.
Example 1:
Consider the formula we treated earlier:
(◊p& ◊q)-◊(p&q)

We have assumed that there is some proposition, p, that is contin-

gent, i.e., p & ◊~p. Consider a substitution instance in which p is
substituted for p and ~p for q:
(◊p & ◊~p)-◊(p&~p)

The antecedent is now equivalent to our first assumption, a), and the
consequent is an obvious necessary falsehood. Consequently, we have
a general counterexample to the alleged inference. Any contingent
proposition and its denial will suffice.
Example 2: (Op & q) ::J□ (p & q)
Let q be a contingently true proposition. This means we have q
and (◊q & ◊~q). Let p be the necessary truth (q v ~q). Substituting,
we obtain
(□ (q V ~q) &q)::J □ ((q V ~q) &q)
44 PROPOSITIONAL MODAL LOGIC

Looking at the consequent, we see that the expression within the

scope of the necessity sign is a conjunction whose first conjunct is
a tautology. Hence that expression is logically equivalent to the sec-
ond conjunct alone. Making this substitution yields
(D(q V ~q) &q)::) □ q.
The antecedent is now true by virtue of the assumption that q is true
and the fact that q v ~q is obviously necessary, and the consequent
contradicts our choice of q as a contingent proposition, that is, ◊ ~q
or ~ □ q.
Exactly why does this constitute a proof of invalidity? First of all,
the theorems or candidates for theorems being examined are univer-
sal generalizations. Every permissible substitution instance of a for-
mula must be true if it is a theorem. The form of every valid infer-
ence is a theorem. The technique just described provides a way of
trying to discover a type of substitution instance of a formula that
will show decisively that not all uniform substitutions within that
formula are true, and therefore it is not a theorem of modal logic.
Secondly, our substitution technique made certain assumptions. Is
that proper here? These assumptions are consistent with using this
as a disproof technique for modal logic because without these as-
sumptions there is no point to having modal logic. Furthermore, these
assumptions seem to be obvious truths.
Any propositional modal formula to which we can give a counter-
example is not a theorem and does not correspond to a valid inference.

2.3.g Exercises

A. Give proofs:
1. □ p- □ (pvq)
2. (p-q)::)(~q-~p)
3. ~◊p::) ~Op
4. (p-r)&(q-s)
□ p& □ q
:. □ rv □ s
5. (p -r) v (q -r)
□ (p&q)
:. □ r

B. Give proofs or counterexamples:

1. (p-q)::)(~◊q::)~◊p)
2. (p-q)::)(◊ ~q::)◊ ~p)
 THE SYSTEM T 45

3. (p:)q):)(~◊q:)~◊p)
4. (p:)q):)(◊~q:)◊~p)
5. [((p&q)-r)& □ qJ-(p-r)
6. [((p&q)-r)&qJ-(p-r)
7. □ p- □ (p&q)
8. ◊(pvq)-(◊pv◊q)
9. (◊p v◊q)-◊(p vq)
10. (~◊p&~◊q)-◊(pvq)
11. (p-r)v(q-r)
D(pvq)
:. □ r
12. (p-r)&(q-s)
D(pvq)
:.D(rvs)
13. (p-r) & (q -s)
□ pv □ q
:. D(rvs)
14. (p-r)v(q-s)
Dpv □ q
:. D(rvs)

C. Symbolize and prove:

1. If a proposition is possible, then its disjunction with any
proposition is possible.
2. If each of two propositions is necessary, then their conjunc-
tion is necessary.
3. If a conjunction is necessary, then each of the conjuncts is
necessary,
4. If either of two propositions is necessary, then their disjunc-
tion is necessary.
5. If one proposition is possible while another is not, then their
disjunction is possible.

D. Symbolize and prove or give a counterexample:

1. If a disjunction is necessary, then either one or the other dis-
junct is necessary.
2. If a proposition is necessary, then its conjunction with any
other proposition is possible.
3. If one proposition is necessary and another is possible, then
their conjunction is possible.
4. If one proposition is possible and another is impossible, then
their conjunction is impossible.
46 PROPOSITIONAL MODAL Lome

E. Prove or give a counterexample:

1. (p-(q -r))-((p & q)-r)
2. ((p&q)-r)-(p-(q-r))
3. ((p&q)-r)-(p-(q:::>r))
4. (◊p& ~◊q):::>~◊(pvq) (compare C.5)
5. ((p&q)-r):::>((p-r)&(q-r))
6. ((p & q)-r) :::>((p-r) v (q -r))
7. ((p v q)-r) :::>((p-r) v (q -r))
8. p-(q-p)
9. p::>(q-p)
10. p-(q::>p)
11. (p-(q-r))-((p-q)-(p-r))
12. ((p-(q-r)) & (p-q))-(p-r)

2.3.h Iterated Modalities and Reduction of Modalities

So far we have not dealt with formulas which have "piled up" or
repeated modal operators, e.g.,
◊ □□ ◊ □ p
◊ □ ◊( □ ◊p::> □ ◊◊q)
◊ □ (◊ □ ◊p& ◊◊( □ ◊ □p-q)).
Such formulas are difficult to understand. Our modal intuitions fail
us when we try to make sense of them. What sense can we make,
under our interpretation of the modal operators, of the difference be-
tween □□□□ p and □□□ p? Thus it seems desirable to look for a
way of reducing repeated modalities in such formulas, if possible.
First, we need to make more precise the concept of a repeated mo-
dality, or, as it is often called, an iterated modality. A sequence of
modal operators containing two or more modal operators consti-
tutes an iterated modality. There are several points to notice in con-
nection with this concept. First, a well-formed formula can contain
several iterated modalities, e.g., □□□ p :::>◊◊q. Secondly, the occur-
rence of one modal operator within the scope of another does not nec-
essarily constitute an iterated modality. For example, in □ (p :::>◊q),
although the ◊ occurs within the scope of the □, we do not have an
iterated modality because it is not a sequence of modal operators con-
taining two or more consecutive modal operators. Thirdly, when nega-
tions occur within a sequence of modal operators, as in □ ~◊~ □ p,
the sequence still is an iterated modality. This is because the ne-
gation always can either be removed or replaced by a single nega-
tion at the beginning of the sequence by substituting equivalences.
 THE SYSTEM T 47

The example above is equivalent to DD D p. The sequence ◊ ~ ◊ p

is equivalent to ~ D ◊ p. Finally, it is important to recall that our
- abbreviates D ( ... ::) ... ). Recalling this enables us to see that
◊(p-q) contains an iterated modality-◊D(p::)q).
Any equivalence that enables us to reduce the number of modal
operators in a sequence of iterated modalities by replacing it with
a shorter sequence is called a reduction law.
These concepts enable us to put some of our questions more pre-
cisely:
1. How are we to understand (or read or interpret) iterated
modalities?
2. Do we have any reduction laws in T?
3. Are there any intuitively correct reduction laws for the
modal concepts as we are interpreting them that should
be part of our system of modal logic?
Let us start with the first question. Consider a formula such as DD p
or ◊ ◊ p. Presumably these are to be read as
It is necessarily true that p is necessarily true,
and
It is possible that p is possibly true.
We recall that the sense of necessity involved is logical, not causal
or epistemic. Now there does not seem to be any particular difficulty
in understanding these formulas. They may seem peculiar, cumber-
some, or even redundant, but they are not unintelligible. They may
say things we would never have occasion to say in ordinary discourse
or even philosophical discourse, but that is no drawback.
Formulas with more and more iterated modalities, however, are go-
ing to be more and more difficult to distinguish from each other by
any means besides counting the modalities. That is, it is hard to con-
ceive of what the difference in the number of modal signs between
□□□ Op and □□□□ Op is supposed to amount to. So while iter-
ated modalities may not be unintelligible, it is difficult to see what
the difference between certain ones is supposed to come to. Indeed,
one is led to wonder whether there is any difference.
Are there any reduction laws in T? We know that we can prove
□□p::) □p and ◊p::)◊◊p. If we could prove □p::) □□p and ◊◊p::)◊p,
this would give us a reduction law that would enable us to reduce
significantly many iterated modalities. But, alas, the latter formulas
cannot be proven in T. So we do not have a logical equivalence, and
48 PROPOSITIONAL MODAL Lome

therefore we lack a reduction law that would enable us to reduce con-

secutive D's or ◊'s. In fact, there are no reduction laws in T. All mo-
dalities in T are irreducible.
Is that as it should be? Clearly the one theorem we have, D p :::>
p,
is acceptable. But are there other theorems that we should have (i.e.,
express modal truths) and would give us reduction laws but that we
lack in T? Here we have before us a major issue distinguishing the
systems T, S4, and S5. We have just observed that if we had
(AS4) Dp:::>DDp,
we could then reduce any pair of D's to a single D. Adding (AS4)
as an axiom to T yields S4. AS4 is often called the "characteristic
formula" of S4. Its presence in addition to what we have would give
us the logical equivalence
Dp-DDp,
and it is a simple exercise to show that we also then have
◊p-◊◊p.

Is it true that if a proposition is necessarily true, then it is a neces-

sary truth that it is necessarily true? If the necessity involved is logi-
cal necessity, then it is hard to think of any reasons for denying AS4.
It seems very likely that any argument or example that establishes
the contingency (or impossibility) of Dp will also establish the con-
tingency (or impossibility) of p and thus fail to be a reason for reject-
ing the axiom AS4.
The explanation is quite clear in terms of the following "possible
worlds" interpretation. Like Leibniz, let us say that a logically neces-
sary truth is true in all possible worlds, and a logically possible propo-
sition is true in at least one possible world. Is it possible for D p to
be true while DD p is false, for some proposition p? This would mean
that there is a possible world in which Dp is true and D Dp is false.
If Dp is true in a world, then p is true in every possible world. But
if D Dp is false in a possible world, then Dp is false in at least one
possible world, and if there is a world in which Dp is false, then there
is a world in which p is false. But now our supposition has led to the
conclusion that there is a possible world in which p is true and p is
false, which is impossible.
It appears, then, that our system T fails to include an impor-
tant modal truth. We turn next to the system S4 which includes
Dp:::>DDp.
 THE SYSTEM 84 49

2.4 The System S4

The system S4 is one of eight modal systems devised by C. I. Lewis,

one of the great modem pioneers of modal logic. It was not the sys-
tem he favored, for reasons we will go into later. But it is something
of a favorite among logicians, primarily because it is well-behaved
and because it has some reduction laws.8 Our natural deduction treat-
ment of it is due to Frederick B. Fitch.

2.4.a Rules

We have said earlier that S4 can be obtained from T by a simple

modification of the rules. We have just now seen that this modifica-
tion has to allow us to prove (AS4) Op:::>□□p, since the possibility
of obtaining this formula distinguishes S4 from T. We can best see
what modification is necessary by asking what prevents us from de-
ducing AS4 in T. If we attempt a proof, the answer is soon obvious:
It is the requirement that we must remove the D when we reiterate
a formula of the form □ pinto a nee intro subproof. A good way to
obtain S4, then, is introduce an S4-reiteration rule which drops that
requirement.
Our rules of S4 will be all the rules of T plus the following S4-
reiteration rule:
S4-reit: Where a formula of the form D p occurs as a previous
line in a proof, it may be entered as a line within a nee
intro subproof with the justification 'S4-reit', unless
that occurrence lies within the scope of a discharged
assumption or within the scope of a terminated nee
intro subproof.
This rule is to be used in conjunction with necessity introduction sub-
proofs, just as the T-reit rule was. The T-reit rule is still available,
but it becomes redundant with the addition of S4-reit.
We illustrate the use of this rule by proving AS4:

l □p □p S4-reit
□p nee intro
□ p:::> □□ p.
The way we modified the rules of T to obtain our rules for S4 makes
it clear that Tis contained in S4, in the sense that any theorem prov-
50 PROPOSITIONAL MODAL Lome

able in Tis provable in S4. And our earlier illustration with AS4 makes
it clear that S4 is not contained in T.

2.4.b E:cercises

A. Show that T-reit is redundant in S4.

B. Prove:
1. ◊ □ ◊p::)◊p
2. (p-q)::)(r-(p-q))
3. □ p ::)(q- □ p)
4. (p::)Oq)::)(p::)(r- □ q))
5. □ ◊p::) □ ◊ □ ◊p
6. □ ◊ □ ◊p::) □ ◊p
7. ◊Op:) ◊ □ ◊ □p
8. ◊ □ ◊ □ p ::)◊ □ p

2.4.c Reduction of Modalities in S4

In S4, we have the following equivalences, which are reduction laws:

RLl □ p- □□ p
RL2 ◊p-◊◊p.
(RL2 is easily derived from RLl.)
From RLl and RL2 we see that any time a modal operator has two
consecutive occurrences in a formula, that formula can be replaced by
an equivalent formula containing a single occurrence of the operator
in that position. More interesting is the reduction of formulas con-
taining mixtures of operators. These laws enable us to replace any
formula containing iterated modal operators with an equivalent for-
mula that contains no more than three iterated modal operators.
There are only two possible ways of combining three modal opera-
tors so that each operator stands next to a different operator:
□ ◊ □p
◊ □ ◊p.
What formulas with fewer operators imply or are implied by these
formulas? You have already provided a proof for one in exercise set
B in section 2.4.b.
There are also two possible ways of combining four modal opera-
tors so that each operator stands next to a different operator:
□ ◊ □ ◊p
◊ □ ◊ □p.
 THE SYSTEM S5 51

Each of these can be reduced to a formula in which p is prefixed by

only two modal operators. Note that you have proven the appropri-
ate equivalence for each of these in exercise set B at the end of sec-
tion 2.4.b.
We see now that many iterated modalities can be reduced in S4.
Four irreducible iterated modalities remain. But S5, the system we
consider next, has no irreducible iterated modalities.

2.5 The System S5

S5, like S4, was first axiomatized by C. I. Lewis. S5 is a favorite

system from a purely formal standpoint because it is a well-behaved
system containing standard propositional logic and having no irre-
ducible iterated modalities. But its primary virtue is being the for-
mal system that best represents the notions of logical necessity, logi-
cal possibility, and logical entailment, where logical necessity is
thought of as "broadly logical necessity" - the concept discussed at
the beginning of the book. While the use of T and S4 would not com-
mit us to any mistaken inferences involving these concepts, these
systems do not provide us with the full range of inferences that are
intuitively correct. We have the full range only in S5.
The characteristic formula of S5 is
◊p:) □ ◊p.
In S5 modal status is always necessary, and the characteristic for-
mula of S5, together with the characteristic formula of S4 (which is
deducible in S5 ), can be seen as saying this. The characteristic for-
mula of S4, □ p :) □ □ p, says that if a proposition is necessary, its
necessity (i.e., its modal status as a necessary proposition) is neces-
sary. Likewise, the characteristic formula of S5 may be seen as the
counterpart claim for possible propositions. ◊ p :) □ ◊p says that if
p has the modal status of being a possible proposition, it is neces-
sary that it is a possible proposition.
S5 best reflects and expresses the possible worlds account of nec-
essary truth, the view that a necessary truth is one that is true in
all possible worlds. This interpretation confirms the plausibility of
the characteristic formulas of S4 and S5. Consider the characteristic
formula of S4 first. Suppose we have some proposition, p, that is true
in every possible world. Since each possible world contains every
proposition true in that world, each world contains not only p but
also the proposition that p is true in every world. If every world con-
tains the proposition that p is true in every world, it is true in every
52 PROPOSITIONAL MODAL Lome

world that p is true in every world, i.e., it is necessary that p is

necessary.
If we read the characteristic formula of S5 in this way we have:
if p is true in some world, then in every world p is true in some world.
If we are talking about broadly logical necessity it is very hard to
see how some proposition could be true in some world (i.e., possible),
and yet fail to be a logical possibility in some other world (i.e., not
necessarily possible). If it is possible that there are unicorns, then
there is some possible world in which it is true that there are uni-
corns. But then it is true in every world that it is true in some world
that there are unicorns. That is to say, it is necessary that it is pos-
sible that there are unicorns. So the characteristic formula of S5 ex-
presses something intuitively acceptable, something there seems to
be no good reason to deny.

2.5.a Rules for S5

The only difference between the rules of T and the rules for S4
comes in the restriction on the reiteration rule. The T-reiteration rule
allows only lines of the form Op to be reiterated into nee intro sub-
proofs and specifies that when these lines are reiterated, the [) must
be dropped. The S4-reiteration rule likewise allows only lines of the
form □ p to be reiterated, but it does not require that the □ be
dropped. We can obtain a rule for S5 by relaxing the restrictions still
more.
A look at the characteristic formula of S5 will show us how to re-
lax the restriction further so as to obtain the proper rule. The char-
acteristic formula of S5

implies that modal status is always necessary. If a proposition is nec-

essary, then it is necessary that it is necessary, and if a proposition
is possible, it is necessary that it is possible. The formula ◊p ::>□ ◊p
contains the clue to the modification needed in our rules to get S5.
Clearly what we must do is relax the restriction that only formulas
of the form □ p may be reiterated into nee intro subproofs so as to
permit reiteration of formulas of the form ◊p as well.
We obtain the system S5 by adopting an S5-reit rule which relaxes
this restriction. Our rule is stated as follows:
S5-reit: Where a modal formula (a formula of the form □ p
or of the form ◊p) occurs as a previous line in a
proof, it may be entered as a line within a nee
 THE SYSTEM 85 53

intro subproof with the justification 'S5-reit,'

unless that occurrence lies within the scope of a
discharged assumption or within the scope of a
terminated nee intro subproof.
The statement of the reiteration rule for S5 makes it clear that
T and S4 are contained in S5 and that S5 is not contained in either
of them. T is the weakest of our systems; S4 is stronger than T but
weaker than S5.

2.5. b Exercises

A. Prove the following formulas using the weakest reiteration rule

that still permits derivation of the formula.
1. □ (pv □ q)::>( □ pv □ q)
2. □ (p v q)::>(Dp v ◊q)
3. D(pvq)::> □(□ pv ◊q)
4. □ (p & Oq)::>(◊p & ◊q)
5. (Dpv □ q)::>D(pv □ q)
6. ◊(p & □ q)::>(◊p & □ q)
7. (◊p & ◊q)::> ◊(p & ◊q)
8. (◊p & □ q) ::>◊(p & □ q)
9. □ (p V ◊q)::>( □p V ◊q)
10. ( □ pv◊q)::> □ (pv◊q)
B. Prove that by adding ◊p ::>□ ◊ p as an axiom to our system T,
the characteristic formula of S4, □ p ::>□ □ p, follows.
Hints: 1) Any axiom is itself necessary, i.e., □ (◊p ::>□ ◊p);
2) any wff with the general form ◊ p ::>□ ◊ p is an axiom;
(Any formula that has that structure, where p is any
wff you like, is an axiom.)
3) Be sure to consider the axiom in both of its guises, i.e.,
both ◊ □ p ::>□ p and the above form.

2.5.c Reduction Laws

We saw earlier that we could reduce the number of consecutive

modal operators in many formulas, given the rules of S4. We did this
with the help of the reduction laws based on AS4:
RLl □ p- □□ p
RL2 ◊p-◊◊p.
Although we could significantly reduce the number of iterated op-
erators where long strings of them occurred, there were certain irre-
54 PROPOSITIONAL MODAL LOGIC

ducible combinations in S4. In S5 we can always reduce a sequence

of iterated operators to a single operator.
We can carry on such a reduction by continuing to reduce pairs
of iterated operators, if we have a way of reducing any possible pair.
Now there are only four possible pairs,
DD
◊ ◊
D ◊
◊ D.

A glance back at RLl and RL2 shows that we have appropriate re-
duction laws for the first two on the list. Furthermore, we have a
start on proving the required equivalences in the case of the latter
two. We have
D◊p:::>◊p
Dp:::>◊Dp

in T and S4. Now in S5, we are able to obtain equivalences and gain
the rernaining reduction laws. Indeed, the formula needed to prove
the necessary equivalence is the characteristic formula of S5. So we
may add to our stock of reduction laws in S5.
RL3 Dp-◊Dp
RL4 ◊p-D◊p
All the iterated combinations that were irreducible in S4 are now re-
ducible to a single modal operator in S5. In fact, whenever we have
iterated modal operators in S5, we may remove all the modal opera-
tors except the last one.
There is also another type of reduction that can be carried out in
S5. 'lb describe this, we introduce the concept of modal degree of a
formula. With iterated modalities, this concept works just as one
would expect: D ◊ D p is a formula t>fdegree 3; D ◊ D ◊ p :::> D ◊p is a
formula of degree 4. This concept also applies to formulas in which
a formula containing modalities itself occurs within the scope of mo-
dalities, even though no iteration occurs, e.g., D (D p :::,D p) has de-
gree 2. The concept of degree encompasses iterated modalities and
other situations as well. The degree of a formula can be precisely cal-
culated as follows:
1) a propositional variable has degree O;
2) if p has degree n, then ~p has degree n;
 PHILOSOPHICAL MATTERS 55

3) ifp has degree n andq has degree m, p vq, p &q, andp:::>q

have the degree of the larger component;
4) if p has degree n, then □ p and ◊p have degree n + I. 9
Not only can all iterated modalities in S5 be reduced to a single modal
operator, but every formula of a degree greater than one can be re-
duced to a first-degree formula in S5.

2.5.d Exercises

A. Prove:
I. □ ◊ □p= □ p
2. ◊ □ ◊p=◊p
3. □ ◊ □ ◊(p- □ ◊q):::>( □ ~pv ◊q)
4. ~(r-◊~p):::>(Dp & ◊r)
5. (~(r-◊~p)&(◊r-q)) :::>(( □ pv◊~q)&◊r)
B. Give the degree and reduce to first degree formulas:
1. □ ◊ □ ◊(◊◊p::) □ ◊p)
2. □ (pv □ q)
3. ◊(Op v ◊◊q)
4. ◊(p& □ q)
5. ◊( □ ◊p:::>q)

2.6 Philosophical Matters

2.6.a Differences between Systems

So far we have emphasized the formal aspects of the differences

between the Systems T, S4, and S5. They differ in size - that is, S5
has more theorems than S4, and S4 has more than T. T fits inside
S4, and S4 is contained in S5. S5 is therefore called the strongest
of these three systems, and T is the weakest of the three. These three
systems also differ in the reduction laws that are available in each
system. T has none, S4 has some, while S5 permits reduction of the
modality of any formula to a first degree modality.
But these are largely formal differences. They do not give a philo-
sophical "picture" of how the systems differ. Now we turn to a more
intuitive picture for thinking of the differences. The way of looking
at the differences presented here is due to Saul Kripke and has be-
come a popular way of dealing with this matter. It uses the idea of
possible worlds and of an accessibility relationship among them. The
56 PROPOSITIONAL MonAL Lome

different modal systems will represent differing degrees of accessi-

bility among possible worlds. But that is getting ahead of the story.
First we build our picture and introduce the necessary terminology.
A possible world is a maximally consistent state of affairs, a to-
tal way a universe could be. We imagine an array of such possible
worlds. We will call a proposition true in a possible world just in
case it would be true if that world were actual. It will be convenient
to think of possible worlds mainly in terms of the propositions true
in them.
Setting up our array of possible worlds, we further suppose that
we are given some relationship R among these worlds. Some may be
related by R while others are not. We will think of this relationship
as "accessibility" or "relative possibility." Depending upon how it be-
haves, we will have pictures or models of one system or another. We
call a proposition possible in a world just in case it is true in some
world accessible to the world in question. Similarly, a proposition is
necessary in a world just in case it is true in every world accessible
to that given world. Depending on how various worlds are accessible
to others in a given model, we might have a proposition necessary
in one possible world but not in another, or a proposition that is pos-
sible in one world but not in another. The theorems of modal logic
are generalizations about relationships of propositions within one
array of possible worlds. By altering properties of the accessibility
relation we can give models of T that do not make true all the theo-
rems of S4, and models of S4 that do not make true all the theorems
of S5, as well as models satisfying S5. The way this works depends
only on structural features of the model- upon properties of the ac-
cessibility relation among possible worlds- and not upon which in-
dividual propositions turn out to be true or false in a given possible
world. The theorems of T, for example, turn out to be true in all mod-
els having a certain structure, regardless of the truth or falsity of
individual propositions in those models. What kinds of structures
correspond to each of our systems?
We start with T. I ts characteristic formula is D p ::Jp. First we give
a model in which this formula is true, and then we give further pic-
tures in which this formula is true but the characteristic formulas
of S4 and S5 are not. The minimal network to model D p ::Jp requires
exactly one possible world accessible to itself in which D p is true.
Given some possible world accessible to itself with D p true in it, we
see that p holds also, because D p means that p is true in every ac-
cessible world. While it may seem trivial to state that the accessibil-
 PHILOSOPHICAL MATTERS 57

ity relation is reflexive here, that is precisely the feature of accessi-

bility guaranteeing that □ p :J p will be true.
But this picture is not very exciting as an attempt to model T,
because p :J □ p is also true in it. This means that our model so far
fails to distinguish T from propositional logic. Given both □ p :J p and
p :J □ p, we would have the reduction law, □ p =p, that would allow
us to eliminate all modalities, and thus our modal logic would col-
lapse back into classical propositional logic. We want a more "inter-
esting" model, one that not only distinguishes T from classical propo-
sitional logic but from S4 and S5 as well.
Let us suppose that we have a network of three possible worlds
accessible or inaccessible to each other as indicated in Figure 1.

p ~p
◊p ◊~p
□p ◊p
□ ◊p
I ► □ ◊p
I

p
◊p
◊~p
□ ◊~p
□ ◊p

Figure 1
- = is accessible to, where a world W1 is accessible to a world W2 just in
case any proposition true in W2 is possible in Wi.

Arrows represent accessibility and slashed arrows indicate lack of

accessibility. We assume that each world is accessible to itself, al-
though this is not indicated by arrows on our diagram. Inside each
58 PROPOSITIONAL MODAL LOGIC

circle is a list of some propositions true in the possible world repre-

sented by that circle. Let us see what happens to some of our char-
acteristic formulas in this model.
The characteristic formula of T, D p :::> p, holds here. Wherever D p
is true, so is p. This can be seen directly in W 1• 'l\vo other instances
of this characteristic formula, D ◊ ~ p :::>◊ ~ p and D ◊ p :::>◊ p, also
hold in W2 and W3 , respectively, and there are no counterexamples.
Indeed, given that any proposition that is necessary in a world is
true in every world accessible to that world, and that every world
is accessible to itself, it is easy to see that there cannot be any counter-
examples.
This model enables us to distinguish T from propositional logic,
something which the first model failed to do. For p holds in W2 while
□ p does not, since there is an accessible world, W3 , in which p is not
true. Thus we have a counterexample (actually a countermodel) to
the formula that would be needed to "collapse" modality, p :::>D p.
The characteristic formulas of S4 and S5 also fail in this model.
□ p occurs only in Wi,and DD p does not hold there because D p does
not appear in W2 , the only other world accessible to Wi. Thus
□ p :::>
DD p, the characteristic formula of S4, does not hold in this
model.
The characteristic formula of S5, ◊ p :::>D ◊ p, appears at first
glance to be satisfied, since ◊ p appears in each world and so does
D ◊p. However, in W3 we find that ◊ ~ p is true while D ◊ ~ p is not.
□ ◊~p is not found in W3 because ◊~p does not occur in Wi, the
other world accessible to W3 • Thus the formula ◊ ~ p :::>D ◊ ~ p is not
true in the model, and since it is an instance of ◊p :::>D ◊p, the char-
acteristic formula of S5, we have here a model which fails to satisfy
S5, even though initially it might have appeared to satisfy it.
The accessibility relation in Figure 1 is reflexive; that is, every
world is accessible to itself. But this relation is not symmetric there;
W1 is accessible to W2, but W2 is not accessible to W1• Nor is it tran-
sitive: Wi is accessible to W2, and W2 is accessible to W3 , but Wi is
not accessible to W3•
What would happen if the accessibility relation were transitive as
well as reflexive? As the reader may suspect, we will have S4 models.
Or, to put it differently, we will not be able to devise models in which
the accessibility relation is reflexive and transitive and in which the
characteristic formula of S4 fails to hold. To illustrate this, we will
use a model similar to the one in Figure 1, except that we alter the
accessibility relations somewhat. We want a model of T and S4 in
which the characteristic formula of S5 fails to hold.
 PHILOSOPHICAL MATTERS 59

p
◊p ~p
◊~p
□p
◊p
□□ p
□ ◊p I
I .. □ ◊p

p
◊p
□p
□□ p
□ ◊p

W2
Figure 2
[-=is accessible to]

First we check the characteristic formulas of T and S4. In each

world in which we have a formula of the form D p, we have a formula
of the form p. The same is true for the characteristic formula of S4.
We have formulas of the form DD p in Wi and W2 , and there we have
Op in each case. Now what of ◊p-::J □ ◊p? Again, although initially
it seems to be satisfied, it is seen to fail. Once again D ◊ p appears
in each world where ◊p appears, seeming to satisfy ◊ p -::J□ ◊ p. But
W3 contains ◊ ~ p without containing D ◊ ~ p, which means that not
all instances of ◊p -::J□ ◊p hold in this model. Where the accessibil-
ity relation is reflexive and transitive but not symmetric, the charac-
teristic formula of S5 is not guaranteed to hold.
If the accessibility relation is symmetrical as well as reflexive and
transitive, then it becomes an equivalence relation on the set of pos-
sible worlds, and every possible world is accessible to every other.
In fact, as Kripke remarks, S5 may be viewed as a modal system that
requires that all the possible worlds be accessible to each other.
This way of looking at the differences between the systems in terms
of the properties of the accessibility relation suggests the possibility
of another system "between" T and S5. Suppose that the accessibil-
60 PROPOSITIONAL MODAL LOGIC

ity relation is reflexive and symmetric, but not transitive. In such

a case we would have a system containing T but not 84, and con-
tained in 85, and whose characteristic formula is p ::::>D ◊p. In fact,
adding this formula as an axiom to the axioms of 84 yields 85 -
exactly what we would expect given the relationships we have just
considered. This formula is sometimes called the Brouwersche axiom,
although the connection with Brouwer is tenuous. 10
Comparing these various modal systems in terms of different prop-
erties of the accessibility relationship offers a way of seeing why 85
is the modal system that best represents the understanding of broadly
logical necessity. That which is broadly logically necessary is true
in all possible worlds, true no matter what. When we use this picture
of possible worlds and accessibility relationships, the necessity of a
proposition in a given world- in the actual world, for example- is the
truth of that proposition in all possible worlds accessible to the ac-
tual world. But this would not coincide with truth in all possible
worlds unless the accessibility relation were reflexive, symmetric, and
transitive. So if we think of broadly logically necessary truths as true
in all possible worlds, 85 seems to be the most adequate modal system.
So far, all we have done is argue that modal inferences properly
follow certain patterns. We have not tried to say which propositions
are broadly logically necessary, nor have we tried to devise any test
for picking them out. Many controversial candidates for necessary
propositions have been proposed by philosophers, but to accept 85
(or any other modal system) does not commit us to accept any phi-
losopher's candidates for broadly logically necessary truths, beyond
the theorems of 85 (or of the other modal system accepted).

2.6.b Some Assumptions Implicit in Our Presentation

When we accept 85 as the most adequate and accurate account

of the logic of broadly logical possibility and necessity, we are mak-
ing a number of assumptions about propositions and modality. We
have seen in our discussion of accessibility that adopting 85 rules
is presuming that broadly logical necessity is a kind of unlimited
necessity- for something to be necessary in this sense is for it to be
true in every possible world, every possible state of affairs. We do
not admit as a possibility that some proposition should be necessary
in one possible world and not be necessary in another. We assume
that whatever is broadly logically necessary in the actual world is
necessary in every possible world. But, as we have just observed, this
does not imply anything about our ability to know whether or not
 PHILOSOPHICAL MATTERS 61

a given proposition is necessary, nor does it imply that we can pro-

vide a completely reliable test by which we can discover whether a
proposition is necessary.
In order to reveal some further assumptions we have been malc-
ing, let us consider a criticism of all the systems we have introduced
(and more besides) made by A. N. Prior. This criticism applies to all
systems that contain as a theorem the formula
◊p-:J◊(pvq)
i.e., if it is possible that p, then it is possible that p or q. Prior tries
to show that this is false by giving counterexamples. You might won-
der how Prior could possibly given an example in which it is true that
p is possible and false that p or q is possible. But this is not exactly
what he does .
. . . let p be 'Only God exists,' and suppose this possible, and let q be
'I don't exist.' Here p is possible ex hypothesi, but could the disjunc-
tion 'Either p or q,' i.e.,
'Either only God exists or I don't exist,'
possibly be true? The peculiarity of it is that the disjunction is unstat-
able unless I exist, and is therefore only statable if both parts of it,
'I don't exist' and 'Only God exists,' are false. In this case, then, 'Pos-
sibly either p or q' is false although 'Possibly p' is true. 11
Prior's contention, notice, is that the proposition in question is stat-
able (by anyone other than God) only if it is false, so it is not possible
for it both to be stated and true. Well, you might say, what does it
matter whether or not anyone can state the proposition truly? After
all there is such a proposition, and if there is, then it is possibly true
even if I cannot state it. But this reply assumes just what Prior wishes
to reject-that there is such a proposition whether statable or not.
Perhaps he is harking back to the medieval idea of regarding a propo-
sition as a dictum, a thing said. Thus he assumes that where some-
thing is unstatable, we do not have a dictum or proposition, and there-
fore since propositions are the bearers of truth or falsity, we do not
have truth or falsity. It is worth noting here that Prior holds this
not only for cases of pronomial reference, such as "I don't exist," but
also for dicta using names which fail to refer - that is, names of non-
existent objects. Thus "Santa Claus doesn't exist" likewise expresses
a dictum that is either false or unstatable, and similarly lacks a truth
value where it is unstatable. This dictum cannot actually be asserted
in such a way that the referring expression actually refers and the
assertion is true.
62 PROPOSITIONAL MODAL Lome

Looking back at the inference Prior is questioning, we see he holds

that it is possible for the antecedent to be true and possible for the
consequent to fail to be true. The consequent fails to be true, not
because it is false, but because it is not statable and hence lacks
truth value.
Prior's alleged counterexample exposes an important assumption.
Accepting the counterexample entails accepting the claim that at
least some propositions are contingent beings - i.e., they can fail to
exist. We have tacitly assumed that all propositions are necessary
beings - that is to say, they all exist and have truth values in all pos-
sible worlds, although of course not all are necessary truths. Now
we can see the importance of this assumption to the adequacy of the
modal systems we have been considering.
Although we will not defend our assumption here, we may explore
a bit further some consequences of denying it. The most important
consequence is that we must revise our accounts of possibility and
necessity, since there are additional cases to be dealt with.
We described necessary propositions as those true in all possible
worlds, and we can continue to use that characterization. But, if we
continue to hold that necessity and possibility are interdefinable while
we allow the possibility that propositions fail to exist in some worlds,
the definition of possibility must be changed. Let us call those
propositions true in all possible worlds strongly necessary. Then the
corresponding sense of possibility becomes weak possibility- the
not-strongly-necessarily false. This means that any propositions that
are not false in every possible world are weakly possible. These in-
clude not only the strongly necessary, and those true in some worlds
but false in others, but also those propositions that are true in every
world in which they exist but do not exist in all worlds, as well as
those false in every world in which they exist but that do not exist
in all worlds. Equivalently, we could say that a proposition is weakly
possible just in case there is some world in which the denial of that
proposition fails to be true.
But now there is a second way to understand these modal notions
of possibility and necessity. Essentially it involves changing where
we put the cases in which the proposition fails to exist in some worlds.
Let us call a proposition weakly necessary if and only if there is no
world in which that proposition exists and is false. This, of course,
counts a proposition as necessary even though it is not true in every
possible world. Correlatively, a proposition is strongly possible if and
only if there is some possible world in which it is true. (If there are
no propositions that exist in some worlds and not in others, then
 PHILOSOPHICAL MATTERS 63

these two accounts of necessity and their corresponding senses of

possibility are simply equivalent.)
This gives us two distinct pairs of modal notions:
strong necessity-weak possibility
weak necessity- strong possibility.
A proposition would be weakly necessary but not strongly necessary
if there were some worlds in which it failed to exist while it was true
in every world in which it did exist. Prior apparently thinks the propo-
sitions we express when each of us says, "I exist," are of this sort.
Likewise, if a proposition that fails to exist in some worlds is false
in every world in which it exists, then it is weakly necessarily false.
Presumably "I do not exist" expresses a proposition of this sort for
Prior.
Having distinguished these pairs of modal concepts, we can see
that Prior's alleged counterexample works only if possibility is taken
as strong possibility. That is, the modal operators in the formula
◊p-::J ◊(p v q) must be interpreted in the sense we have called strong
possibility in order for the theorem to turn out false. The formula
continues to be true for weak possibility. So for Prior's counterexam-
ple to work, it must be possible for some propositions to fail to exist,
and possibility must be defined as strong possibility.
But there is a strange result if we follow Prior here. In the pas-
sages where this counterexample is presented, 12 Prior suggests that
in worlds where I don't exist, the propositions I now express by say-
ing "I exist" and "I don't exist" also fail to exist. If that were the case,
then the proposition I express by assertively uttering the words "l
exist" is a (weakly) necessary truth. This seems to be an unhappily
implausible consequence if here we are trying to capture the notion
of "broadly logical necessity" described earlier.
On the other hand, defining possibility as weak possibility and
adopting Prior's position on "I exist" yields the conclusion that con-
tradictions are (weakly) possible. Since I exist now, the propositions
I express by saying "I exist" and "It's false that I exist" both exist.
Hence their conjunction, a proposition normally taken to be neces-
sarily false or impossible, exists. But since I presumably fail to exist
in some other worlds, this conjunction does not exist in some other
possible worlds. However, it then follows by the definition of weakly
possible that this conjunctive proposition is possible. Hence a propo-
sition that exists and is usually taken to be impossible is (weakly)
possible, given Prior' s assumptions.
There are probably many lines to pursue on this matter. Although
64 PROPOSITIONAL MODAL Lome

we will not pursue them here, we will be forced to look at similar is-
sues later on in dealing with modality and quantification when we
deal with the existence of individuals. Now we tum to another ques-
tion raised not only by these conundrums but also by the fact that
there are so many systems of modal logic: Is there one correct sys-
tem of modal logic?

2.6.c Is There Only One Correct System of Modal Logic?

We have before us three systems of modal logic, and there are many
more systems which we have not considered. Are they all correct?
Is only one correct?
In an Aristotelian Society symposium addressed to these ques-
tions, E. J. Lemmon 13 suggests an analogy with a situation in mathe-
matics. There are many systems of geometry: Euclidean geometry,
and a host of non-Euclidean geometries. Now there is one weak sense
in which all of these geometries are correct: They are all consistent
formal systems. They can each be modelled so that all the theorems
of the system tum out to be true in the model. But which one is true
of actual physical space? Which is true of the lines and triangles and
circles, etc., with which we have to do in our daily lives? It seems that
only one system can be correct and that it is an empirical matter to
find out which one.
Alas, the question is not quite so straightforward as this. For to
answer the question in this way, we must make some assumptions
about the accuracy of our measurements - usually that our instru-
ments remain constant at different times and places. But, continues
Lemmon,
If we make this assumption, we shall find, according to Einstein that
in fact there are triangles whose angles do not add up to two right angles:
on this assumption, physical space does not conform to Euclidean ge-
ometry. On the other hand, we can preserv!) the Euclidean model for
physical space by sacrificing the assumption, by admitting instead that
one's measuring instruments are deformed in certain ways at different
points. 14

Thus, which geometry turns out to be correct depends on other theo-

retical considerations. Any answer to the question requires a full
description of the situation, and even if we have as full a descrip-
tion as we can obtain, the facts at our disposal may not dictate the
answer.
The analogy contains several morals applicable to deciding which
 PHILOSOPHICAL MATTERS 65

modal logic is correct. First, it is not sufficient to ask about the for-
mal systems alone or even the formal system interpreted with the
help of an abstract model, such as a set theoretical model. We must
ask about the system under its intended interpretation, some actual
application. Then we can sensibly ask whether one system rather
than another is correct. Secondly, our intended interpretation must
be clear enough to enable us to answer questions about the truth or
falsity of theorems once they are interpreted. And, finally, we must
consider the effect adopting the system will have on other theories
we hold (or want to hold).
There is another consideration important to our present case. We
have before us three systems that are all compatible. The larger ones
contain the smaller ones. So if S5 is correct, both S4 and T contain
no falsehoods; rather, they do not contain all the modal truths. In
such a situation T and S4 would not be incorrect but insufficient.
Yet the correct system should be as adequate as possible; it should
contain all the relevant truths.
Although only one system is most adequate under an intended
interpretation, and we have suggested that S5 is the preferred sys-
tem where necessity is understood as broadly logical necessity, it does
not mean that there is no interest in the other modal systems, in-
cluding ones not presented here. For there are other ways of under-
standing necessity, as we have just seen in our brief consideration
of Prior, and there are still other quite different interpretations of
modal operators that philosophers have considered. Relative to these
other interpretations, one of the other modal systems may be the cor-
rect one. Let us look at some of the alternative interpretations to get
an idea of the possibilities.
We have mentioned earlier that C. I. Lewis, the "inventor" of S4
and S5, rejects these systems in favor of a weaker one, S2. Lewis be-
gan exploring modal logic because of his dissatisfaction with ':J' as
the formalization of the concept of implication, as used in Principia
Mathematica and in nearly all symbolic logic textbooks today. He
thinks that one proposition implies another when it is impossible both
for the implying proposition to be true and for the implied proposi-
tion to be false. He further thinks of implication in terms of deduci-
bility. Lewis uses a fishhook as his symbol for implication and he
interprets the expression p -3 q as meaning that q is deducible from
p. However he defines p -3 q in the same way we have defined p -q
-as O(p:Jq).
He employs his interpretation when he rejects S4 and S5 for con-
taining (p-3q)-3((q-3r)-3(p-3r)) as a theorem:
66 PROPOSITIONAL MODAL LOGIC

I doubt whether this proposition should be regarded as a valid prin-

ciple of deduction: it would never lead to any inference p -S r which
would be questionable when p -S q and q -S r are given premises; but
it gives the inference (q -S r)-S (p -Sr) whenever p -Sq is a premise.
Except as an elliptical statement for "((p -Sq) & (q-S r))-S (p-S r) and
p -S q is true," this inference seems dubious. 15

Lewis is here claiming that this formula is at least dubious and per-
haps false when interpreted as being about deducibility. He does not
doubt that when q is deducible from p and r is also deducible from
q, it can be deduced that r is deducible from p. But he is reluctant
to believe that when q can be deduced from p, we can always go on
to deduce that r's deducibility from pis deducible from r's deduci-
bility from q.
In S4 and S5
□ [ □ (p=>q)=> □(□ (q □ (p:::>r))]
:::>r):::>
is provable as a theorem, and it seems acceptable under our inter-
pretation. This suggests that Lewis either rejects our interpretation
of □ and ◊, or that he rejects other principles that are needed for
the deduction of this formula. I think it is most plausible to suppose
that he is reading □ as something other than broadly logical neces-
sity, since he formally defines p -8 q the very same way we define
p -q. He obviously takes the concept of necessity to be more closely
connected with deducibility than we do.
Arthur Prior has studied tense-logical interpretations of several
modal systems. 16 One way of obtaining a tense-logical interpretations
of the modal systems we have considered is to read the operators as
follows:
◊p = itis or was or will be the case that p
□p = it is, was, and will be the
case that p (or it is always
the case that p)
p = it is (now) the case the p.
Obviously some of the basic axioms are true on this interpretation.
□p =>p = If it is, was, and will be the case that p, then it
is (now) the case that p;
□ (p =>q) =>(□ p =>□ q) = If it is always the case the p =>q,
then if it is always the case that p, then it is always the
case that q.
So is R - if p is a theorem of propositional calculus, then p is, was, and
will be the case. Other tense-logical interpretations are also possible.
 PHILOSOPHICAL MATTERS 67

Jaakko Hintikka has developed an epistemic interpretation using

the structure of S4. 17 Here the operators are read
□p = Smith (or some specific individual) knows that p
(written K p); 8

◊p = It is possible for all Smith knows that p (written

Psp);
while p = it is true that p.
Obviously it is important to keep careful track of the subscripts on
the "modal" operators, since most of the inferential relationships will
hold only within one given individual's knowledge and belief, and not
between the beliefs of different individuals. For example, Smith's
knowing p will imply that he, Smith, believes p, but Smith's know-
ing p will imply nothing about Jones's beliefs.
a) □ p = ~ ◊ ~p is interpreted as
K 8 p = ~Ps ~p, which is read as "Smith knows p iff it is
not possible, for all S knows, that ~p";
b) □ p::)p is interpreted as
K 8 p::) p, which is read as "If Smith knows p, then p is
true";
c) □ (p:)q):)( □p:) □ q) is interpreted as
K 8 (p::)q)::)(K 8 p:)K 5 q), which is read as "If Smith knows
that p implies q, then if Smith knows that p, then Smith
knows that q";
d) □ p ::) □ □ p is interpreted as
K 8 p::)K 8 K 8 p, which is read as "If Smith knows p, then
Smith knows that he knows p."
The wary reader may be suspicious of some of these, but we will not
examine them further here. Suffice it to say that Hintikka is aware
of the possible grounds for suspicion and he defends his position on
these points. Needless to add, he develops this interpretation with
more skill and subtlety than is exhibited here.
We could continue examining interpretations of modal systems, 18
but the point is clear. Many interpretations can be given for the vari-
ous systems. When we ask whether one system is correct or which
system is correct, we must always ask those questions relative to some
interpretation. Different interpretations are required for different pur-
poses, and different systems are likely to be preferred for different
interpretations.
 3
QUANTIFICATION

Before we turn to quantified modal logic, we must specify which

symbols and rules we will use for predicate logic. The system of predi-
cate logic- quantifier logic, quantification theory, or first-order logic
- used here is the same as that presented in nearly any standard in-
troductory textbook. Despite differences in symbols and in the ways
rules are specified, exactly the same formulas turn out to be theo-
rems and the same argument forms turn out to be valid. Here are
the symbols and rules used in this text.

3.1 Symbols

Quantification theory adds two operations to those employed and

symbolized in propositional logic.
OTHER COMMONLY
OPERATION SYMBOL EXAMPLE USED SYMBOLS

Universal Quantification (x) (x)Ax ('v'x), Vx, nx

Existential Quantification (3x) (3x)Ax (Ex),Ax, :Ex

Lower case letters from the end of the alphabet are used as variables,
and Roman capitals with a variable or variables written immediately
to the right represent what are variously called predicates, senten-
tial functions, open sentences, propositional functions and relations.
For example, the expression Ax might represent the predicate x is
an aardvark, and Axy might symbolize x is the aunt of y.
We also adopt the usual characterization of free and bound occur-
rences of variables. The occurrence of a given variable will be bound
in a given formula when it either occurs in a quantifier or within the
scope of a quantifier containing that variable. An occurrence of a vari-
able that is not bound is a free occurrence.

69
70 QUANTIFICATION

3.2 Interpretation

The universal quantifier is designed to capture certain ordinary

uses of "all" and "every," and the existential quantifier represents that
use of "some" which is more precisely expressed as "at least one."
Where Gx =xis a Greek andHx =xis human, (x)(Gx::JHx) symbol-
izes the statement that all Greeks are humans and (3x)(Gx &Hx)
symbolizes the statement that some Greeks are human (or, more
precisely, that at least one Greek is human).
We will understand the quantifiers to range over existing objects.
It may seem strange that we bother to say that, but there are two
matters at stake here.
First, there are two ways to read quantifiers- the "objectual" way
just mentioned, and a substitutional way. The substitutional way
construes formulas of the form (3x) (Ux) as follows: There is a true
substitution instance of the formula (open sentence) ax.This has
the consequence that uses of existential quantifiers carry with them
no implications about what objects there are. The objectual view of
quantification, on the other hand, reads such formulas so as to im-
ply that there exists an object of which the predicate is true when
we have a true existential statement. The difference between these
views can be clarified with an example. It is true that Fafner (or
Santa Claus) does not exist. According to the substitutionalist such
a truth licenses the existential generalization, (3x) (x does not exist).
Or similarly, from Bucephalus no longer exists, we may infer (3x) (x
no longer exists). The objectualist demurs in these and similar cases.
By objectualist lights, (3x) (x does not exist) expresses something
self-contradictory and so any inference of this from a truth must be
invalid. 19
The second matter at stake here is related to the first and will come
up again later. I tis the question of what we quantify over-the range
of the quantifiers. We can restrict the range of our quantifiers. For
example, in an arithmetic textbook, we might understand our quan-
tifiers to be ranging only over integers, or over rational numbers, or
over real numbers. In such a case the formula (3x) (xis prime or x
is over 7 feet tall) is understood to mean that there is a number that
is either prime or over seven feet tall. It is clear that in such a con-
text the open sentence, x is prime or x is over 7 feet tall, is satisfied
by the number three and not by Ralph Sampson. It is possible, of
course, to expand our universe of discourse beyond numbers. We may
add sets, persons, material objects - anything that we would call a
 EQUIVALENCES 71

concrete object or an abstract object. Is it possible to expand to our

universe of discourse beyond the abstract and concrete objects which
actually exist? Well, there have been proposals that we quantify over
fictional entities or over possible but nonactual objects. Here, how-
ever, we will be conservative and limit ourselves to existing things.
This proposal is not intended to settle any philosophical arguments
about what does and does not exist; we will consider that to be be-
yond the scope of our concern here.

3.3 Equivalences

Like the modal operators, the quantifiers are interdefinable. We

will officially consider the universal quantifier to be undefined and
the existential quantifier to be defined in terms of it. (We could, if
we chose, do it the other way around.) Here is a list of equivalences,
which may be used in proofs, expressing the relationships between
the quantifiers.
(3x)ax = ~(x)~ax
(x) ax = ~ (3x) ~ ax
~(3x)ax =(x)~ax
~ (x) ax = (3x) ~ ax
We will label our use of these rules QE (for Quantifier Equivalence).

3.4 Rules

In quantification theory we have an expanded means of symboliz-

ing and two additional operations (just one if we count only primi-
tives). 'Ib continue our natural deduction treatment, we require intro-
duction and elimination rules for the universal quantifier and for the
existential quantifier.
Statements of rules vary from text to text. Certain restrictions are
necessary, but it is possible to introduce them in different ways, and
most textbooks employ some orthographic device to insure compli-
ance with the restrictions. The intuitive idea behind each rule usu-
ally is simpler than the technical statement of the rule.
Let us use a special symbol to facilitate the statement of our rules.
ax will stand for a formula possibly containing free occurrences 20
of the variable x. (x) ax will be the universal quantification of ax.
72 QUANTIFICATION

Universal Instantiation (UI)

From (x) ax
infer ay (or ax) where y is free and replaces all free oc-
currences of x in ax.

Universal Generalization (VG)

From ay (1) where ax is obtained from ay by re-

infer (x) ax placing all free occurrences of y by x;
(2) where no free occurrences of x in ay are
bound in (x)ax (This restriction is to
prevent us from accidentally binding a
variable that should remain free.);
(3) where y does not occur free in an as-
sumption within whose scope Uy (the
formula being generalized) lies. (This in-
cludes EI assumptions; see below.)
This set of restrictions is a technical (syntactic) way of preventing
certain fallacious inferences. Just remember these rules of thumb: Do
not apply UG to a variable introduced by EI; and do not apply UG
to a variable introduced in an assumption until you have discharged
the assumption.

Existential Generalization (EG)

From ax where we need not replace all occur-

infer (3y)ay rences of x by y.

Existential Instantiation (El)

From (3x)ax
use
IT where Uy is the result of replacing all
EI ~p ::: free occurrences of x in ax with y;
where y is a variable that does not oc-
cur in any line of the proof prior to its
to infer p EI concl
introduction in the EI subproof, and y
does not occur free in p.
This rule appears more complicated than the others. However, this
form of the rule with its restrictions fits some common sense insights
 EQUIVALENCES 73

about the reasoning that involves Existential Instantiation. The first

is that if it is true that something has property a, then it is permis-
sible to assume that some arbitrarily selected individual in the uni-
verse of discourse has a and draw conclusions from that assumption,
provided that we use no special features of the individual chosen.
Secondly, our assumption is just that-an assumption. We do not
infer that this arbitrarily selected individual has a, and the conclu-
sion we draw in closing the EI proof must not involve that "instance."
Our form of the EI rule explicitly represents these insights.
The "EI asp" line captures the first insight. We assume that some
individual has a. The restriction on the choice of the variable insures
that the choice is arbitrary and prevents the use of any feature pecu-
liar to the individual chosen.
The second insight is captured by the use of a subproof that per-
mits deductions from our chosen "instance," but that does not permit
any conclusion containing that "instance" to be entered as a line out-
side the subproof. Thus the rule permits us to assume an instance of
an existential generalization while it forbids us to bring conclusions
about that particular instance outside the scope of the assumption.
The statement given of the EI rule can be regarded as an abbrevia-
tion of a use of implication introduction and UG together with a fa-
miliar theorem about the scopes of quantifiers. Consider the follow-
ing proof schema:
(3x)Ux
(considered this time as

17UyJp
an assumption for impl
intro)
Impl intro
(x)(UxJp) UG
(x) (UxJp)J[(3x)UxJp] (Theorem)
(3x)UxJp MP
p MP
The restrictions on the EI rule are just those needed to guarantee
that this proof will always work. If the subproof were not carried out
in terms of a "new" variable, then it would not be guaranteed that
UG could be used. And if p contained any free occurrences of the
variable introduced in the subproof, then the theorem about quan-
tifiers could not be used.
This shows that the form of EI adopted here is both formally and
intuitively a good one for a natural deduction system.
74 QUANTIFICATION

For those who need to become accustomed to these rules, we pre-

sent some sample proofs and offer a few exercises. An important
"trick" to keep in mind when using the EI rule is to deduce within
the EI subproof the conclusion you are working for and then bring
it outside of the subproof.

Prove (x) (Ax -:JBx)-:J((3x) ~Bx -:J(3x) ~Ax)

(xl{Ax-:JBx) Asp

l
(3x)~Bx) Asp
~Bx EI asp
Ax-:JBx UI
~Ax MT
(3x)~Ax EG
(3x)~Ax EI concl
(3x) ~Bx :J(3x) ~Ax impl intro
(x) (Ax:JBx):J((3x) ~Bx:J(3x) ~Ax) impl intro

Prove (x) (Fx -:JQ) -:J((3x)Fx :J Q)

(Q here represents any formula containing no free occurrences of x.)
(x)(Fx:JQ) Asp
x) Fx Asp

~
EI asp
IQX:)
Q UI
MT
Q EI concl
(3x)Fx-:JQ impl intro
(x) (Fx :J Q) :J ((3x)Fx :J Q) impl intro

Prove ((3x)Fx:JQ):J(x)(Fx:JQ)
j(3x)Fx :J Q) Asp

~
Asp
,~~Fx EG
1

MP
Fx:JQ impl intro
(x)(Fx:JQ) UG
((3x)Fx :J Q) :J (x)(Fx :J Q) impl intro
 EQUIVALENCES 75

Exercises

1. (p ::>(x)Qx) ::>(x)(p ::>Qx)

2. (3x) (p ::>Qx) == (p ::>(3x)Qx)
3. (x)(Ax::>Bx)::>[(x)( ~Ax:> ~Cx)::>(x)( ~Bx:> ~Cx)]

Since the UI rule permits the deduction of formulas containing

free variables, it is necessary to add a final word about the theorems
of quantifier logic. The theorems are any closed formulas deducible
by means of the rules given here, including the earlier rules for propo-
sitional logic. A closed formula is one that contains no free occur-
rences of any variable. Of course it is a simple matter to close a for-
mula we deduce containing free occurrences of some variables: we
apply UG, or if we like, EG.
 4
QUANTIFIED MODAL LOGIC

This chapter presents quantified modal logic within the framework

of propositional modal logic and quantification theory already given.
After a brief look at the need for quantified modal logic, we examine
some traditional modal distinctions that have an important bearing
on our development of quantified modal logic. The most important
is the distinction between modality de re and modality de dicto. We
use this distinction to explain what our modal signs, □ and ◊, mean
when applied to predicates, and we discuss translation of modal claims
from natural language into our symbolic language. After a brief look
at some objections to quantified modal logic, we discuss and sum-
marize the philosophical considerations we are trying to satisfy in
the system presented here.
With these philosophical considerations in hand we consider three
versions of quantified modal logic that have a legitimate claim to the
title "Quantified S5." The first and easiest to develop is a version
studied by Saul Kripke in 1959. Both it and a subsequent modifica-
tion by Kripke in 1963 have unhappy philosophical consequences,
however. A third version, called Actualism, is expounded and defended
as the best system of quantified modal logic relative to a possible
worlds interpretation. We present a natural deduction version of ac-
tualism equivalent to an axiomatization by Thomas L. Jager. This
system is discussed and used by Alvin Plantinga in The Nature of
Necessity and other writings.
The last three sections of the chapter show how the system of
quantified modal logic developed here provides superior resolution
of certain problems associated with modal logic.

4.1 Motivation
Why extend modal logic by adding quantification? What benefits
do we expect? Are there inferences we can shed light on that are not
illuminated by the methods developed so far?

77
78 QUANTIFIED MODAL Lome

We develop our answer with the help of an example.

1. All men are necessarily rational.
2. Socrates is a man.
3. : . Socrates is necessarily rational.
4. :. Socrates is (actually) rational.
5. :. Socrates is possibly rational.
The necessity has to be represented formally in (3)in order to account
for the inference of (4) and (5). Yet we cannot say that it is a neces-
sary truth that Socrates is rational (since it is not a necessary truth
because Socrates could have failed to exist). On the other hand (3)
follows from (1) and (2), which are both true, so it is true; and it im-
plies (4) and (5).
This example suggests two conclusions. First of all, the necessity
here indicated figures in the inference. We cannot simply say that
this argument uses a predicate, "necessarily rational," that needs no
special representation. While this suggestion might help us explain
how (3) follows from (1) and (2), we cannot account logically for the
inference of (4) and (5). The other conclusion suggested by the ex-
ample is that the sense of necessity here, while perhaps related to
broadly logical necessity, is not the same.
Some explanation of necessity, in this present sense, needs to be
given and it should be related to the sense introduced for proposi-
tional modal logic. This will be done here by considering a pair of
old distinctions common to medieval discussions of modality. Then
it will be easier to pick out the appropriate sense of necessity and
relate it to broadly logical necessity.

4.2 Two Distinctions

The venerable distinctions are (a) the de re/de dicto distinction

and (b) the distinction between the composite sense and the divided
sense of modal terms. These distinctions will be considered only as
they apply to our modal terms, possibility and necessity. It will be
clear that each distinction has a far wider range of applicability.
Medieval discussions of modal arguments frequently appeal to a
distinction between modality de re and modality de dicto. Modality
de re is modality thought of as applying to a thing (res), more pre-
cisely, as a way a thing possesses a property. For example, one thing
might be said to possess a property necessarily, or something can
be said to possibly have some given property, as in the claims that
Socrates is necessarily rational or Socrates is possibly a sailor.
 Two DISTINCTIONS 79

Modality de dicto is the modality applied to a statement (dictum).

It refers to the manner or mode of a statement's being true. For ex-
ample, it is necessarily true that all bachelors are unmarried. Here
it is the statement (the dictum) that is said to be necessary. More
exactly, it is the statement's being true that is necessary.
Often an ordinary sentence, such as "Socrates is rational neces-
sarily," is ambiguous. The word "necessarily" might be taken to refer
to Socrates's possession of rationality, as in
Socrates is necessarily rational,
or to the necessity of the statement, as in
It is necessary that Socrates is rational.
Since the former is true and the latter is false, it is obvious that there
is some point to keeping them distinct, and that the result of con-
fusing these two senses in an argument can be disastrous.
The de dicto sense of necessity can be readily identified with
broadly logical necessity, as this was described earlier. That is not
to say, however, that broadly logical necessity is what the Medievals
who discussed de dicto modality always had in mind. Further con-
sideration of de dicto modality will be postponed until after we look
at a related matter.
I turn first to another modal distinction, one which is occasion-
ally linked with the de re/de dicto distinction and sometimes con-
fused with it. This is the distinction between a modality taken in the
composite sense (in sensu composito) and one taken in the divided
sense (in sensu diviso). The locus classicus for explanations of this
distinction is a passage in Aristotle's De Sophisticis Elenchis (166a):
For the meaning is not the same if one divides the words and if one
combines them in saying that 'it is possible to walk-while-sitting' [and
write while not writing]. The same applies to the latter phrase, too, if
one combines the words 'to write-while-not-writing': for then it means
that he has the power to write and not to write at once; whereas if one
does not combine them, it means that when he is not writing he has
the power to write. 21

Resorting to our symbolism, we can represent Aristotle as distin-

guishing
◊(3x) (xis a man & xis not writing & xis writing)
(taking the phrase in composition or compositely) from
(3x) (xis a man & xis not writing & ◊(xis writing))
80 QUANTIFIED MODAL LOGIC

(taking the phrase in division or dividedly and assuming prior to the

explanation that we can make some sense of this use of the ◊ ). The
importance of making this distinction is clear when we observe that
the latter statement is true while the former is necessarily false.
It should be observed here that this distinction is not the same
as the de re/de dicto distinction, but is simply a distinction between
two ways of representing the scope of a modality. It may appear that
because the first symbolization of Aristotle's example has the mo-
dality applying to a dictum and the second to a thing's possession
of a property, the distinctions come to the same thing. But we can
illustrate the distinction between the composite and the divided sense,
eliminate any de dicto reading, and probably more accurately repre-
sent Aristotle's meaning as follows:
(3x) (xis a man & ◊(xis writing & xis not writing))
(composite sense)
as compared with
(3x) (xis a man & xis writing & ◊(xis not writing))
(divided sense).
Similar illustrations can be provided where the modality is de dicto
throughout. For example,
It is possible for it not to be raining when it is raining
can be read
compositely as ◊(~R &R)
and dividedly as (◊~R)&R.
Thus it is clear that the distinction between taking a modality in the
composite sense and the divided sense is different from the de dicto/
de re distinction, although they may coincide in many cases.
These distinctions are linked with the previously discussed dis-
tinction between the necessity of the consequent and the necessity
of the consequence by St. Thomas Aquinas. In a well-known passage
in his Summa Contra Gentiles, he invokes these distinctions to re-
solve the question of whether God's foreknowledge of a contingent
action (or truth) necessitates that action (or truth):
If each thing is known by God as seen by Him in the present, what
is known by God will then have to be. Thus, it is necessary that Soc-
rates be seated from the fact that he is seen seated. But this is not
absolutely necessary or, as some say, with the necessity of the conse-
quent; it is necessary conditionally, or with the necessity of the con-
 Two DISTINCTIONS 81

sequence. For this is a necessary conditional proposition: if he is seen

sitting, he is sitting. Hence, although the conditional proposition may
be changed to a categorical one, to read what is seen sitting must nec-
essarily be sitting, it is clear that the proposition is true if understood
of what is said de dicta, and compositely; but it is false if understood
of what is meant de re, and dividedly. Thus in these and all similar
arguments used by tl\ose who oppose God's knowledge of contingents,
the fallacy of composition and division takes place. 22
Thomas's resolution of the problem here seems to be correct, but the
apparent conflating of distinctions is mistaken.
Thomas here points out that the claim
If God sees that Socrates is seated, then necessarily, Soc-
rates is seated
is ambiguous. It could mean
Necessarily, if God sees that Socrates is seated, then Soc-
rates is seated,
which expresses the necessity of the consequence, or it could mean
If God sees that Socrates is seated, then it is a necessary
truth that Socrates is seated,
which expresses the necessity of the consequent. Although the first
reading expresses something true, it does not yield the conclusion
that it is a necessary truth that Socrates is seated, given the addi-
tional premise that God sees that Socrates is seated. On the other
hand, the latter reading, from which the undesirable conclusion does
validly follow, is false. Thus Aquinas uses the distinction between
necessity of the consequence and necessity of the consequent to show
that the proffered argument is invalid or unsound. Either way we are
spared the conclusion he wants to avoid.
The examples just considered make it clear that we have here three
distinct pairs of concepts, which may overlap in their applications.
St. Thomas, in the passage just considered, seems to think that in
his example the distinctions coincide, that applying any one of them
gives the same result. But it is not clear whether Aquinas thinks this
is true in general. The distinctions between the composite and divided
senses and between de re and de dicto are apparently more at home
when applied to categorical propositions. So Aquinas turns the am-
biguous hypothetical into an equivalent ambiguous categorical propo-
sition to make his point. Thus
What is seen sitting must necessarily be sitting
82 QUANTIFIED MODAL Lome
is true if understood compositely, i.e., with necessity applying to the
composite:
Necessarily, what is seen sitting is sitting.
Obviously this is also a de dicto reading. The modality is external
to the statement; it is applied to the dictum. However, understand-
ing Aquinas's ambiguous example in the divided sense means apply-
ing the modality to the predicate without simultaneously applying
it to the subject-
What is seen sitting is sitting necessarily.
Here the thing (res) that is seen to be sitting, whatever it might be,
is said to have the property of sitting and to have it in a necessary
manner or mode. In the present example, this claim is false, since
what is seen to be sitting is Socrates and he does not have the prop-
erty of sitting necessarily.
Despite the coincidence of the three distinctions in St. Thomas's
example, it remains clear that we cannot claim that in general these
distinctions will coincide or that each distinction has the same ba-
sis. The distinctions between necessity of the consequent and ne-
cessity of the consequence, and that between the composite and the
divided sense, seem simply to be distinctions of scope of modali-
ties. It is not true in general that the terms of these pairs pick out
a different kind of modality- that "necessarily" or "possibly" have
different meanings, depending on whether the modality is understood
compositely or dividedly.
Returning now to our distinction between modality de re and mo-
dality de dicto, we might think that here we have a distinction be-
tween two different kinds of modality- between the manner in which
some individual has a property and the manner in which a statement
(dictum) is true. Statements of modality de re frequently make use
of the adverb 'essentially' in place of 'necessarily.' This association
with essences has, in part anyway, served to make the notion of mo-
dality de re suspect. Some philosophers who are suspicious of essences
have suggested that modality de re is intelligible or acceptable only
if it can be explained in terms of modality de dicto. On the other hand,
such eminent medieval logicians as Peter Abelard, Peter of Spain,
and William of Sherwood thought of modality de re as basic, choos-
ing to explain de dicto modality in terms of the essential possession
of truth by a statement.
Here I will make no attempt to decide which type is basic, if in-
deed either one should be considered "reducible" to the other. Like-
 Two DISTINCTIONS 83

wise, I will not attempt to refute arguments against the legitimacy

of either kind of modality. And I will not try to assess the basis of
either kind of modality. All of these issues are interesting topics, wor-
thy of further investigation. 23
We are going to take de re modality to express the essential pos-
session of a property by an individual. We use the □ to represent
de re as well as de dicto modality. In our symbolism, we will continue
to regard an occurrence of a modality sign as representing modality
de dicto if and only if that formula and only that formula within the
scope of modality contains no free occurrences of any individual vari-
ables. The easiest way to apply this test is as follows: Delete all parts
of the whole formula except for what appears within the scope of the
modal sign under consideration. If what remains contains no free oc-
currences of any individual variable, the modality is de dicto. Corre-
spondingly, a modality is considered de re if and only if the formula
within the scope of the modality does contain at least one free occur-
rence of an individual variable. Hence, in the following example, the
first (outermost) occurrence of □ is modality de dicto and the second
occurrence of □ is modality de re:
□ (x) D (Cx :JHx).

In giving rules for quantified modal logic, we frequently will be

forced to take stands on complex and hotly disputed metaphysical
issues involved in understanding essence and existence, and de re and
de dicto modality. Part of our purpose is to exhibit how and where
these issues arise, and what the consequences are of various posi-
tions that can be taken on these issues.
Before we can see what rules it would be appropriate to have for
□ in the context of quantification theory, we must first make sense
of applying the □ to predicates. Prior to this chapter we have been
reading (interpreting) □ as meaning "necessarily" or "it is necessar-
ily true that." But this reading will not do when we apply the □ to
predicates (or to symbolic representations of predicates). Reading
(x) □ ax as "for anything x, x has a is a necessary truth," is unsatis-
factory. For one thing, the expression 'x has a• does not express a
truth (or a falsehood), let alone a necessary one. So how can an op-
erator like □, as we have been reading it, be sensibly attached to it?
The answer seems to be that it cannot. On the other hand, if we read
the □ as "essentially," and say, "for any x, x has a essentially," then
we are using a different interpretation of the □, and we must justify
our continued use of the symbol □ and establish a clear and com-
patible connection with our earlier de dicto interpretation.
84 QUANTIFIED MODAL Loo IC

Our strategy here is the latter approach. This strategy amounts

to assuming that the logic of "essentially" and of essential proper-
ties, parallels that of "necessarily" and of necessary truths, and that
we can find some straightforward logical connections between them.
Much of what is done subsequently is designed to justify or at least
vindicate these assumptions.

4.3 Symbolization

How to translate philosophical idioms into symbols becomes in-

creasingly subtle and difficult in quantified modal logic. These prob-
lems frequently center on the scope of the modality and problem of
rendering claims about properties in first-order logic. Consideration
of an example will illustrate the point.
In Meditation VI, Descartes announces that the body is essen-
tially divisible while the soul is essentially indivisible. Consider the
first claim, that the body is essentially divisible. How is this claim
best understood?
1) □ (x) (x is a body::>x divisible) 24
2) (x) (x is a body::, □ (x is divisible))
3) □ (x) (x is a body::, □ (x is divisible))
4) (x) □ (x is a body::, x is divisible)
5) (x) □ (x is a body::, □ (x is divisible))
(1) is a straightforward de dicto reading. It is said to be a necessary
truth that everything that is a body is divisible, or, alternatively, that
there is a necessary (essential) connection between being a body and
being divisible. (2) says something quite different: that everything
that is a body, (regardless of whether it is a body accidentally or es-
sentially) has the property of being divisible essentially, although
there need not be any necessary connection between having a body
and being essentially divisible. To put it a bit differently, (2) might
be true even if it is possible for something to be a body without be-
ing essentially divisible, as long as everything that is a body is es-
sentially divisible.
The third reading (3) combines the first two. It says that it is a
necessary truth that anything that is a body is essentially divisible.
It excludes the possibility just mentioned above.
The fourth reading is best compared with the first. It is a de re
reading, unlike (1), and it says that given any thing there is, it is an
essential property of it that it is divisible if it is a body. The essential
 SOME OBJECTIONS TO QUANTIFIED MODAL LOG IC 85

property here is a hypothetical property- an "if-then" property. For

any thing there is, it could not have had a body and yet been indivisi-
ble. But to say that everything there is is essentially such that if it
is a body then it is divisible does not preclude the possibility that
there might have been something that was a body but nevertheless
was indivisible. Saying that it is a necessary truth that whatever is
a body is divisible does preclude this latter possibility. Reading (1)
implies (4), but (4) does not imply (1).
In reading (5), just like (4), we have an essential connection between
two properties with the additional feature that one of the properties
is said to be essential. (5) says that for all the things there are (but
not necessarily for all the things there could be), none of them could
both be a body and fail to be divisible essentially. For any thing there
is, being essentially divisible is essentially connected with (implied
by) its being a body.
Which of these readings might Descartes have had in mind when
he says that the body is essentially divisible? Obviously we would
have to take account of the context in the Meditations, Descartes's
metaphysics, and even of the history of metaphysics to produce a de-
finitive answer to that question. I happen to think that the first read-
ing is the best, but I believe a case could be made for some others
as well.
It is worth noting here while we are thinking about what these
readings mean, that (3) is the strongest. It implies each of the others,
and no one of these others implies it. Once we have a set of rules in
place, we will be able to confirm this intuition in our symbolic system.

Exercise: Symbolize, giving what you regard as the most plausible

reading. Be able to explain your symbolization. Some of the exer-
cises are genuinely ambiguous. Give the various possible readings
in such cases.
(1) Everything exists necessarily.
(2) Everything exists essentially.
(3) Possibly persons have bodies, but they can exist without
them.

4.4 Some Objections to Quantified Modal Logic

Of those philosophers who have objected to quantified modal logic,

one of the most notable is W. V. 0. Quine. Quine's central objection
86 QUANTIFIED MODAL LOGIC

is that if we permit quantification into de dicto modal contexts, we

generate fallacies, but if we interpret those contexts as expressing
de T"emodality, then we are committed to "Aristotelian Essentialism,"
which for him is an unacceptable metaphysical view.
Quine does not couch his objections in terms of a de re/de dicto
distinction. He prefers to speak of modality or necessity that attaches
to a thing as contrasted with modality or necessity that attaches to
a way of speaking. In Quine's view," ... necessity resides in the way
we say things, and not in the things we talk about." 25 Necessity that
resides in the way we say things is, for Quine, a barely acceptable
way to interpret de dicto necessity, analytic truth, and Lewis's strict
sense of necessity. Quine observes that
According to the strict sense of 'necessarily' and 'possibly,' these state-
ments would be regarded as true:
(15) 9 is necessarily greater than 7,
(16) Necessarily if there is life on the Evening Star then there is life
on the Evening Star,
(17) The number of planets is possibly less than 7,
and these as false:
(18) The number of planets is necessarily greater than 7,
(19) Necessarily if there is life on the Evening Star then there is life
on the Morning Star,
(20) 9 is possibly less than 7.26

But, he continues, there is a serious problem with these contexts of

modal terms,
... for substitution on the basis of the true identities:
(24) The number of planets= 9,
(25) The Evening Star= the Morning Star
turns the truths (15)-(17) into the falsehoods (18)-(20). 27
In making his objection, Quine takes what we have called the de
dicto reading of modal proposition as the only reading that is philo-
sophically acceptable. He then observes that substitution of identi-
ties into de dicto modal contexts enables us to infer falsehoods from
truths by otherwise impeccable rules of inference, and he concludes
that we ought not to permit quantification into de dicto modal
contexts.
But the advocate of quantified modal logic need not disagree with
Quine on the points just made. Indeed advocates of quantified modal
logic typically do not want to quantify into de dicto modal contexts.
Rather they disagree with Quine about the propriety of recognizing
 SOME OBJECTIONS TO QUANTIFIED MODAL Lome 87

de re modalities - modalities attaching to things. This is a disagree-

ment, as Quine observes, about the tenability of "essentialism," the
views that things have some properties essentially and others con-
tingently. Quine flatly rejects this position, although it is difficult
to see what his reasons are beyond the already stated claim that mo-
dality attaches only to statements, not to things.
Those who find essentialism true or even tenable can distinguish
two claims in
Nine is necessarily greater than seven.
One is the de dicta claim that
It is a necessary truth that nine is greater than seven;
the other is the de re claim that
Nine has the property of being greater than seven essen-
tially or necessarily.
It is this latter claim that is thought to provide a context into which
we may legitimately quantify and substitute.
In short, Quine's objections to quantified modal logic do not cen-
ter on purely formal aspects of quantified modal logic, but on a cer-
tain interpretation of the modal operations. Since we have already
said we do not intend to interpret the D as indicating de dicta mo-
dality when the D has as its scope an open sentence, we need not
worry about Quine's objection that quantification is illegitimate and
that substitutivity fails in those contexts. Instead, we must turn to
the task of explaining more clearly what we mean by calling these
contexts de re and reading the D as "essentially." 28
Before we turn directly to the question of essential properties, we
should remember the situation. When we dealt with propositional
modal logic, we considered the meaning of the modal operator before
we turned to formal systems, and we judged formal systems by their
adequacy to the primary meanings of the modal terms. At this stage
of the discussion we already have first-order logic and we already
have a system of modal logic, S5 (or S4 or T, for those who prefer
the weaker systems). We will pursue the "finer" questions of inter-
pretation and of metaphysics by combining the modal systems with
first-order logic and raising our questions in a formal and philosophi-
cal context, rather than by trying to anticipate and answer questions
prior to introducing the formalism. This procedure may also have an
advantage in helping us identify the questions on which we will have
to take stands, and enable us to avoid (or not, at our pleasure) some
88 QUANTIFIED MODAL Lome
controversies that do not have to be resolved in order for us to agree
on quantified modal logic.

4.5 Essential Properties and Possible Worlds

We have been thinking about modality in terms of possible worlds.

But we must not think of a possible world as some kind of concrete
object sitting out there waiting to be inspected. Nor should we think
of it as some shadowy domain of possible but non-actual objects.
Possible worlds are abstract entities: maximally consistent sets of
statements. Given that statements exist and exist as abstract enti-
ties independent of human thought, possible worlds exist all right,
and exist independently of whether or not we choose to think about
them. Their existence, however, is the same sort of existence we as-
cribe to sets or to numbers.
We have described necessary statements as being true in all pos-
sible worlds and possible statements as those true in at least one
possible world. That description was fine for the de dicto modalities
considered in chapter II. But now our project is to make sense of
de re modal claims such as
(1) Socrates is necessarily (or essentially) rational
and (2) Socrates possibly has a Roman nose,
or general statements like
(3) All men are essentially rational
and (4) Some men are possibly snubnosed.
This involves saying something about what it means for an individual
to exist in a possible world and to have properties in a possible world.
Perhaps the most straightforward initial suggestions for under-
standing statements like (1) and (2) are to take (1) as telling us that
Socrates is rational in every possible world and (2) as telling us that
Socrates is Roman-nosed in at least one possible world. Furthermore,
these suggestions seem to have the additional advantage of allowing
us to reduce de re modal statements to de dicto ones. For example,
the statement that Socrates is necessarily or essentially rational could
be read as saying that the statement that Socrates is rational is true
in every possible world. Likewise, to say that Socrates is possibly
Roman-nosed could be taken as equivalent to saying that in some
world it is true that Socrates has a Roman nose.
If we choose to adopt this course, we must face up to certain con-
 ESSENTIAL PROPERTIES AND POSSIBLE WORLDS 89

sequences. One of the first is that contingent entities cannot be said

to have essential properties. Presumably rationality is one of Socrates'
essential properties. However, it is not true in every possible world
that Socrates is rational, and hence, according to the reading we are
considering, Socrates is not essentially rational. Pursuing the point
just a bit further, it seems that the reason why there are worlds in
which it is not true that Socrates is rational is that Socrates does
not exist in those worlds. He is a contingent being, existing in some
worlds and not others. Obviously it will be true in general that con-
tingent beings cannot have essential properties, given this present
de dicto way of understanding statements about essential and pos-
sible properties. This result, however, runs counter to the strong in-
tuitions many philosophers have had about essential properties.
Some philosophers have suggested that individuals can have
properties in worlds in which they do not exist. On this suggestion,
then, it might be true in every world that Socrates is rational even
though in some worlds it is false that Socrates exists. Again, how-
ever, this is a suggestion that runs counter to the intuitions of many
philosophers.
The best course of action at this point, it seems, is to look for an
alternative way of understanding having a property essentially. But
before we do that let us make sure we understand some of these locu-
tions that we have begun to throw around. In particular, we should
say what it means to exist in a world and to have a property in a
world. An individual is said to exist in a world just in case it is im-
possible for that world to be actual and that individual to fail to ex-
ist. Alternatively, we could say that the individual would exist, were
the world in question actual. An individual has a property in a world
if and only if it is impossible for that world to be actual while the
individual lacks the property.
Now we are ready to try to give a properly de re reading of state-
ments like Socrates is essentially rational. An individual is said to
have a given property essentially if and only if it is not possible that
the individual exist and lack the property. Alternatively, an individ-
ual must have the property in every world in which it exists. This
account lacks the counterintuitive implications of the proposal just
considered above, but it has some implications of its own that should
be observed. The first implication is that anything that exists has
existence essentially. This may seem implausible initially, but it does
accord with the fact that philosophers have found existence to be a
peculiar property at best, so peculiar that many philosophers have
been led to declare that existence is not a property.
90 QUANTIFIED MODAL Lome

We can deal with possible properties in a similar way. Socrates is

possibly Roman-nosed if in some possible world in which Socrates
exists, he has the property of being Roman-nosed.
There is another assumption usually made with these accounts of
essential properties and possible properties. 29 It is that if individuals
- contingent individuals like Socrates - exist in exactly one possible
world, then all their properties would be essential to them, their es-
sential properties would be the same as their possible properties, and
they would have no accidental properties. Once again, this is a view
that runs counter to our modal intuitions.
There are modal theories that have all "world-bound" individuals
- individuals who exist in one and only one possible world. This sort
of theory was held by Leibniz. An interesting way of preserving our
modal beliefs while holding that all individuals are "world-bound" is
to have counterparts in other worlds for a given individual in a given
world. Leibniz sometimes seems to have thought of matters this way
(see his Correspondence with Arnauld). Recently David Lewis has
worked out a theory of this sort. 30
I prefer to stick with what I think is our ordinary view. Individu-
als exist in some worlds and not in others. They are not "world-bound"
and the set of individuals can vary from world to world. Some pos-
sible worlds lack individuals who exist in the actual world and some
possible worlds contain individuals who do not actually exist.
There are at least two accusations one might be tempted to bring
at this point. One is the charge that we are now committed to hold-
ing that there are things that do not exist. But all we are committed
to saying is that there are sets of properties that would have been
exemplified if some other world had been actual. The same sort of
answer can be given to the charge that we are committed to the ex-
istence of Santa Claus, Superman, Captain Nemo, and Hamlet. In-
deed, these seem to have some status in the actual world. But we
may reply that such fictional characters, although they may even bear
names, are not individuals but sets of properties that may or may
not describe individuals. 31

4.6 Considerations Bearing on Quantified Modal Logic

Our informal look at possible worlds has generated several consid-

erations that will have some bearing on what we will accept as a cor-
rect system of quantified modal logic. Of course, when we speak of
a "correct" system, we mean one that is correct relative to a certain
CONSIDERATIONS BEARING ON QUANTIFIED MODAL LOGIC 91

interpretation. So our look at possible worlds, essential properties,

and so forth shows us certain things that our system, under the in-
tended interpretation, should either make turn out true or at .least
should not rule out.
Here is a list of the considerations that have emerged so far:
1. Only one world is actual.
2. The only actually existing individuals are the individuals
who exist in the actual world.
3. Individuals exist in more than one possible world.
4. There can be individuals existing in some possible worlds
who do not exist in the actual world.
5. There are individuals existing in the actual world who fail
to exist in other possible worlds.
6. Individuals have some of their properties essentially and
other properties accidentally.
7. An individual has a property essentially just in case that
individual has that property in every world in which that
individual exists.
8. The de re understanding of 'necessarily' or' □' is as 'essen-
tially'.
9. Individuals do not have properties in worlds in which they
do not exist.
10. There are names of individuals that refer to that individual
in every world in which that individual exists (rigid desig-
nators).
Most of these considerations have to do with existence, and there
is a good reason for that. We are introducing quantified modal logic
by adding quantification to propositional modal logic. On the usual
understanding of the quantifiers, quantifiers range over all and only
existing individuals. A universally quantified formula is true just ex-
actly when every actually existing individual satisfies (makes true)
the formula preceded by the universal quantifier, and an existentially
quantified formula is true just in case some existing individual sat-
isfies (makes true) the formula preceded by that existentially quan-
tified formula. We want to preserve this understanding when we in-
troduce modality; we do not want our quantifiers to be understood
as ranging over nonexisting things. Quantifiers that occur within the
scope of modal signs, that is to say, quantifiers that occur within de
dicto contexts, receive no special reading. Quantifiers into de re modal
contexts are understood to be ranging over actually existing individu-
als. Perhaps some examples will be helpful.
92 QUANTIFIED MODAL Loa IC
(A) D (x)(x is human ::Jxis rational) is to be understood as say-
ing that in each possible world it is true that all humans
are rational.
(B) (3x)(x) is a human & ◊(x high-jumps 8 feet)) is under-
stood as saying that there is an actually existing individ-
ual who is human and in some possible world that indi-
vidual high-jumps 8 feet.
Some of the considerations we listed above are reasons to preserve
this understanding of quantifiers in quantified modal logic.

4.7 The Systems

We turn at last to some systems of quantified modal logic. The

proposals we will consider are all of fairly recent origin, reflecting the
fact that only recently have serious advances been made in this branch
of logic, despite the long history of the modal concepts. We will be
looking at quantified S5, although anyone who wants to deal only
with quantified T or quantified S4 can so restrict things by allowing
use only of T-reit or of S4-reit, depending on the system desired. The
quantification rules and other restrictions remain the same from sys-
tem to system.

4.8 A Kripke System (1959)

The simplest and most straightforward way for us to generate a

system of quantified modal logic at this point is to add the quanti-
fier rules given in chapter III to the modal logic and propositional
logic rules we already have. All we need do is extend the definition
of a well-formed formula so that predicates (i.e., expressions contain-
ing free variables) are also governed by our operators and subject to
our rules. In effect we have already presupposed an understanding
of this.
The resulting system of quantified modal logic is equivalent to one
proposed and studied by Saul Kripke in 1959, and by Hughes and
Cresswell (1968). Kripke took his system to be quantified S5. Later
he revised that claim, for reasons we will consider shortly. This Kripke
system is easy to use, requires no special restrictions, and presents
no special technical problems. It is simply the combination of the
methods of the first three chapters. However, it is crucially impor-
 A KRIPKE SYSTEM (1959) 93

tant to note one thing. The nee intro subproof rule uses the reiter-
ated lines as assumptions. Any line introduced into a nee intro sub-
proof (this includes poss elim subproofs) by a reiteration rule (T-reit,
S4-reit, S5-reit) is an assumption within that subproof and thus is
subject to restriction (3) in the statement of the Universal General-
ization rule. In other words, if we bring a formula containing a free
variable into a nee intro or poss elim subproof by means of a reitera-
tion rule, we may not apply UG to that variable within that subproof.
'lb illustrate this system, we will prove a few theorems.
First, we prove a quantified de re version of the characteristic for-
mula of S5:
(x)◊Fx::J(x) □ ◊Fx
(x)◊Fx asp
◊Fx UI
□ l◊Fx S5-reit
□ ◊Fx nee intro
(x) □ ◊Fx UG
(x)◊Fx::J(x) □ ◊Fx impl intro
The following more interesting formula is also provable in this
system:
◊(3x)Fx=(3x)◊Fx
◊(3x)Fx asp
-(3x)◊Fx asp
◊-(3x)◊Fx poss intro
DI ◊-(3x)◊Fx S5-reit
□ ◊-(3x)◊Fx nee intro
-◊ □ (3x)◊Fx def of □ ,◊
□ (3xFx asp for poss elim
EI asp
Fx poss intro
◊Fx S5-reit
(3x)◊Fx EG
(3x)◊Fx nee intro
□ (3x)◊Fx EI
◊ □ (3x)◊Fx poss elim
(3x)◊Fx RAA
◊(3x)Fx ::J(3x)◊Fx impl intro
94 QUANTIFIED MODAL LOG IC

and
(3x)◊Fx asp
~◊(3x)Fx asp
□ (x)~Fx Def, DN
□ 1(x)~Fx T-reit
~Fx UI
□ ~Fx nee intro
(x) □ ~Fx UG
~(3x)◊Fx Def, DN
◊(3x)Fx RAA
(3x)◊Fx::> ◊(3x)Fx impl intro

Exercises

1. (x) (Fx -Gx) ::>((x)(Gx-Hx) ::>(x) (Fx -Hx))

2. ( □ (x)(Fx::>Gx) & ~(3x)◊Gx)::>(x) □ ~Fx
3. ◊(x)Fx::>(x)◊Fx
4. 'fry to explain why the converse of (3), (x)◊ Fx ::>◊(x)Fx, should
not be valid. 'fry to think of a counterexample. What restriction
on our rules blocks the attempt to prove it?
5. (x)(Fx-Gx)::>((x)Fx-(x)Gx)

Alas, mixing quantification and modality is not as easy as this

system makes it appear. The example and exercises you have just
finished are designed to illustrate where the difficulties are. The diffi-
culties philosophers have had with this Kripke system center on the
logical relationships between de dicto and de re modalities. There are
four notorious formulas which are the focal points of controversies:
Barcan Formula: 32 (x) □ ax ::>□ (x)ax
Converse Barcan Formula: □ (x)ax ::>(x) □ ax
Buridan Formula: 33 ◊(x)ax::>(x)◊ax
Converse Buridan Formula: (x)◊ax::> ◊(x)ax

These four formulas, as we are interpreting them, express relations

of modality de dicto to modality de re. The Barcan Formula, if part
of a system of modal logic, would sanction the inference of a de dicto
necessity from a de re counterpart, and the Converse Barcan For-
mula licenses the reverse inference. The Buridan and Converse Buri-
dan Formulas would do the same for de dicto and de re possibility.
As anyone who has read the text and done the exercises now knows,
the system under consideration has as theorems the Barcan, Con-
verse Barcan, and Buridan formulas. It also is easy to see that the
 OBJECTIONS TO Tins SYSTEM 95

corresponding versions of quantified T and quantified S4 also con-

tain the Buridan Formula and the Converse Barcan Formula, but not
the Barcan Formula, since the proof of the latter requires the use of
S5 reiteration. The Barcan Formula can be added to these systems
without generating any formal inconsistency, however.

4.9 Objections to This System

We have observed earlier that objections to systems of modal logic

are typically not merely formal objections. They are objections to
a certain interpretation of a formal system. The objections considered
here are also of this latter sort. These objections presuppose that we
are trying to present a certain type of possible worlds interpretation.
The objections are ways in which the present formal system, under
interpretation, fails to represent the truth of certain modal claims.
The objections to this first attempt to add quantification to the
S5 system of modal logic can be seen clearly if we pursue two related
questions: 1) How does our view of possible worlds fit with this sys-
tem? and 2) what is the relationship between de re and de dicto mo-
dality in this system?
We start with the second question. The answer is provided for the
most part by the presence in the system of certain formulas, formu-
las that relate the representations of modality de dicto and modality
de re under the interpretation we are considering. We have in this
system as a theorem, the Barcan Formula
(x) □ ax:) □ (x)ax,

the Converse Barcan Formula

□ (x)ax:) (x) □ ax,

and the Buridan Formula

◊(x)Ux:J(x)◊Ux.

The Barcan Formula and its converse make de dicto and de re neces-
sity equivalent in this system. The Buridan Formula allows us to in-
fer de re possibility from de dicto possibility. We have already noted
that its converse is not a theorem of this system.
The Barcan and Buridan Formulas have proven to be highly con-
troversial under our present interpretation. Indeed, the situation is
even worse than that. These formulas seem to express false claims
96 QUANTIFIED MODAL LOG IC

under our interpretation. Counterexamples to both may be found in

Plantinga's The Nature of Necessity (Oxford, 1974), chap. 4, §9.
Plantinga uses a counterexample to the Buridan Formula given
by Buridan himself in the 14th century. It seems to be possible that
God should not have created anything and that therefore it is pos-
sible that nothing exists besides God. But it does not follow from
this that everything that in fact exists is possibly identical with God.
For example, you and I, who are actually existing things, are not
even possibly identical with God. There is no possible world in which
I am God and, for that matter, none in which you are either.
Likewise, the Barcan formula does not express a necessary truth.
Imagine a universe in which the only existing things are essentially
immaterial objects-God, sets, numbers, properties, etc. Then it would
be true that everything is essentially immaterial, but it would not
follow that it is a necessary truth that everything is an immaterial
object, since God could (and did) create some material objects.
On the other hand, it seems correct that the Converse Barcan for-
mula should be included as a theorem of our system of modal logic,
and also correct that the Converse Buridan Formula should be ex-
cluded. Concocting a counterexample to the latter is left as an exer-
cise for the reader.
Other difficulties occasioned by the Barcan and Buridan formu-
las show up when we consider what sort of possible worlds are re-
quired to satisfy the present system of quantified modal logic. The
counterexamples make use of sets of possible worlds with different
individuals in them. But any possible worlds model of our system
will have to have the same individuals in every possible world.
'lb see this, suppose that we have a possible world in which every-
thing has a certain property a essentially. That is to say, everything
in this world has a and has a in every world in which it exists. As
we have just seen from the counterexample to the Barcan Formula,
we cannot count on everything in another world having a if there
are individuals existing in the latter world who do not exist in the
former. Having the Barcan Formula valid means in effect that no
world can have individuals in it who do not exist in every other world.
Since the actual world is the standard for the things there are, the
Barcan Formula, if valid, would have the effect of making it the case
that no possible world has any individuals who do not exist in the
actual world, a strongly counterintuitive conclusion.
This conclusion can be evaded, but only at the price of other im-
plausibilities. One way to evade this conclusion is to hold that things
can have properties in worlds in which they do not exist. Another
 KRIPKE's REVISED S5 (1963) 97

is to say that all essential properties are the kind of properties noth-
ing can fail to have in any possible world- properties like being either
red or non-red and like being such that 2 + 2 = 4. These sorts of prop-
erties are sometimes called trivially essential properties. So to have
a quantified modal logic that can represent significantly essential
properties and do so without requiring that individuals have proper-
ties in worlds in which they do not exist or requiring that no possible
world have any individuals that do not exist in the actual world, we
will have to jettison the Barcan Formula.
We should consider, too, what effect including the Buridan For-
mula has on our possible worlds model. By the Buridan Formula, if
there actually is something that has an essential property, then in
every world there is something that has that property. Now if each
individual is essentially unique, i.e., has some property essentially
that cannot be possessed by any other individual, then no individual
in the actual world can be absent from any other possible world. For
example, if being identical with Socrates is essential to Socrates, and
only Socrates can have this property, as seems plausible, then there
are properties of this sort. But then we have the conclusion that no
world lacks any individual who exists in the actual world.
Combining the results of considering these two formulas, we have
the conclusion that exactly the same individuals exist in all possible
worlds. But that implies, of course, that all individuals exist neces-
sarily. So either we are stuck with that blatant falsehood, or we have
a modal system that is satisfied only by necessary beings - those
things that exist necessarily. This means that the system cannot deal
with essential properties of contingent beings, such as you and me
and our earthly possessions. Either way it seems desirable to look
for another system.

4.10 Kripke's Revised S5 (1963)

The problem is this. Just a few years before Kripke's 1959 paper,
Arthur Prior had published a deduction of the Barcan formula from
the axioms of standard quantification theory and S5. 34 It seemed
therefore that the Barcan Formula (and the Buridan and Converse
Barcan Formulas) were simply part of quantified S5 and that it was
impossible to get rid of them without going to a different system.
But in 1963 Kripke claimed that there is a subtle fallacy in Prior's
proof and that the system we have just considered is an improper
version of S5.
98 QUANTIFIED MODAL Lome

In presentations of quantification theory, one of the axioms is cus-

tomarily stated as follows:
(x) ax-::;ay.

Strictly speaking, it is an open sentence, not a proposition, as it is

here stated. Of course, Kripke pointed out, it is understood to mean
(y)((x)ax-::; Uy)
and ordinarily not writing down that initial quantifier creates no prob-
lems. However, Prior's derivation of the Barcan Formula requires
applying the rule of necessitation (If p is a theorem, then □p is a
theorem) to the formulation of the axiom with the initial quantifier
dropped. The proof will not work if we may use only the second,
stricter form of the axiom. What I have called the stricter form of
the axiom is also referred to as the universal closure of the earlier
form. A universal closure of a formula is obtained by binding all free
occurrences of variables in the formula with universal quantifiers pre-
fixed to the formula.
Kripke's "revision" of S5 involves no different formulations of the
axioms, just a change in the way quantification is understood to apply
to them. We now understand only the universal closures of these for-
mulas to be axioms. Kripke's contention is that this should have been
the proper understanding all along. The effect of this understanding
is that neither the Barcan Formula, the Converse Barcan Formula,
nor the Buridan Formula are theorems of quantified S5. They may
be added without generating inconsistency, but they are not part of
the system.
We can obtain this "revised" system with our natural deduction
rules by adding certain restrictions to the necessity introduction rule
and to the possibility elimination rule.

The necessity introduction rule

(for revised S5 quantified modal logic)

Recall that this rule provides a way of terminating a nee intro sub-
proof. The nee intro subproof rule remains as it was. The additional
restrictions introduced here will only affect the derivation of certain
formulas containing free variables and formulas obtained by a gen-
eralization rule. Propositional logic will not be affected at all.
nee intro
A nee intro subproof may be terminated at any point, and any line,
p, occurring within the subproof and that is not within the scope of
 KRIPKE's REVISED 85 (1963) 99

an undischarged assumption within the subproof, may be entered as

a line prefixed by D with the justification "nee intro," provided that
(i) the derivation of this line, p, does not depend on the application of
UG or EG to a formula containing a free occurrence of a variable which
also has a free occurrence in a line preceding the opening of the nee
intro subproof in question, and provided that (ii) p contains no free oc-
currence of a variable which has been introduced into the nee intro
subproof by UI or EL (The mention of EI here is, strictly speaking,
unnecessary, because the restrictions on the EI rule will also block the
cases involving EI that the restriction is designed to forbid.)

The first proviso of this revised rule uses a technical concept of

"depends on," which is defined as follows.
A line lj of a proof (or subproof) depends on a line l; iff (i)
l; :::>lj or (ii) lj is justified as a direct consequence of rules
or equivalences from some premises, at least one of which
depends on l;.
The general idea is that if a line (a formula) is used in the deduction
of another line in a proof, the latter depends on the former.

Possibility elimination (revised quantified modal logic)

The rule of possibility elimination also needs to be restricted be-

cause, as we have seen earlier, it is officially an abbreviation of a spe-
cial case of nee intro.
poss elim
Where ◊p appears as a line in a proof (and does not lie within the
scope of a discharged assumption), and where an impl intro subproof
opened with p as its assumption within the scope of a nee intro sub-
proof yields a conclusion q, both subproofs may be simultaneously ter-
minated and ◊q entered as the next line with the justification poss
elim, provided that (i) neither q nor any line on which q depends results
from the application of UG or EG to a formula containing a free occur-
rence of a variable that has a free occurrence preceding the opening
of the nee intro subproof in question, and provided that (ii) q contains
no free occurrence of a variable introduced into the nee intro subproof
by UI or EL
The point of these restrictions or provisos is to prevent interchange
of quantifiers and modal signs. This is done by forbidding the ap-
plication of generalization rules within modal contexts (i.e., nee intro
subproofs) to variables having free occurrence outside those contexts
(the first proviso), and by forbidding us to bring out of modal con-
100 QUANTIFIED MODAL LOG IC

texts free variables that were introduced by instantiation within the

modal context (the second proviso).
A good way to test and illustrate these restrictions is to apply them
to proofs of the Barcan and Converse Barcan formulas that would
be acceptable in the earlier system. Consider this proof of the Con-
verse Barcan formula:
□ (x)ax:::)(x) □ ax

1. □ (x)ax asp
2. □ I (x)ax T-reit
3. ax UI
4. □ ax* nee intro
5. (x) □ ax UG
6. □ (x)ax:::)(x) □ ax implintro
*Step 4 now is forbidden by the second pro-
viso, which requires that the step to which nee
intro is applied contain no free occurrences of
a variable introduced by use of an instantiation
rule within the subproof. Line 3 contains a free
occurrence of a variable introduced into the nee
intro subproof by UI.
As further exercises, look back at the proofs given in Section 9
of this chapter of the Barcan and Converse Barcan formulas and de-
termine which restrictions are now violated in those proofs. It may
also be a useful exercise to attempt proofs of the Buridan and Con-
verse Buridan formulas (◊(x)ax:::)(x)◊ax and (x)◊ax:::)◊(x)ax) to
see how such attempts run afoul of the restrictions.
The system under discussion includes as its theorems all and only
the closed formulas deducible by means of the natural deduction rules
given. A closed formula is one that contains no free occurrences of
variables. Many formulas containing free occurrences of variables are
deducible, but they are not theorems of the system. Kripke says in
stipulating this that such open formulas only make sense if we under-
stand them as implicitly closed anyway, so we are not depriving our-
selves of anything we should have, and we are getting rid of some
undesirable formulas. This system he regards (with some justice) as
the proper version of quantified S5. It has the advantage of not yield-
ing the Barcan and Buridan formulas, which run counter to the sense
of the modalities we want to capture. The system also has the advan-
tage of not requiring that the same individuals "populate" each pos-
sible world. We should remind ourselves here that these advantages
 REFLECTIONS ON KRIPKE's REVISION 101

are advantages only when we consider the systems under our intended
interpretation.

4.11 Reflections on Kripke's Revision

This system of quantified S5 presented by Saul Kripke in 1963

is better (for our purposes) than the version found in his 1959 paper.
But it is still not fully satisfactory. Our objections will once again
be directed against the intended interpretation. While the system as
a formal system may be perfectly good for some other applications,
here we are interested in its adequacy as the proper structure for the
possible worlds picture of modality, including de re and de dicta re-
lationships.
First, a deficiency. This version of quantified modal logic includes
none of the principles relating modality de re and modality de dicta:
the Barcan and Converse Barcan formulas as well as the Buridan and
Converse Buridan formulas are false under the intended interpreta-
tion. We have already endorsed counterexamples to the Barcan and
Buridan formulas. We have not yet said anything about the Converse
Barcan formula
□ (x)ax::)(x) □ ax.

Earlier it was suggested that the Converse Barcan Formula is ac-

ceptable. Yet Kripke offers a counterexample (strictly speaking, a
countermodel) to it right along with his counterexample to the Bar-
can Formula. We need to look further at this to see how his revised
system offers a different account of modality than what we are look-
ing for.
Kripke presents a counterexample along the following lines. Sup-
pose there are two possible worlds, the first containing as its only
individuals Ronald and Nancy Reagan, and the second containing
only Ronald. In the first world each is bipedal, in the second world
Ronald is bipedal but Nancy is not, since she does not exist in that
world. In each world, however, everything is bipedal. Hence it is true
in every world (all two of them) that everything is bipedal, that is,
it is necessary that everything is bipedal. But it is false that every-
thing is necessarily bipedal because Nancy is not bipedal in the sec-
ond world. Thus, the Converse Barcan Formula does not express some-
thing true in all possible worlds and is not a principle of quantified
modal logic.
This counterexample serves to make clear that Kripke's interpre-
102 QUANTIFIED MODAL LOGIC

tation of D differs from what we were looking for, and that the ab-
sence of the Converse Buridan Formula is a deficiency. Kripke reads
the Converse Barcan Formula as saying that if it is true in every pos-
sible world that everything in that world has property a, then a is
true of everything there is and true of it in every possible world.
Since Kripke also assumes, rightly, that things do not have atomic
properties in worlds in which they do not exist (his interpretation
makes it false in such a world that the individual has the property),
his system is incapable of expressing atomic essential properties of
entities that do not exist in every possible world. This means that
no contingently existing individuals, such as Socrates, can be said
to possess essentially such properties as being human or being ra-
tional, assuming these to be atomic properties. For it is not true of
Socrates in every possible world that he is human or that he is ra-
tional. Kripke's system is only capable of expressing trivially essen-
tial properties of contingently existing entities by using the □. A
trivially essential property is one that every individual has neces-
sarily, such as being red or non-red or, perhaps, the property of being
such that 2 + 2 = 4.
If, on the other hand, we understand having a property essentially
as not lacking that property in any world in which it exists, and if
we want D as attached to open sentences to express possessing a
property essentially, then it can be seen that the Converse Barcan
Formula should be a theorem of our system. For the CBF simply ex-
presses what is a necessary truth: if it is true in every possible world
that everything is a, then everything has a in every world in which
it exists. And we continue to assume that individuals do not have
properties in worlds in which they do not exist.
There is also a metaphysical difficulty with Kripke's interpreta-
tion of quantified S5 that we will not pursue in detail here: it seems
to require that there are some individuals that do not exist. Accord-
ing to the interpretation he proposes, for each possible world there
is a set of individuals that comprises the individuals that exist in
that possible world. There will be some "overlap" from world to world,
but also some individuals "lost" and others "added." 'Th.king this in-
terpretation, there must be a union of all these sets of individuals -
the total set of all individuals in all possible worlds. Now if we sup-
pose, as seems entirely plausible, that the actual world does not
contain all the individuals there could be, then some other possible
worlds "contain" individuals not to be found in the actual world. But
if this is so, then there must be a union of all the sets of individuals
in all the possible worlds that contains objects that do not actually
 'fiIE ACTUALISTIC VERSION OF POSSIBLE WORLDS 103

exist. Thus, anyone who accepts this kind of interpretation of quan-

tified S5 is committed to the implausible claim that there are some
things that do not exist. 35

4.12 The Actualistic Version of Possible Worlds36

Despite the claims in the propositional modal logic section of this

book that S5 best reflects the truth about matters modal, both ver-
sions of quantified S5 considered here have been metaphysically un-
satisfactory. We will see shortly, however, that only a slight modifica-
tion is required to produce a satisfactory system. But first we should
take a closer look at the metaphysical picture of possible worlds the
system must fit - the actualist view of possible worlds.
Actualism is a species of what textbooks on metaphysics tradi-
tionally call realism. However, actualism is an attempt to avoid the
excesses that have sometimes been associated with realism. While
propositions, properties, and states of affairs all exist for the Actual-
ist, there are no nonexistent objects, nor could there be any. Further-
more, the "serious actualist" holds that objects do not have proper-
ties in worlds in which they do not exist, and he believes it is possible
that there should be objects distinct from all the objects there in fact
are. The actualist rejects any interpreted modal system attempting to
represent broadly logical necessity that fails to honor these truths.
But we should begin at the beginning-possible worlds. What does
the actualist think they are? We will explain them in terms of states
of affairs. We assume that there are states of affairs. There are pres-
ent states of affairs like Ronald Reagan's being President, past ones
like Roger Maris's hitting his 61st home run, fut,ure ones like the Sec-
ond Coming of Christ; there are permanent states of affairs such as
any equilateral triangle being equiangular and 7 plus 5 being 12; there
are states of affairs that never have been and never will be actual,
such as Napoleon's being very tall; and there are states of affairs that
could not be actual, such as the Barber of Seville's shaving all and
only those persons in Seville who do not shave themselves. All of these
states of affairs exist; but not all of them do, or could, actually obtain.
A possible world is a special state of affairs. It is, of course, first
of all a possible state of affairs. (Remember we are not trying to ex-
plain the meaning of possibility here.) Secondly, to be a possible world,
a possible state of affairs must be maximal. What does it mean for
a state of affairs to be maximal?
'lb understand this, we introduce two other concepts, that of one
104 QUANTIFIED MODAL Lome

state of affairs including another and that of one state of affairs pre-
cluding another. Let S 1 and S 2 stand for any states of affairs (not
necessarily different ones). Then S 1 is said to include S 2 if and only if
it is not possible for S 1 to be actual without S 2 also being actual, and
S 1 precludes S 2 if and only if it is impossible that S 1 and S 2 should
both be actual. So, for example, the state of affairs of Roger Maris's
hitting his 61st home run includes the state of affairs of Roger Maris's
having hit his 30th home run, while the state of affairs of something
being an obtuse triangle precludes that thing's being an equilateral
triangle.
These concepts now enable us to define a maximal state of affairs.
A state of affairs, M, is maximal just in case for any state of affairs,
S, M either includes S or precludes S.
According to this definition, given a maximal state of affairs and
any old state of affairs, the latter state of affairs is either already
part of the maximal state of affairs or cannot be added to the maxi-
mal one without yielding an inconsistent state of affairs as a result.
Any genuine enlargement of a maximal state of affairs will be in-
consistent.
Notice, however, that a maximal state of affairs need not be com-
posed of a large number of constituent states of affairs. It may be,
but it may also be composed of just one- a self-inconsistent one, such
as 7 + 5 = 13, which will, of course, preclude every state of affairs.
On the other hand, however, a check of the definition of includes shows
that inconsistent states of affairs tum out to include every state of
affairs, too. The point is that any inconsistent state of affairs is
maximal.
It is clear that we want nothing to do with these inconsistent maxi-
mal states of affairs. We define a possible world, then, as a state of
affairs that is both maximal and consistent. This explanation ac-
counts for possible worlds without the need to introduce any shadowy
metaphysical entities or strange doctrines. Possible worlds can be
nicely explained in terms of entities with which everyone is quite
familiar.
Not only is the notion of a possible world thus explained in quite
familiar terms, it is also explained in a way that makes it perfectly
plausible to suppose that there are possible worlds. If there are states
of affairs, as there surely are, then there are possible worlds. It has
to be kept in mind, of course, that these are abstract entities. It is
not as though there is some place at which possible worlds exist. All
possible worlds, including the one that has been actualized, are ab-
stract entities. Now, from this it does not follow that the physical
world is an abstract entity, but rather that the existence of a pos-
 ACTUALISM AND QUANTIFICATION 105

sible world differs from the actuality of a possible world. All of a huge
number of possible worlds exist; at most one can be and is actual.

4.13 Actualism and Quantification

Inasmuch as we have criticized other interpretations of modal logic

for requiring that there are things that do not exist or that things
can have properties in worlds in which they do not exist, we should
see how the actualist system and interpretation handles these prob-
lems. We begin by spelling out how quantifiers are to be interpreted
in our system.
One's first inclination is to start with the reading of ordinary quan-
tification theory and extend that kind of interpretation to possible
worlds. Quantification is customarily construed as being over all ex-
isting objects.
On our ordinary reading of the quantifiers, there will be a domain
of quantification - a set of objects that are the values of variables.
This domain may be restricted for various purposes, e.g., mathemati-
cians may confine the domain to numbers or sets in the study of
arithmetic or algebra, but ordinarily this domain will consist of every-
thing there is. Of course, philosophers and others notoriously disagree
about what there is, but this is a metaphysical disagreement, not one
about interpreting quantifiers.
Quantified statements are interpreted with reference to the do-
main. Universally quantified statements are understood as saying
that every member of the domain satisfies the condition that follows
the quantifier. Or, stated less technically, the predicate or property
denoted by the expression following the universal quantifier is true
of everything in the domain. And likewise, following the definition
of the existential quantifier in terms of the universal quantifier, an
existentially quantified statement is understood as claiming that it
is false that everything in the domain fails to satisfy the condition
expressed by the expression following the quantifier.
Now it might seem natural to continue to use this interpretation
when working with quantified modal logic, but we have already hinted
that there are certain difficulties involved with it. The set of objects
existing in a possible world may differ from world to world, and we
want to understand quantification in a possible world to be over the
objects that exist in that world (i.e., over the objects that would ex-
ist if that world were actual). However, the only domains of contin-
gent objects actually available are the set of objects that actually
exist and domains that are subsets of that set. If we try to get by
106 QUANTIFIED MODAL LOGIC

without allowing for objects that exist in other possible worlds but
not in the actual world, the alternatives are unpalatable. We either
(1) have to explain quantification in a possible world when part or
all of its domain does not exist, or (2) explain how something can have
a property (at least a nontrivial property) in a world in which it does
not exist, or (3) hold that there are things that do not exist, or (4)
hold that there are no possible worlds containing objects not con-
tained in the actual world.
Assuming we want to hold to a standard reading of quantifiers
and we want some possible worlds to "contain" individuals who do
not exist in the actual world, how shall we construe domains of quan-
tification for possible worlds? The domains of quantification for the
various possible worlds will have to be subsets of actually existing
things and be necessarily existing things in order to satisfy the re-
quirements we have laid down. Alvin Plantinga suggests that we use
essences - individual essences.
What are individual essences and how can it be made clear that
they will serve as the domain of quantifiers in modal contexts? We
have said that an object has a property in a world if it would have
been true that it has the property had that world been actual. For
an object to have a property essentially is to have the property in
every world in which it exists. The introduction of one further con-
cept will enable us to define individual essences. Let us define what
it means for a property to be exemplified in a possible world. A prop-
erty is exemplified in a possible world if the proposition that some-
thing has that property would have been true had that world been
actual. This definition enables us to offer our official definition of an
individual essence. An individual essence is a property that is exem-
plified in some possible world, and in every possible world any thing
that has this property has it essentially, and nothing else in any pos-
sible world has this property. 37 Although a given individual essence
need not be exemplified in any particular possible world, essences
are properties, and since properties presumably are necessary beings,
as such, they exist in every possible world. 38 So, for example, Julius
Caesar's essence exists in every possible world even though Julius
Caesar fails to exist in some possible worlds. We can exploit this fea-
ture of essences to explain quantification in modal contexts.
'lb conclude our explanation, we need to define one more concept,
that of coexemplification. We will say that two properties are coex-
emplified with each other in a possible world when the proposition
that something has both properties would have been true had that
possible world been actual.
 ACTUALISM AND QUANTIFICATION 107

With this, we are in a position to say how the actualist who adopts
quantification over domains of essences reads various types of propo-
sitions. Consider first of all ordinary universally quantified proposi-
tions and existentially quantified ones.
(x) (x is a man::)x is mortal)
is understood as follows: any essence in the domain of essences that
are actually exemplified (i.e., any essence in the domain of the actual
world) is such that if it is coexemplified with the property of being
a man, then it is coexemplified with the property of being mortal.
The existentially quantified statement
(3x) (x is man & x is mortal)
is read like this: Some essence in the domain of essences exemplified
in the actual world is coexemplified in the actual world with the prop-
erties of being a man and being mortal. Some existing thing, which
of course has an individual essence and which is an example of (ex-
emplifies) that essence, has the properties of being human and mortal.
The advantages of this actualist view begin to emerge more clearly
as we consider statements which mix modalities and quantifiers. First
we look at quantifiers within the scope of a modality:
◊(x) (xis a man::)x is mortal)
is interpreted as meaning that in some possible world, every individ-
ual essence in the domain of essences that would be exemplified if
that world were actual is coexemplified with the property of being
mortal if a man. Or, more simply, ◊(x)Ux means that in some pos-
sible world, every individual essence that would be exemplified if that
world were actual is coexemplified with property a. Note that we say
"would be exemplified" and not "would exist." Every individual es-
sence exists in every possible world. But not all are exemplified in
every possible world. Similarly, ◊(3x)ax is interpreted as saying
that in some possible world, some individual essence (in the domain
of essences of that world) is coexemplified with the property a. This
reading boils down to saying that Opis true just in case there is a
possible world in which p is true.
How about examples of modalities within the scope of quantifi-
ers, as in
(x)◊(Ux)
and
(3x)◊ax?
108 QUANTIFIED MODAL LOGIC

Here we are no longer dealing with the truth of a proposition in a

possible world-the modality of a dictum (i.e., modality de dicto)-
but with claims about objects of the actual world (essences exempli-
fied in the actual world) that assert that these essences are in some
(or all) possible world(s) coexemplified with a certain property or prop-
erties. (x)◊ax says that every essence in the domain of the actual
world (those essences exemplified in the actual world) is such that
in some possible world that essence is coexemplified with the prop-
erty a. The second, (3x)◊ax, says that some essence exemplified in
the actual world is in some possible world coexemplified with the pro-
perty a.
Now consider a related statement that embeds the modality deeper:
Some man is possibly mortal, i.e.,
(3x) (xis a man & ◊(xis mortal)).
This is read as saying there is an essence in the domain of the actual
world, coexemplified with the property of being a man, and this es-
sence is coexemplified in some possible world with the property of
being mortal.

Exercises

A. Symbolize and interpret the following. Where the statement is

ambiguous, give the possible readings and try to say which is
more plausible.
1. Possibly (all) minds can exist apart from (all) bodies.
2. (All) minds can exist apart from bodies.
3. Possibly there is a golden mountain.
4. There is something that could be a golden mountain.
5. All men are rational essentially.
6. All bachelors are essentially unmarried.
B. Argue in terms of the present interpretation for your answers to
the following:
1. Does
a) (3x)◊(x is F)-◊(3x) (xis F)?
b) ◊(3x)(x is F) -(3x)◊(x is F)?
2. Is ◊(3x) (x is a flying horse) necessary or contingent?

4.14 Actualistic Quantified Modal Logic

Despite our claims in the propositional modal logic section of this

book that S5 best reflects the truth about matters modal, the first
 ACTUALISTIC QUANTIFIED MODAL Lome 109

two versions of quantified S5 considered here have proven metaphysi-

cally unsatisfactory. The satisfactory metaphysical view we have
called actualism- there neither are nor could be any nonexistent ob-
jects. We also hold that something must exist in a world to have
properties in a world, and we hold that other possible worlds have
things existing in them that do not exist in the actual world. Now
we want a system that under the interpretation outlined in the pre-
vious section yields or permits these conclusions and at the same
time enables us to demonstrate the validity of standard modal in-
ferences.
The actualist system of quantified modal logic for which we give
natural deduction rules was axiomatized by Thomas L. Jager. 39 It
is very close to the Kripke system, differing only with respect to
some theorems of quantified modal logic. There are actualist versions
of T, S4, and S5, and it is a very simple matter to move from one
system to the other, just as it was in propositional modal logic. Jager's
system A is a quantified modal logic obtained by adding quantifica-
tion in a certain way to S5 propositional modal logic; actualist ver-
sions of T and S4 can be obtained in the same way.
The differences between this actualistic system and the other two
systems presented here can be summarized in terms of what they
do with the formulas governing the mixing of quantification and
modality-the Barcan and Converse Barcan formulas, and the Buri-
dan formula. (All three systems exclude the Converse Buridan for-
mula.) The first Kripke version of S5 had all three, and the second
rejected all three. The actualistic system excludes the Barcan and
Buridan formulas but includes the Converse Barcan formula as a
theorem.

4.14.a Rules

We obtain the actualistic system of quantified modal logic by tak-

ing our rules for propositional modal logic, adding quantification rules,
and restricting the use of the rules nee intro and poss elim. The re-
strictions will apply to these rules only in contexts involving quan-
tifiers and free occurrences of variables. The restriction on the nee
intro rule is designed to prevent us from "losing'' variables that have
a free occurrence in lines used in deducing a conclusion to which we
want to apply the nee intro rule. The restriction on the poss elim rule
is designed to prevent us from "gaining" free occurrences of variables
between the point of making an assumption for the subproof used
in the rule and the use of the poss elim rule.
Actualistic quantified modal logic then, has the following rules:
110 QUANTIFIED MODAL Lome

Nee elim
Poss intro
Nee intro subproof rule
Reit (T-reit, S4-reit, S5-reit)
Quantification rules (UI, UG, EI, EG)
Nee introA
A nee intro subproof may be terminated at any point, and any line,
p, occurring within the subproof and that is not within the scope of
an assumption undischarged at that point in the nee intro subproof
may be entered as a line prefixed by □ with the justification "nee in-
tro," provided that the line p contains a free occurrence of every vari-
able that has a free occurrence in any line within the nee intro subproof
on which p depends.
Poss elimA
Where ◊ p appears as a line in a proof (and not within the scope of a
discharged assumption) and where an impl intro subproof opened with
p as its assumption within the scope of an nee intro subproof yields
a conclusion, q, both subproofs maybe terminated simultaneously and
◊q entered as the next line in the proof with the justification poss
elim, provided that every variable which has a free occurrence in q also
has a free occurrence in p.

The best way to see how the restrictions work and how to use them
is to consider some examples. Consider the following proof and be
prepared to compare it with the succeeding one:

Prove: (x) □ ((Fx&Gx):JHx):J(x)( □ (Fx & Gx):J □Hx)

1. (x) D ((Fx & Gx) :JHx) asp
~ D~&ili) ~
3. D((Fx & Gx) :JHx) UI
4. D Fx & Gx T-reit
5. (Fx & Gx) :JHx T-reit
6. Hx MP
7. □Hx nee intro
8. D (Fx & Gx) :J □Hx impl intro 2-7
9. (x)(D(Fx & Gx) :J □Hx) 8,UG
10. (x)□((Fx&Gx):JHx):J(x)(□(Fx&Gx):J □Hx) impl intro 1-9

Here the only step deduced within the nee intro subproof is line
6, derived from lines 4 and 5. Lines 4 and 5 contain free occurrences
of the variable x, and x occurs free in line 6, thus satisfying the re-
 AcTUALISTIC QUANTIFIED MODAL Lome 111

striction. Compare this with a similar formula for which a similar

"proof" does not go through because of the restriction:

Prove: (x)(y) □ ((Fx&Gy) :>Hx)::, (x)(y)( □ (Fx&Gy)::, □ Hx))

I. (x)( ) □ ((Fx&Gy) :>Hx) asp

2. □ (Fx&Gy) asp
3. □ ((Fx&Gy):>Hx) UI
4. □ Fx&Gy T-reit
5. (Fx&Gy) :>Hx T-reit
*6. Hx MP
7. □Hx nee intro
8. □ (Fx&Gy):> □Hx impl intro
9. (x)(y)( □ (Fx&Gy):> □Hx) 8 UG
10. (x)(y) □ ((Fx&Gy) :>Hx)::, impl intro
(x)(y)( □ (Fx&Gy)::, □Hx)

The proofs may seem identical, but this latter proof violates the
restriction on nee intro subproofs in line 6, the starred line. Line 6
depends on lines 4 and 5. Lines 4 and 5 each contain free occurrences
of two variables, x and y, while line 6 contains only an occurrence
of x. The free variable y has been "lost." Hence the restriction has
not been satisfied and this second "proof" is fallacious.
Next, consider a pair of proofs illustrating the use and misuse of
the poss elim rule:

Prove: (x)(◊Ax::, ◊(Ax v Bx))

I. ◊(Ax) asp for impl intro
2. asp for poss elim
3. l~vBx add
4. ◊(AxvBx) poss elim
5. ◊Ax::, ◊(Ax v Bx) impl intro
6. (x)(◊Ax::, ◊(Ax v Bx)) UG
When the poss elim rule is applied to line 3 to derive line 4, the re-
striction is satisfied. Only one variable, x; occurs in line 3 and it is
the same one that occurs in line 2, the assumption opening the poss
elim subproof. Hence, no "new" variables were "gained" within the
subproof.
Compare the above proof with the one below. The one below, like
the preceding, appears to prove a formula corresponding to a theo-
rem of quantification theory.
112 QUANTIFIED MODAL LOGIC

Prove: (y)(x)(◊Ax:::> ◊(Ax v By))

1. ◊ asp for impl intro
2. ~ asp for poss elim
3. IAxvBy add
*4. AxvBy) poss elim -violates restriction
5. ◊Ax::>◊(AxvBy) impl intro
6. (y)(x)(◊Ax :::>◊(Ax v By)) UG (twice)
Although this proof parallels its predecessor, this latter proof con-
tains a violation of the restriction on poss elim in line 4. This time
there is a free occurrence of the variable y that does not occur free
in line 2, the assumption for the application of poss elim.
What happens now to the most exciting formulas such as the Bar-
can and Converse Barcan formulas? The Converse Barcan formula
is easily proven:

Prove: □ (x)Fx::>(x) □ Fx

1. □ xFx asp
2. □ 1(x)Fx T-reit
3. Fx UI
4. □ Fx nee intro
5. (x) □ Fx UG
6. □ (x)Fx::>(x) □Fx impl intro
The restriction on nee intro subproofs is trivially satisfied since the
only deduction within such a subproof is the deduction of line 3 from
2, and line 2 contains no free occurrences of any variable. Attempts
to prove the Barcan Formula run afoul of restrictions, however. The
reader can verify this by turning back to the proof of this formula
in section 9, earlier in this chapter. (There the formula appears in an
equivalent form, ◊(3x)Fx:::>(3x)◊Fx.) The Barcan Formula is not a
theorem of this system, which is just what we intended.

4.14.b Exercises

A. Symbolize
1. There is a being that possibly fails _toexist.
2. Possibly there is something brown that could be blue.
3. Necessarily everything that is red is colored.
4. Everything that is red is necessarily colored.
5. Everything that is red is necessarily colored essentially.
6. Anything that is six feet tall might not have been six feet tall.
 EXISTENCE AND MODALITY 113

7. Everything that is essentially rational is visible.

8. Some things that are essentially rational are not essentially
embodied.
B. Prove:
9. (x) □ ((Fx&Gx):JFx)
10. (x)(y)( □ (Fx&Gy):JFx)
11. (x) □ ((Fx&Lx) :JFx)

C. Give Proofs:
12. □ (x)((Ax&Bx):J ~Cx)
(x) □Ax
(x) □ (Bx:JCx) :.(x)~◊Bx
13. (x)( □Ax:::, □Bx)
□ (x)(Bx:::, Cx)
:.(x)( □Ax:::, □ Cx)

14. (y) □ ((x)Ax:JBy)

:.□ (x)Ax:J(y) □By
15. (x)(y) □ (Gxy :::,Fxy)
(x)(y) □ (Fxy:::, Cy)
(x) □ (Cx ::,Bx)
(x)(y) □ Gxy :.(y) □By

D. Attempt to construct proofs of the Buridan and Converse

Buridan formulas, and observe which restrictions on our rules
prevent the proofs from going through.

4.15 Existence and Modality

Metaphysics abounds with puzzles about existence, possible ex-

istence, and necessary existence. Since the interpretation we have
adopted has interesting implications for some of these puzzles, it will
be worthwhile to take a brief look at a few here.
We have said earlier that any property a thing has in every pos-
sible world in which it exists is an essential property of the thing.
Paraphrasing that in the terms we have adopted, we would say that
something has a certain property essentially just in case the essence
of that thing is coexemplified with that property in every possible
world whose domain contains that essence. It is easy to see that,
whichever characterization we use, existence (assuming it is a prop-
erty) is an essential property of everything that exists. This may seem
highly paradoxical at first, since it seems well-nigh axiomatic that
114 QUANTIFIED MODAL LOGIC

many of the things that exist exist contingently. In order to remove

this air of paradox, we shall have to make some further distinctions.
Necessity is usually opposed to contingency, while essences or es·
sential properties are usually contrasted with accidents or acciden-
tal properties. An accident is a property a thing has but that it could
lack and still be what it is. Given this characterization, it seems plau-
sible to deny that existence is an accidental property of any thing
that has it. Yet the existence of a good many things, ourselves in-
cluded, seems to us contingent. Suppose then we call a being con·
tingent if it exists (i.e., its essence is exemplified) in some worlds but
not in others, while a being is necessary or has necessary existence
just in case it exists in every possible world. The air of paradox is
dispelled, or at least greatly diminished, when we see that we exist
essentially but not necessarily. You exist, and therefore you exist es-
sentially, but it does not follow that you exist necessarily.
We may also use the de re - de die to distinction to explain the dif ·
ference between necessary and essential existence. What was just
called necessary existence can be explained as the truth of a proposi-
tion asserting existence of the thing in question in every possible
world. Here necessity is predicated of a proposition ascribing exis-
tence to the thing. In the case of essential existence, however, the
property of existence is predicated of the thing in question, and the
thing is said to have this property essentially.
This in tum sheds some light on negative existential propositions,
propositions that deny the existence of an individual. The classical
paradox holds that it is impossible for negative singular existentials
to be true because our reference implies the existence of that whose
existence we are denying. This might seem to be a special problem
for actualists. By saying that my older brother does not exist, I ap·
pear to be asserting something about my older brother. But accord·
ing to the actualist we cannot predicate properties of things that do
not exist. Hence, my older brother has to exist for me to deny that
he exists, and so what I am trying to affirm implies its own falsehood.
We have just seen a way of dealing with this. Someone affirming
a negative existential may be denying the truth of a proposition or
she may be attributing a property, nonexistence, to something. Any-
one doing the latter is trying to affirm the impossible. There is no
possible world in which such a de re claim is true. However, the for-
mer sort of claim can truly be made in any world in which the essence
referred to by the name is not coexemplified with existence in that
world. It is foolishness to try to say of my older brother that he lacks
existence; that will always be necessarily false. But it is sensible, and
 THE PROBLEM OF 'TuANSWORLD IDENTITY 115

even true, to deny that my older brother exists, i.e., to deny that I
have an older brother.

4.16 The Problem of Transworld Identity

The system and interpretation of modal logic with which we have

been working holds that some individuals exist in different possible
worlds. Indeed, that feature enables us to distinguish a thing's acci-
dental from its essential properties. But we perhaps should pause
a moment to acknowledge that not all proponents of modal logic and
possible worlds have been happy with that idea. One of the well-
springs of the possible worlds tradition, Leibniz, apparently thought
that any given individual exists in one and only one possible world.
Indeed, he suggests that God organizes all the possible individuals
into mutually compatible (compossible, he says) groups and each in-
dividual fits into exactly one. In spite of this, however, Leibniz seems
to feel quite free in his correspondence with Arnauld to speak of Adam
in other possible worlds, although he will also allow that, strictly
speaking, this other Adam is actually a counterpart, and not the same
individual (or individual concept). 40 At any rate, Leibniz seems to
subscribe to the view that an individual is to be found in at most
one possible world. This has been explored more recently by David
Lewis. 41
What are the attractions of such a view? Leibniz apparently be-
lieves that all of a thing's properties are essential to it and that to
hold less was to impugn the omnipotence and omniscience of God
(Correspondence with Amauld, chap. IX). His arguments require a
more careful look at his entire metaphysics, and so we shall not pur-
sue them here.
There is another argument raised recently against the view that
an individual may exist in more than one possible world. The argu-
ment is the allegation that there are insuperable difficulties involved
in identifying individuals from one world to the next. 42 A common
way of generating the desired sense of bewilderment involves asking
us to consider two individuals and to imagine a series of possible
worlds in which these individuals exchange more and more proper-
ties. Consider Chisholm's example. We begin with Adam and Noah
in the actual world. Next we "move" to a possible world in which Adam
lives 950 years and Noah lives 930 years. Then we go to a possible
world in which Adam lives 950 years and fathers Shem, while Noah
lives 930 years and fathers Seth. If we continue this long enough,
116 QUANTIFIED MODAL Lome

the argument goes, we will no longer be able to tell who is Adam and
who is Noah. There do not seem to be any clear criteria for identify-
ing Adam and identifying Noah such that in any given possible world
we can tell which is which. So we are invited to surmise that perhaps
there are no individual essences of objects that would make trans-
world identification possible. Individual essences are supposed to in-
dividuate things, and they do not seem to be able to do that.
Initially it is important to note two things about this alleged dif-
ficulty. First of all, it should not be taken to imply that there are no
individual essences. It does not follow that a thing has no individual
essence. Rather, the argument suggests that all of a thing's proper-
ties are essential to it- a very implausible position. Secondly, the argu-
ment or puzzle generates an epistemic claim about our inability (in
some cases, anyway) to tell what individual essence a thing exempli-
fies or a claim about our inability to tell whether some thing x in
one possible world exemplifies the same individual essence as y in
some other possible world. But then it invites us to jump to the meta-
physical conclusion that there are no individual essences that can be
exemplified in more than one possible world. This conclusion does
not follow.
It is also important in this connection not to be misled by the "pic-
ture" we have been using. It is helpful to think instead in terms of
the official definitions and descriptions given earlier. Recall that possi-
ble worlds and individual essences are abstract objects. A possible
world is a maximally consistent state of affairs, and an individual
essence is a set of properties (or complex property) such that anything
that has (exemplifies) it, has (exemplifies) it essentially, and no other
thing can have (exemplify) it. Assuming that properties exist neces-
sarily, there is no question about the existence of individual essences,
and there is no question that they individuate (i.e.,whatever exempli-
fies one is a distinct individual). There is also no difficulty in the pos-
sibility that essences should be exemplified in more than one possi-
ble world.
This view leaves open questions about what properties might be
essential to various individuals and how we might be able to ascer-
tain whether a given individual has a given property essentially. But
this is not apt to confuse us about the nature of essences unless we
insist on using the metaphor of visually peering at two incompletely
known individuals in two possible worlds and supposing that we
should be able to tell whether they have the same essence.
 SUPERIORITY OF ACTUALISM 117

4.17 Superiority of Actualism

One insuperable difficulty of other modal systems we considered

was their inability to express the truths
There could have been an object distinct from every ac-
tual object,
and
There could have been more objects than there are,
without needing to postulate nonexistent objects or ascribing pro-
perties to objects in worlds in which they do not exist. The discus-
sion of the previous sections provides the necessary clue for giving
readings of these claims without allowing anti-actualist implications.
It will be true that there could have been an object distinct from
every actual object just in case there is a possible world in which some
essence is exemplified (would be exemplified if that world were ac-
tual) that is not exemplified in the actual world. The present inter-
pretation has the virtue of restoring to the original statement its
possibility. The companion claim, there could have been more objects
than there are, requires that there be more individual essences than
are actually exemplified and that an even larger number of them be
able to be coexemplified. Again, our interpretation maintains the
plausibility of this plausible claim.
Modal actualism thus provides an interpretation of possibility and
necessity that preserves and enhances the insights that tradition-
ally have been part of modal logic. It makes clearer the relationships
between de re and de dicto modality and does not quantify into de
dicto modal contexts. Actualism does not include some of the meta-
physical implausibilities of its near competitors and it provides some
illuminating insights into the relationship between existence and
modality.
 APPENDIX I

AXIOMATIZATION OF T, S4, & S5

Although C. I. Lewis was aware that his modal systems contained

the propositional calculus, he did not axiomatize his systems by add-
ing modal axioms to an axiomatization of the system of Principia
Mathematica. Godel appears to have been first to axiomatize a series
of modal systems by adding modal axioms to a complete basis for
the classical propositional calculus. Here we follow Godel.
We start with a complete axiomatization of propositional calcu-
lus, taken from Mendelson's Introduction to Mathematical Logic (Van
Nostrand, 1979), p. 31. The axioms are actually axiom schemata; the
letters represent any well-formed formula. Hence, any uniform re-
placement of letters in the schemata by well-formed formulas results
in an axiom. Given the usual conventions about the number of propo-
sitional variables available, each axiom schema represents an infinite
number of axioms. The propositional calculus base is as follows:
Axioms: 1. p::>(q::>p)
2. (p ::>(q ::>r) ::>((p ::>q) ::>(p ::>r))
3. ( ~q ::>~p) :>(( ~q ::>p) ::>q)
Rule: Rl. p, p::>ql :.q(modus ponens)

We obtain the system T by adding a definition, a rule, and two axioms:

T = Propositional Calculus (1-3 and Rl) plus
4. □p::>p
5. □ (p:>q):>( □p:> □ q)
Rule: R2. I- pf.•. I- □p

Rule R2 says that if a formula p is a theorem (i.e., an axiom or a con-

sequence of axioms), then □p is a theorem.
S4 comes by adding its so-called "characteristic formula" as an
axiom.
S4: T plus 6. □p ::> □□p

119
120 APPENDIX I
S5 comes by adding its "characteristic formula" to T.
S5: T plus 7. ~ □p ::) □ ~ □p
(7.) is the form Godel gave. It also appears as
7'. ◊p::) □ ◊p
and 7". ◊ □p ::) □p.
The form (7) is used in order to specify the axiom solely in terms that
are primitive, i.e., undefined, within the system.
In an important paper on modal logic, Saul Kripke uses this basis
for T as a set of sufficient conditions for calling a modal propositional
calculus normal. Non-normal systems are ones that do not satisfy
R2.
In an interesting paper that groups some modal logics into vari-
ous "families," entitled "New Foundations for Lewis Modal Systems,"
E. J. Lemmon defines a Lewis Modal System as follows:
A Lewis Modal System is a system that
(i) contains the full classical propositional calculus;
(ii) is contained in S5;
(iii) admits substitutability of tautologous equivalents;
(iv) possesses as theorems:
Fl. □ (p:)q):)(0p:)0q)
F2. □ (p&q)=( □p& □ q)
F3. (D(p:)q)&D(q:)r)):) □ (p:)r)
F4. □p::)~ □ ~p.

It is easy to see that T, S4, and S5 are Lewis Modal Systems in the
defined sense, even though T was not among the systems Lewis him-
self proposed.
 APPENDIX II

PROVING THE EQUIVALENCE OF

ALTERNATIVE FORMULATIONS
OFT, S4 and S5

We have made a number of references to alternative formulations

of the systems we have presented. We have presented each system
by means of natural deduction rules and, in Appendix I, by means
of axioms. How do we know that we have the system we want? And
how do we know that these presentations give the same system?
These systems are usually defined by certain axiomatizations of
them. The system Tis a certain set of theorems, derivable from vari-
ous specified sets of axioms. The axioms presented for each system
in Appendix I are not the original refining ones, but they are sets
that yield exactly the same theorems. That different axiomatizations
yield the same system can be shown quite straightforwardly.
Suppose we have two sets of axioms and rules, alleged to be axio-
matizations of the same system. We must show that whatever is a
theorem according to the one axiomatization is also a theorem ac-
cording to the other. We do this by showing that all the axioms of
the one formulation are provable as theorems in the other formula-
tion, and we show how the effect of any rule of the one formulation
is obtainable using the rules and axioms of the other. This is, in ef-
fect, to show how to turn a proof of a theorem from the one set of
axioms and rules into a proof from the other set of axioms and rules.
When this can be done, we conclude that the one system contains
the other.
The procedure is basically the same for proving an axiomatiza-
tion for a system equivalent to a set of natural deduction rules. We
show that the axioms can each be proven via the natural deduction
rules, and that the effect of any rules can also be obtained via the
natural deduction rules. This shows that the natural deduction sys-
tem contains the axiomatic system. Then we show that the effect of

121
122 APPENDIX II

each natural deduction rule can be obtained by means of the axioms

and rules of the axiomatic system. Once we show this, we can con-
clude that the axiomatic system contains the natural deduction sys-
tem. If each contains the other, they are equivalent.
We illustrate this with the modal system T. We will assume here
that our natural deduction formulation of the propositional calculus
given in chapter 1 and the axiomatization of it (Al-A3 and RI of Ap-
pendix I) are equivalent. We complete the proof for the modal axioms
and rules.

A. Proof that any modal inference obtainable in the natural deduc-

tion formulation of T can be obtained from the axioms of T:
Natural deduction rule Derivation from axioms of T
nee elim
I. □p given 1. □p given
2. :.p nee elim 2. □p:)p axiom A4
3. :.p MP
nee intro (with T-reit)
1. □p given 1. □p given
2. □ (p :)q) given 2. □ (p :)q) given
3. □ p T-reit 3. □ (p:)q):)(0p:)0q) A5
4. (p:)q) T-reit 4. □p:) □ q MP
5. q MP 5. □ q MP
6. □ q nee intro

B. Proof that any modal formula that can be shown to be a theorem

from the axioms of T can also be derived from the natural deduc-
tion rules:
Derivation from
Axiom natural deduction rules
A4. □p:)p assumption
nee elim
impl intro
□( :)q)

1' □ ~~q
p

I ~q
□p:) □ q
□ (p :)q) :)(Op:) □ q)
 APPENDIX II 123

R2 If I- p then I- Op If I- p, then there is a PC proof of p

from Al-A3 and by our assumption
there is an ND proof of p (which uses
no premises). I.e.

a nee intro subproof: 0 h

Th obtain Op, simply put this within

p
Op
Since each formulation contains the other, they are equivalent.
 APPENDIX III:
AN EFFECTIVE DECISION PROCEDURE

One of the disadvantages of the natural deduction formulations

of logical systems, such as the one we have given, is that they do
not provide an effective procedure for determining theoremhood. An
effective procedure is a mechanical technique for determining in a
finite number of steps whether or not a formula is a theorem. We
know that if we can derive a formula with our rules it is valid, but
if we find ourselves unable to come up with a proof on some occasion,
we do not know whether that is due to a lack of ingenuity on our
part or to the invalidity of the formula. Furthermore, our rules en-
able us to go in circles; we could just repeat lines or sequence of lines
ad infinitum and never get to the end of a proof that could be fin-
ished. We are not required to use rules in a fixed order and not to
repeat them indefinitely.
There are mechanical proof and disproof procedures for modal
propositional logics, just as there are for classical propositional logic.
The easiest ones to use are based on a technique developed by E. W.
Beth, exploited by Kripke and Jeffrey and Zeman (q.v.).Procedures
of this type for modal logic are developed in detail by Slaght, "Modal
'Iree Constructions," and Davidson, Jackson, and Parzetter, "Modal
'Irees for T and S5," in the Notre Dame Journal of Formal Logic
Volume 18, No. 4, October, 1977, pp. 517-526 and pp. 602-605,
respectively.
Tableau techniques are given for T, S4, and S5 and proven to be
decision procedures in Zeman, J. Jay, Modal Logic, The Lewis-Modal
Systems. Oxford: Clarendon Press, 1973, chaps. 9 and 14.

125
 NOTES

1. For an excellent article presenting some central questions and main

arguments, see Richard Cartwright's article "Propositions," in R. J. Butler,
ed., Analytical Philosophy, First Series (Blackwell, 1962), pp. 81-103.
2. See Augustine's On the Free Choice of the Wilt book III, chap. 3, and
Boethius, The Consolation of Philosophy, book V, prose 6.
3. See Kretzmann, Kenny, and Pinborg, eds., The Cambridge History
of Later Medieval Philosophy (Cambridge, 1982), p. 343. The translation is
"loose" in the sense that I have rendered the modalities as applying only to
propositions.
4. See B. Mates, Stoic Logic (University of California, 1953), chap. IV,
§1.
5. See Anderson and Belnap, Entailment (Princeton, 1975).
6. E.g., by Lemmon in "New Foundations for Lewis Modal Systems,"
The Journal of Symbolic Logic, vol. 24 (1959), 1-14.
7. See Purtill, Logic for Philosophers (Harper & Row, 1971), chap. 7.3,
and Purtill, "Four Valued Thbles for Modal Logic," Notre Dame Journal of
Formal Logic, vol. 11, no. 4, (October, 1970).
8. Even Anderson and Belnap find the implicational fragment of S4 ac-
ceptable as an account of "implies," although they do not accept the system
as a whole.
9. This is a slight modification of a definition given by Perry, "Modali-
ties in the Survey System of Strict Implication," The Journal of Symbolic
Logic, vol. 4 (1939), 144, reported by Hughes and Cresswell, An Introduction
to Modal Logic (Methuen, 1968), p. 50.
10. See Hughes and Cresswell, Introduction to Modal Logic (Methuen,
1968), p. 58, n. 37.
11. A. N. Prior, Time and Modality (Oxford, 1957), p. 49.
12. Ibid., and Prior and Fine, Worlds, Times, and Selves (Duckworth,
1977), pp. 102-3.
13. Lemmon, "Is There One Correct System of Modal Logic?" The Aris-
totelian Society, Supp. Vol XXXIII, 1959, pp. 23-40. I draw heavily on Lem-
mon's article in what follows.
14. Ibid., p. 24.
15. Lewis and Langford, Symbolic Logic (Dover, 1959), p. 496.

127
128 NOTES TO PAGES 66-97

16. See his Time and Modality (Oxford, 1957), Past, Present, and Future
(Oxford, 1967), and Papers on Time and Tense (Oxford, 1968).
17. Hintikka, Knowledge and Belief (Cornell, 1962).
18. Lemmon mentions several more in the paper cited in note 13.
19. Those interested in pursuing this may consult Leonard Linsky, "'l\vo
Concepts of Quantification," Nous, vol. 6 (1972), 224-39. [Reprinted as an
appendix to Linsky, Names and Descriptions (Chicago, 1977).] See also Ruth
Barcan Marcus, "Interpreting Quantification," Inquiry, vol. 5 (1962), 252-59,
and Nuel Belnap and J. Michael Dunn, "The Substitutional Interpretation
of Quantifiers" Nous, vol. 2 (1968), 177-85; Susan Haack, Philosophy of Logics
(Cambridge, 1978), chap. 4.
20. This technical term was defined a few pages earlier.
21. Quoted from Aristotle, The Works of Aristotle, ed. W. D. Ross, vol. I,
(Oxford, 1928), 166a pp. 24-30.
22. SCG, I, 67. 'Ir. Anton Pegis (Notre Dame, 1975). Note that the "fal-
lacy of composition and division" here mentioned is that of confusing the
composite and the divided senses of some expression.
23. Several of these topics will arise again in "Some Objections to Quan-
tified Modal Logic."
24. 'Divisible' itself appears to be a modal term. Here we will understand
it to mean 'having at least two parts.'
25. Quine, "Three Grades of Modal Involvement," The Ways of Paradox
(Random House, 1966), p. 174.
26. Quine, "Reference and Modality," in From a Logical Point of View, sec-
ond edition (Harper and Row, 1961), p. 143.
27. Ibid., p. 144.
28. We will not pursue any further Quine's objections and possible re-
sponses. Anyone interested in pursuing the issue in detail should consult
Leonard Linsky's anthology, Reference and Modality (Oxford, 1971), and the
appendix to Alvin Plantinga's The Nature of Necessity (Oxford, 1974), where
Quine's objection is discussed in detail.
29. And accidental properties as well. An accidental property is a prop-
erty an individual has in the actual world but lacks in some other possible
world in which the individual exists.
30. "Counterpart Theory and Quantified Modal Logic," Journal of
Philosophy, vol. 65 (March 1968), pp. 113-26.
31. For more on fictional characters, see N. Wolterstorff's Works and
Worlds of Art (Oxford, 1980), pp. 134-58.
32. What I give here is equivalent to the form ◊(3x)Ux:J(3x)◊ax
which was first discussed by Ruth Barcan, now Ruth Barcan Marcus, in 1946.
33. So named by Alvin Plantinga, The Nature of Necessity (Oxford, 1974),
p. 58, after Jean Buridan, who denied the truth of the natural interpretation
of it.
34. "Quantification and Modality in S5," Journal of Symbolic Logic, vol.
21, no. 1 (March 1958), 60-62.
 NOTES TO PAGES 103-115 129

35. This objection is discussed in greater detail in chap. 7 of Plantinga's

The Nature of Necessity and also in his ''Actualism and Possible Worlds,"
Theoria, vol. 42 (1976), 139-60.
36. This section is based on Alvin Plantinga's ''Actualism and Possible
Worlds," Theoria, vol. 42 (1976), 139-60. The name "actualism" is from Rob-
ert M. Adams, ''Theories of Actuality," Nous, vol. 8 (1974), 211-31. This "ver-
sion" is an interpretation of quantified modal logic or an applied semantics.
37. These definitions come from Plantinga's The Nature of Necessity,
chap. 5, and Thomas Jager's ''An Actualistic Semantics for Quantified Modal
Logic," Notre Dame Journal of Formal Logic, vol. 23, no. 3 (July 1982), 336-
37.
38. At any rate, if one takes a realist view of properties, this will follow.
Actualism in possible worlds obviously comports very well with metaphysi-
cal realism, even if it does not strictly require it.
39. ''An Actualistic Semantics for Quantified Modal Logic," Notre Dame
Journal of Formal Logic, vol. 23 (1982), 335-49.
40. Cf. The Leibniz-Arnauld Correspondence, ed. & tr., H. T. Mason (Man-
chester University Press and Barnes and Noble, 1967), pp. 15-16.
41. "Counterpart Theory and Quantified Modal Logic," Journal of Phi-
losophy, vol. 65 (1968), pp. 113-26.
42. For a perspicuous statement of this problem, see Chisholm's "Iden-
tity through Possible Worlds; Some Questions," Nous, vol. 1 (1967), 1-8. Re-
printed in Loux, The Possible and the Actual (Cornell, 1979).
 BIBLIOGRAPHY

Ackrill, J. L. Aristotle's Categories and De Interpretatione. Clarendon Aris-

totle Series, ed. J. L. Ackrill. Oxford: Clarendon Press, 1963.
Adams, Robert M. "Theories of Actuality," Nous, vol. 8 (1974), 211-31.
Anderson, A. R., and Belnap, Nuel D., Jr. Entailment: The Logic of Rele-
vance and Necessity. Princeton: Princeton University Press, 1975.
Aristotle, De Sophisticis Elenchis, in The Works of Aristotle Translated Into
English, vol. I, W. D. Ross. Oxford: Clarendon Press, 1928. Newly reis-
sued in The Complete Works of Aristotle, ed. Jonathan Barnes. Prince-
ton: Princeton University Press, 1984.
Augustine. On Free Wil~ in Augustine: Earlier Writings. 'Ir. John H. S. Bur-
leigh. Philadelphia: The Westminster Press, 1953.
Barcan, Ruth. "A Functional Calculus of First Order Based on Strict Implica-
tion," Journal of Symbolic Logic, vol. 11, (1946), 1-16.
Belnap, Nuel D., Jr., and J. Michael Dunn, "The Substitutional Interpreta-
tion of Quantifiers," Nous, vol. 2, (1968), 177-85.
Boethius, A. M. S., The Consolation of Philosophy. 'Ir. Richard Green. The
Library of Liberal Arts. Indianapolis: The Bobbs Merrill Company, Inc.,
1962.
Cartwright, Richard. "Propositions," in Analytical Philosophy, First Series.
Ed. R. J. Butler. Oxford: Basil Blackwell, 1962. 81-103.
Chisholm, Roderick. "Identity through Possible Worlds: Some Questions,"
Nous, vol. 1 (1967), 1-8. Reprinted in Loux, The Possible and the Actual
(Cornell, 1979).
Curley, E. M.
1972 "Lewis and Entailment," Philosophical Studies. vol. 23, 198-204.
1975 ''The Development of Lewis' Theory of Strict Implication," Notre
Dame Journal of Formal Logic. vol. 16, no. 4 (Oct. 1975), 517-27.
Davidson, B., F. C. Jackson, and R. Pargetter. "Modal 'Irees for T and S5,"
Notre Dame Journal of Formal Logic, vol. 18, no. 4 (October 1977), 602-07.
Fitch, Frederick B. Symbolic Logic. New York: Ronald Press, 1952.
Haack, Susan. Philosophy of Logics. Cambridge: Cambridge University Press,
1978.
Hintikka, Jaakko. Knowledge and Belief. Ithaca: Cornell University Press,
1962.

131
132 BIBLIOGRAPHY

Hughes, G. E., and M. J. Cresswell. An Introduction to Modal Logic. Lon-

don: Methuen, 1968.
Jager, Thomas. "An Actualistic Semantics for Quantified Modal Logic." Notre
Dame Journal of Formal Logic, vol. 23, no. 3 (July 1982), 335-49.
Kretzmann, N., A. Kenny, and J. Pinborg, eds. The Cambridge History of
Later Medieval Phil,osophy. Cambridge: Cambridge University Press, 1982.
Kripke, Saul.
1959 •~ Completeness Theorem in Modal Logic," The Journal of Sym-
bolic Logic. vol. 24, no. 1 (March 1959), 1-14.
1963a "Semantical Considerations on Modal Logic," Acta Philosophica
Fennica. Fasc. XVI, 83-94. Reprinted in Leonard Linsky, ed. Reference
and Modality. Oxford: Oxford University Press, 1971.
1963b "Semantical Analysis of Modal Logic I: Normal Modal Proposi-
tional Calculi," Zeitschrift fur mathematische Logik und Grundlagen
der Mathematik. vol. 9, 63-96.
Leibniz, G. W.
1686 Discourse on Metaphysics, in Philosophical Papers and Letters.
Second edition. Ed. Leroy E. Loemker. Dordrecht: Reidel Publishing
Co., 1969.
1686 Leibniz-Arnauld Correspondence. 'fr. H. T. Mason. New York:
Barnes & Noble, Inc., 1967.
1714 Monadology, in Philosophical Papers and Letters. Second edition.
Ed. Leroy E. Loemker. Dordrecht: Reidel Publishing Co., 1969.
Lemmon, E. J.
1957 "New Foundations for Lewis Modal Systems," The Journal of Sym-
bolic Logic. vol. 22 (June 1957), 176-86.
1959 "Is There Only One Correct System of Modal Logic?" The Aristo-
telian Society, Supplementary Volume XXXIII, 1959. London: Harri-
son and Son, Ltd., 23-40.
Lewis, C. I., and C.H. Langford. Symbolic Logic. Second Edition. New York:
Dover Publications, 1959.
Lewis, David. "Counterpart Theory and Quantified Modal Logic." Journal
of Philosophy. vol. 65 (March 1968), 113-28.
Linsky, Leonard.
1971 Ed. Reference and Modality. Oxford: Oxford University Press.
1972 "'l\vo Concepts of Quantification." Nous. vol. 6, 224-39.
Reprinted as an appendix to Linsky, Names and Descriptions, Chicago:
University of Chicago Press, 1977.
Loux, Michael, ed. The Possible and the Actual. Ithaca: Cornell University
Press, 1979.
Marcus, Ruth Barcan (see also Barcan, Ruth). "Interpreting Quantification."
Inquiry, vol. 5 (1962), 252-59.
Mates, Benson. Stoic Logic. University of California Publications in Phi-
losophy, vol. 26. Berkeley and Los Angeles: University of California Press,
1953.
 BIBLIOGRAPHY 133

Mendelson, Elliot. Introduction to Mathematical Logic. Second edition. New

York: Van Nostrand, 1979.
Plantinga, Alvin.
1974 The Nature of Necessity. Oxford: Clarendon Press.
1976 "Actualism and Possible Worlds." Theoria, vol. 42, 139-60.
Reprinted in M. Loux, The Possible and the Actuat Ithaca: Cornell
University Press, 1979.
Prior, Arthur Norman.
1955 Formal Logic. Oxford: Clarendon Press.
1956 "Modality and Quantification in S5," The Journal of Symbolic Logic,
vol. 21, 60-62.
1957 Time and Modality. Oxford: Clarendon Press.
1967 Past, Present, and Future. Oxford: Clarendon Press.
Prior, Arthur Norman, and Kit Fine. Worlds, Times, and Selves. Amherst:
University of Massachusetts Press and London: Duckworth, 1977.
Purtill, Richard.
1970 "Four-Valued Tables for Modal Logic," Notre Dame Journal of For-
mal Logic, vol. 11, no. 4 (October 1970).
1971 Logic for Philosophers. New York: Harper and Row.
Quine, Willard Van Orman.
1961 From a Logical Point of View. Second edition. Harper 'lbrchbook
Edition. New York: Harper and Row.
1966 The Ways of Paradox and Other Essays. New York: Random House.
Slaght, Ralph L. "Modal Tree Constructions," Notre Dame Journal of Formal
Logic. vol. 18, no. 4 (October 1977), 519-26.
Thomas Aquinas, St. Summa Contra Gentiles. 'fr. Anton C. Pegis. Notre
Dame: University of Notre Dame Press, 1975. Reprint of On the Truth
of the Catholic Faith. Garden City, N.Y.: Hanover House, 1955.
Whitehead, Alfred North, and Bertrand Russell. Principia Mathematica to
*56. Cambridge: Cambridge University Press, 1962.
Zeman, J. Jay. Modal Logic: The Lewis-Modal Systems. Oxford: Clarendon
Press, 1973.