O X F O R D L O G IC G U ID E S : 25

General Editors

DO V G A B B A Y
ANGUS M A C IN TY R E
D A N A SCOTT
 O XFO R D LOGIC GUIDES

1. Jane Bridge: B egin n in g m od el th eo ry : th e com p leten ess th eorem and som e

co n seq u en ces
2. Michael Dummett: E lem en ts o f intuitionism
3. A. S. Troelstra: C h o ice se q u e n c es : a ch a p ter o fin tu itio n istic m athem atics
4. J. L. Bell: B o o lea n -v a lu ed m od els and in d ep en d en ce p r o o fs in s e t th eory
( 1st edition)
5. Krister Segerberg: C la ssica l p ro p o sitio n a l o p era to rs : an e x e r c is e in th e
fo u n d a tio n s o f lo g ic
6. G. C. Smith: The B o o le -D e M orga n co rresp o n d en ce 1 8 4 2 -1 8 6 4
7. Alec Fisher: F orm a l n u m ber th eo ry and com p u ta b ility : a w ork b o o k
8. Anand Pillay: A n in trod u ction to sta bility th eo ry
9. Η. E. Rose: S u brecu rsion : fu n ctio n s and h iera rch ies
10. Michael Hallett: C an torian s e t th eo ry an d lim itation o f size
11. R. Mansfield and G. Weitkamp: R ecu rsiv e a sp ects o f d escrip tiv e s e t th eory
12. J. L. Bell: B oo lea n -v a lu ed m od els and in d ep en d en ce p r o o fs in s e t th eo ry
(2nd edition)
13. Melvin Fitting: C om pu tability th eo ry : sem a n tics an d lo g ic program m in g
14. J. L. Bell: T o p o ses an d lo c a l s e t th eo ries: an in trod u ction
15. Richard Kaye: M od els o fP e a n o a rith m etic
16. J. Chapman and F. Rowbottom: R ela tive ca te g o r y th eo ry an d g eo m etric
m orphism s
17. S. Shapiro: F ou n d ation s w ith ou tfou n d a tion alism
18. J. P. Cleave: A study o f lo g ics
19. R. M. Smullyan: G od eT s in com p leten ess th eorem s
20. T. E. Forster: S et th eo ry w ith a u n iversal s e t
21. C. McLarty: E lem en tary c a te g o r ie s , elem en ta ry to p o ses
22. R. M. Smullyan: R ecu rsion th eo ry f o r m etam athem atics
23. Peter Clote and Jan Krajicek: A rith m etic , p r o o f th eoryt and com pu tation a l
co m p lex ity
24. A. Tarski and J. Tarski: In trod u ction to lo g ic an d to th e m eth od olog y o f
d ed u ctive s c ie n ce s (4th edition)
25 . Grzegorz Malinowski: M a n y-va lu ed lo g ics
Many-Valued Logics
Grzegorz Malinowski
University of-Lodz

C L A R E N D O N PRESS · O X F O R D
1993
 Oxford University Press, Walton Street, O xford 0 X 2 6D P

Oxford New York Toronto

D elhi Bombay Calcutta Madras Karachi
Kuala Lumpur Singapore Hong Kong Tokyo
N airobi D a res Salaam Cape Town
M elbourne Auckland Madrid
and associated com panies in
Berlin Ibadan

Oxford is a trade mark o f Oxford University Press

Published in the United States by

Oxford University P ress Inc., New York

© G rzegorz Malinowski, 1993

A ll rights reserved. No part o f this publication may be reproduced,

stored in a retrieval system, o r transmitted, in any form o r by any means,
without the prior perm ission in writing o f Oxford University Press.
Within the UK, exceptions are allowed in respect o f any fa ir dealing fo r the
purpose o f research o r private study, o r criticism o r review, as perm itted
under the Copyright, D esigns and Patents Act, 1988, o r in the case o f
reprographic reproduction in accordance with the terms o f the licences
issued by the Copyright Licensing Agency. Enquiries concerning
reproduction outside these terms and in other countries should be
sent to the Rights Department, Oxford University Press,
at the address above.

This book is sold subject to the condition that it shall not, by way
o f trade o r otherwise, be lent, re-sold', hired out, o r otherwise circulated
without the publisher’s p rior consent in any form o f binding o r cover
other than that in which it is published and without a similar condition
including this condition being imposed on the subsequent purchaser.

A catalogue record fo r this book is available from the British Library

Library o f Congress Cataloging in Publication Data

Malinowski, Grzegorz.
M any-valued logics / G rzegorz Malinowski,
p. cm. — (O xford logic guides; 25)
Includes bibliographical references and indexes.
1. M any-valued logic. I. Title. II. Series.
QA9.45.M 35 1993
511.3— dc20 93-36413 CIP
ISBN0 -1 9 -8 5 3 7 8 7 -5

Typeset by the author

Printed in G reat Britain on acid-free paper by
Bookcraft (Bath) Ltd, M idsom er Norton
Contents
Introduction 1

1. The classical logic 5

1.1 Truth-tables 5
1.2 Propositional tautologies 7
1.3 Functional completeness 8
1.4 Axiomatization o f CPC 9
1.5 Predicate calculus 11
1.6 Algebraization 12

2. The third logical value o f Lukasiewicz 16

2.1 Lukasiewicz and the Lvov-Warsaw school 16

2.2 Three-valued logic 17
2.3 Modality and three-valuedness 19
2.4 Interpretation difficulties 21

3. Logic algebras and matrices 24

3.1 Language and logic algebras 24

3.2 Functional completeness o f finite algebras 26
3.3 Logical matrices 28

4. Many-valuedness 30

4.1 Tw o criteria 30
4.2 Structurality and many-valuedness 32
4.3 Finiteness and deduction 33

5. Lukasiewicz logics 36

5.1 Fundamental properties 36

5.2 Definability in Lukasiewicz matrices 38
5.3 Axiomatization 39
5.4 Algebraic interpretations 40

6. Post logics 44

6.1 Post matrices 44

VI Contents

6.2 Interpretation 46
6.3 Algebraic form o f Post logics 47
6.4 Axiomatization o f functionally complete systems
o f n- valued logic 48

7. Three-valuedness o f Kleene and Bochvar 51

7.1 Logic o f indeterminacy 51

7.2 Bochvar logic 54
7.3 Partial logics 56

8. Standard properties o f many-valued constructions 60

8.1 Standard conditions 61

8.2 Axiomatization 62
8.3 Standard matrix consequence 63

9. Probability and many-valuedness 66

9.1 Logical probability 66

9.2 Operationalistic conception o f subjective probability 68

10. Classical characterization o f many-valued logics 72

10.1 Suszko’s thesis 72

10.2 Scott’s method 74
10.3 Urquhart’s interpretation 77

11. Quantifiers in many-valued logic 79

11.1 Ordinary predicate calculi 79

11.2 Set theory and many-valued logic 81
11.3 Generalized quantifiers 83

12. Intuitionism and the modal logics o f Lewis 87

12.1 Intuitionistic logic 87

12.2 Modal logics S4 and S5 91
12.3 Remarks on quantification 95

13. Fuzzy sets and Zadeh logic 98

13.1 Fuzzy sets and logics o f imprecise concepts 98

13.2 Fuzzy logic 101

14. Applications and significance o f the topic 105

14.1 Independence o f axioms 105

14.2 Formalization o f intensional functions 106
 Contents vii

14.3 Many-valued algebras and switching theory 109

14.4 Many-valuedness in computer science 111

Bibliography 113

Author index 125

Subject index 127

Introduction
The vital part o f the studies o f logic seeks to determine structural cri­
teria for propositional validity and deals with formal inference relations.
A suitable starting point for any analysis o f these problems consists in
the selection o f a set o f propositions from among all grammatically well-
formed sentences, the members o f which satisfy some specified syntactical
and semantical conditions.
The assumption stating that to every proposition it may be ascribed
exactly one o f the two logical values, tru th or fa ls ity , called the p r in c ip le
o f b iv a le n c e , constitutes the basis o f classical logic. It determines both
the subject matter and the scope o f applicability o f the logic, the main
systems o f which are the classical propositional calculus (C P C ) and the
(first-order) predicate calculus (quantifier calculus). CPC is a theory o f all
truth-functional propositional connectives, i.e. sentence-argument proposi­
tional functors having the property that the logical value o f any complex
sentence formed with their use is determined uniquely by the logical values
o f its components. The p red ica te ca lcu lu s is formed by introducing to the
language system, with its semantics adequately extended, the symbols o f
name-argument propositional functors representing the names o f proper­
ties and relations and name quantifiers. It renders possible the profound
analysis o f propositions within the principle o f bivalence paradigm.
The most natural and straightforward step beyond the two-valued logic
is the introduction o f more logical values, rejecting simultaneously the prin­
ciple o f bivalence. The indirect way, on the other hand, consists either in the
revision o f the ‘bunch’ o f sentence connectives (then non-truth-functional
connectives are introduced into the language o f propositional calculus) or,
after having questioned some classical laws, in modifying the connectives
characterized by them. The multiple-valued truth-tables constitute the ba­
sis o f the first method, whereas in the other case they are procured as tools
for the procedures o f decidability o f logical theorems.
The roots o f many-valued logics can be traced back to Aristotle (4th cen­
tury BC) who considered, within the modal framework, fu tu r e co n tin g en ts
sentences. In Ch. IX o f his treatise D e In te r p r e ta tio n e Aristotle provides
the time-honoured sentence-example representing this category: ‘There
will be a sea-battle tomorrow’ . The philosopher from Stagira emphasizes
the fact that fu tu r e co n tin g en ts are neither actually true nor actually false,
which suggests the existence o f a ‘third’ logical status o f propositions. The
2 Introduction

prehistory o f many-valued logics falls in the Middle Ages and was made
by Duns Scott, William of Ockham and Peter de Rivo (Louvain). At the
turn o f the 19th century some attempts to create non-classical logical con­
structions, three-valued mainly, appeared: in 1897 Hugh M acColl’s inves­
tigations concentrated on the so-called ‘ three-dimensional logic’ , Charles
S. Peirce (1839-1914) worked on ‘ trychotomic mathematics’ founded on
the ‘ triadic logic’ , whereas Nicolai A. Vasil’ev presented a system o f ‘ imag­
inary non-Aristotelian logic’ in which propositions may be ‘affirmative’ ,
‘ negative’ or ‘indifferent’ .
The final, thoroughly successful, formulation o f many-valued logical
constructions was proposed as a result o f the truth-table method applied to
the classical logic by Frege (1879), Peirce (1885) and others; and then the
logical matrices method (Lukasiewicz, Post). The ‘ era o f many-valuedness’
was finally inaugurated in 1920 by Lukasiewicz and Post. After many
years o f investigation Lukasiewicz (1920) enriched the set o f the classical
logical values with an intermediate value and laid down the principles of
a three-valued propositional calculus. Post (1920), on the other hand, de­
fined (finite) many-valued ‘ logical algebras’ . It is worth mentioning that in
the 1930s some non-classical logical constructions appeared formalizing in-
tensional (non-truth-functional) sentence connectives by means o f axioms,
i.e. Lewis modal logics and intuitionistic logic which codifies the principles
o f some significant trend in the philosophy o f mathematics initiated by
Brouwer in 1907.
This book will present an elementary exposition o f the topics connected
with many-valued logics. Our discussion will focus on the constructions
being ‘ many-valued’ at their origin, i.e. on those obtained by the use o f
matrices. Therefore, the matrix method has been chosen as the most suit­
able way o f presenting the subject. The opening Chapter 1 , ‘ The classical
logic’ , contains background material and is devoted to the fundamentals of
the classical truth-functional logic. First, it will serve as an introduction
for those who are not acquainted with elementary logic at all. Its second
role is to collect and organize the material in a way which makes it eas­
ier to compare and distinguish between classical and many-valued logic.
Chapter j2, ‘ The third logical value o f Lukasiewicz’ , is an overview o f the
origin and basic properties of the first three-valued system o f propositional
logic. There, one may also find a discussion on the problem o f intuitive
interpretation o f Lukasiewicz’s third value and logic.
The next two chapters have a more general character. In Chapter 5,
‘ Logic algebras and matrices’ , interpretation structures o f propositional
languages in the Fregean framework are derived. A discussion o f functional
completeness, the algebraic property which warrants the biggest expressive
power o f the corresponding bunch of connectives, then follows. In the
end, matrices, i.e. algebras with sets o f designated elements and their basic
properties, are constructed. Chapter 4 , ;Many-valuedness’ , aims at showing
 Introduction 3

the complexity o f (the answer to) the question on how one can and must
understand the quality we are interested in. It also collects together some
established methods o f propositional calculi for the same purpose.
Chapter 5, ‘Lukasiewicz logics’ , is an account o f the work o f Lukasiewicz
and his followers, concerning the whole family of Lukasiewicz systems, i.e.
based on arbitrary finite and infinite Lukasiewicz matrices. Chapter 6,
‘ Post logics’ , is a similar account o f the investigation concerning finitely
many-valued functionally complete propositional logics introduced by E.
Post. And, subsequently, in Chapter 7, ‘ Three-valuedness o f Kleene and
Bochvar’ , the fundamentals of the two important epistemically motivated
systems o f (three-valued) logic are discussed. This part o f the book ends
with an overall exposition of constructions motivated by partiality in as­
sociating truth-values to sentences, i.e. supervaluations and partial logic,
which are semantically an extension o f Kleene (and, to an extent, Bochvar)
logic.
Chapter 8 , ‘ Standard properties o f many-valued constructions’ , is based
on the very important work o f Rosser and Turquette on the problem o f ax-
iomatization o f systems o f many-valued logic based on a finite set o f values.
The standard properties mentioned in the title are those o f popular con­
nectives warranting the workability o f a general method o f axiomatization
o f many-valued systems. In Section 8.3 similar properties are established
for matrix consequence operations.
Chapter 9, ‘ Probability and many-valuedness’ , is a discussion o f the
relation between many-valued logic and probability. Though similar on
the surface, each o f the two formalisms has its own merits. And, taken as
they stand, they are different. Our aim is to discuss relevant matters. The
second section describes an ingenious Giles’ solution bridging the so-called
subjective probability and infinite-valued Lukasiewicz logic.
Three kinds o f descriptions o f many-valued constructions in terms o f
zero-one valuations and the interpretations associated with them are the
topic o f interest in Chapter 10, ‘ Classical characterization o f many-valued
logic’ . In the following Chapter 11, ‘ Quantifiers in many-valued logic’ ,
there is not only an account o f the properties of quantification in the area o f
consideration but also a discussion of an important, relevant topic: namely,
Section 11.2 concentrates on set theory or, more precisely, on the question
o f the foundation o f set theory on many-valued logic.
Chapter 12, ‘ Intuitionism and modal logic’ , is, from one point o f view,
a concise introduction to the two axiomatic constructions o f non-classical
logic. The great significance o f these constructions in the history o f logic
and their exceptional properties both explain our interest in the topic.
Another important reason for considering them here is to have an insight
into logics which, though motivated otherwise, ultimately are (infinitely)
many-valued.
In Chapter 13, ‘ Fuzzy sets and Zadeh logic’ , one o f the most interesting
4 Introduction

but most controversial conceptions inspired by logical many-valuedness,

‘fuzziness’ , and its application to the formal analysis of everyday reasoning
and to other purposes, are discussed. Finally, Chapter 14 , ‘ Application
and significance o f the topic’ , offers some examples o f the applications o f
many-valued matrices to philosophical logic and shows the role o f many-
valuedness in such very practical areas as switching theory and computer
science.
The text is followed by a selective Bibliography covering the subjects
discussed. Monographs and expository papers on several issues concerning
many-valuedness are marked by an asterisk *. A full bibliography o f many­
valued logic up to 1965 can be found in Rescher (1969) and its continuation
for 1966-1974 in W olf (1977). An important and useful source covering the
period o f relevant investigation up to 1985 is in Rautenberg (1987).
1. The classical logic

This chapter contains background material. It collects and organizes the

fundamentals o f standard logic in a way which will make it easier to com­
pare and distinguish between the classical and many-valued logics.
The classical propositional calculus (CPC) is a basic system o f the two­
valued logic. The standard line o f approach here is a theory o f the following
sentence connectives taken from everyday reasoning:

negation —1 (-•a means ‘not a ’ )

implication —► (a —►β means ‘ if a , then /?’)
disjunction V ( a V ]8 means ‘ a or /?’ )
conjunction Λ (a Λ β means ‘ a and /?’ )
equivalence = (a = β means ‘ a if and only if /?’)

The set o f formulae o f the classical propositional calculus, For , contains de-
numerably many propositional variables p, <f, r , . . . ,p i, ςχ, r*i,. . . represent­
ing simple (atomic) sentences and all those expressions which are formed
by means o f propositional variables, connectives and brackets according to
the rule:

(For) If a,/3 G For, then “ «(a), (a ) —> (/?),

(a ) V (β), (a) Λ (β), (a ) = (β) € For.

The formulae o f CPC stand for schemes o f utterances in which expressions

corresponding to the mentioned sentence connectives are used. Brackets
appearing in formulae serve as punctuation, like in mathematics. Unless it
causes ambiguity, they can be omitted. Owing to the intuitive, commonly
accepted convention -i is treated as the most binding connective, the others
being ordered in terms o f decreasing strength as follows: V and Λ (to the
same degree), — = . For instance, using this convention, the formula
(r Λ (-»p)) —> ((q V (-»r)) A(q = (p V r ))) may be simplified to the following
one: r A -ip —> (qV -»r) Λ (q = p V r).

1.1 Truth-tables
The simplest and most readily interpretable formalization o f CPC is pro­
vided by truth-tables. Truth-tables determine the role and the meaning
6 1. The classical logic

o f the propositional connectives assigning to them functions whose argu­

ments and values are 0 (zero) and 1 (one), denoting respectively falsity
and truth. The truth-tables for the negation one-argument function and
the two-argument functions in the remaining cases are:

a β a —►β a V /3 a Α β Ol — β
a “O' 0 0 1 0 0 1
0 1 0 1 1 1 0 0
1 0 1 0 0 1 0 0
1 1 1 1 1 1

The crucial point about the above tables is that they specify the intu­
itive features o f the connectives in the following way:

(-i) Negation o f a (~oc) is a true sentence iff a is a false sentence

(—►) Implication o f a and β (a —> β ) is a false sentence iff a is a true
sentence and β is a false sentence
(V ) Disjunction o f a ,/? (a V β ) is a false sentence iff a and β are both
false sentences
(A ) Conjunction o f α,/? (α Λ β ) is a true sentence iff a and β are both
true sentences
(= ) Equivalence o f a ,/? (a = β ) is a true sentence iff a and β have the
same logical value.

Each formula of CPC may then be seen as a function o f (many, in

general) variables ranging over the set { 0, 1}, and hence the truth-table
method can be extended onto all the subsets Z C For . A function v :
Z —►{ 0 ,1 } will be referred to as a logical valuation o f the set Z o f formulae
whenever for any α,/3 £ Z

v ( - 1a ) = 1 iff v (a ) = 0
v (a —►β ) = 0 iff v (a ) = 1 and v ( β ) = 0
v (a V β) = 0 iff v (a) = t; (/?) = 0
v (a A β) = l iff v (α ) = v ( β ) = 1
v (a = β) = 1 iff v (a ) = v ( β ) .

Observe that each valuation v of For (and, similarly, o f any Z C For)

is uniquely determined by its ‘ reduct’ v/Var to the set o f propositional
variables Var C For (v/\far(Z), where Var(Z) is the set o f all variables
appearing in formulae o f Z)\ the values o f compound formulae can be
established by carrying out the arithmetic in truth-tables.
 Propositional tautologies 7

1.2 Propositional tautologies

The formulae which are exclusive schemas o f true sentences are called tau­
tologies. Accordingly, the set TAUT o f all tautologies o f CPC contains all
the formulae that take on the designated value 1 at every valuation:

TAUT = { a G For : να = 1 for every valuation v : For —►{ 0 ,1 } } .

Since any formula a £ For comprises finitely many propositional variables,

the set TAUT is decidable: the number o f all possible substitutions o f
variables in a is a function o f the cardinal number o f \hr(a) and amounts
to 2n for n variables in a.
It is worth while mentioning that some formulae o f this group were
considered logical laws in antiquity. O f special interest among them are:

(TO) —I- ip — p (law of double negation)

( T l) p V -ip (law of the excluded middle)
(T2) -'(p Λ -ψ ) (principle of contradiction)
(T3) (p->q) A p - > q (modus ponens)
(T4) (p -*· q) Λ ->g -»· -φ (modus tollens)

Notice that to test the validity o f (T 0 )-(T 2 ) it suffices to consider two

substitutions: v(p) = 1 and v(p) = 0. And, for verifying (T3), one has to
work through four cases. Incidentally, let us mention the shorter truth-table
technique (or reductio ad absurdum method) which consists o f confirming
the existence or not o f a valuation falsifying a given formula (i.e. assigning
0 to it).
The tautologies of CPC are the laws o f classical logic expressed in
propositional language. Their interpretation adducing customary prag­
matic rules o f a natural language allows us to determine both their content
and consequences. For example, ( T l) and (T2) are usually ‘read’ : given
two contradictory propositions p, -ip:

* at least one o f them is true; ( T l)

* at least one of them is false; (T2).

Observe that, thus interpreted, ( T l) and (T2) together represent the logical
principle o f bivalence. Furthermore, the tautologies (T3) and (T4) warrant
the deductivity of the following rules of inference:

* If (it is true that) Φ —> Φ, and (it is true that) Φ,

then (it is true that) Φ; (T3)
* If (it is true that) Φ —►Φ, and (it is true that) -ιφ ,
then (it is true that) -ιφ; (T4).

The former, known as the detachment rule (or modus ponens ), is a stan­
dard rule o f inference used in most syntactical formalizations o f systems of
8 1. The classical logic

classical and non-classical logic.

1.3 Functional completeness

All the hitherto described propositional connectives exhibit the same prop­
erty o f extensionality with respect to truth and falsity. This is due to
the fact that adequate functions characterized in terms o f truth-tables are
merely truth functions, i.e. the logical value o f any compound proposition
depends only on the logical values of (simpler) sentence components (and
not, for example, on their denotations). For that reason, the considered
connectives are often called truth-functional.
As everyone is quick to point out, the given stock
hardly exhausts (even in a small range) the list o f all
a β a -f- β
possible truth-functional connectives. For the sake
o f illustration let us at least mention an example of 0 0 0
0 1 1
the so-called exclusive disjunction that corresponds
to the everyday expression: either... or. This con­ 1 0 1
1 1 0
nective is characterized by means o f the intuitive
truth-table presented here.
The generalized truth-table method allows us to settle the class o f all
truth-functional propositional connectives and, consequently, to define the
full two-element algebra o f functions on the set {0 ,1 }. For a natural n > 1,
an π-argument truth-functional connective is then characterized by the
table:

«1 <*2 ...
ή t2
Z1 n W1
t2
τ2 <? V)2

b2n t2
b2n <5» W2n

where {{£*, 1 < i < 2n} is the set o f all (different from each other)
n-element variations o f { 0, 1} and w i , . . . , w2* is a sequence o f values from
{0 ,1 }. One thing to notice is that there are 22* such connectives: 22 = 4
- unary, 22 = 16 - binary, 22 = 256 - ternary ... etc. Two nullary
connectives (constants) 0 and 1 complete the list.
Post (1921) showed that each truth function is definable as a super­
position o f elements o f some sets o f unary and binary functions: here the
standard set o f connectives suffices. The property just specified is referred
to as functional completeness. In that sense the classical propositional cal­
culus is said to be functionally complete, and, hence, its logic concerns
all (and not just some) truth-functional connectives. It is, however, worth
 Axiomatization o f CPC 9

mentioning that the postulate is also satisfied by some proper subsets of

the set o f basic CPC connectives. The set consisting o f the negation ( - 1)
and implication (—►) may serve as an important example, the connectives
o f disjunction, conjunction and equivalence being defined by:

aVjS =df -1a —►β

ot Λ β =df -ι(α —►—>/3)
α= β =df -.((a — β) — -(/? — a))

(To verify these definitions the truth-tables defining the formulae on the
right should be compiled and compared with the tables o f the connectives
actually ‘ defined’ .)

1.4 Axiom atization o f CPC

The principal syntactical characterizations o f the classical propositional
calculus are produced by means o f the axiomatic method. Particular ax-
iomatizations of CPC systems vary in primitives (i.e. connectives), axioms
and rules o f inference. Each o f them, however, aims at bringing out a small
set o f tautologies to serve as axioms and some rules o f inference that allow
the derivation (from the axioms adopted) o f all the remaining tautologies
expressed in terms o f the chosen connectives. An equivalent, though more
handy, means o f axiomatization is to give some schemas o f axioms (one has
to use symbols like a ,/? ,7 . . . rather than p, ς, r . . . ) and to give up the rule
o f substitution, which in the former case is necessary (see below). The lat­
ter method is o f particular importance whenever the propositional calculus
has to be extended to the predicate calculus (consult 1.5).
The rules o f most o f the known axiomatizations o f CPC are the substi­
tution rule (SUB) and the detachment rule (M P), the latter being used
only when implication is a primitive o f a system. A substitution o f a
formula Φ (ρχ,. . . ,p n) having at most P i ,. . . ,p n variables is any formula
βΦ = Φ (β ρ ι/ρ ι,. . . , epn/pn) obtained from Φ as a result o f the simultane­
ous introduction o f some formulae e p i,. . . , epn in place o f p i , . . . ,p n. The
substitution rule:

(SUB) Φ / βΦ

states that if a formula Φ is a theorem (o f a given system) then any o f

its substitution instances is a theorem as well. The detachment rule is a
two-premise rule:

(M P) Φ —►Φ, Φ / Φ.

It is not hard to see that TAUT is closed under (SUB) and (M P) - if a

premise(s) is(are) a tautology(ies) so is the conclusion, i.e. it belongs to
TAUT (see 1.2).
10 1. The classical logic

The best-known axiomatization o f CPC with -i, — V, Λ, Ξ as primi­

tives and (M P) and (SUB) as rules is the axiom system

( A l) (p - » q) -► ((? -»· r) -► (p -> r))

(A 2) p —* { q —*p)
(A 3) ( ? —► (?—►? )) —*· (p —» i )
(A 4) p -► -i-ip
(A 5) -i-ip —*· p
(A 6) ( p —* q ) —* h q -»■ ->p )
(A 7) pAq-^p
(A 8) p f\q —* q
(A 9) ( ρ - * ϊ ) ^ ( ( ρ ^ ί ,) - ' ( ρ - * ? Λ ) · ) )
(A10) p —> p V q
(A ll) q-^pMq
(A12) (p —* r) —* ((q —>r) —* (pV q —* r))
(A13) (p = ? ) ^ ( p - ?)
(A14) (p ΞΞg) - > ( q - > p)
(A15) { p —*q) ~* ((« - * p ) ^ ( p = « ))

originated by Hilbert and Bernays (1934), (1939). Other axiom systems o f

CPC also worth mentioning are the ( - 1, V)-system o f Whitehead and Rus­
sell (1910) and the (-«, —^-system o f Lukasiewicz (1929), both using (M P)
and (SUB) and equipped with definitions o f missing standard connectives
(compare 1.3).
Given an axiom system, any o f its axioms as well as each formula derived
from the axioms with the adopted rules o f inference is its theorem. The
adequacy problem is a matter o f basic importance: an axiom system is
adequate (for CPC) whenever the set of tautologies formulated by means
o f its primitive connectives coincides with the set o f theorems o f the system.
The proof o f adequacy usually comprises two parts: the proof of soundness
and the completeness proof. An axiomatization is sound provided that
each theorem is at the same time a tautology - this property is easily
elicited: one should show that all the axioms are tautologies and that the
accepted rules o f inference do not lead out of the set o f tautologies (and,
hence, that any formula derived from the axioms is also a tautology). An
axiomatization is complete if all tautologies (built up o f its primitives) are
its theorems. To prove the completeness is a non-trivial and not easy task:
the first proof presented by Post (1921) for the Russell and Whitehead
system o f CPC employed so-called normal forms. The method o f normal
forms is but one o f the acknowledged methods; the best survey o f the most
important methods o f proving completeness is to be found in Surma (1973).
 Predicate calculus 11

1.5 Predicate calculus

Logical analysis o f the internal structure o f simple sentences, i.e. sentences
without any propositional connectives, leads to the predicate calculus which
is an extension o f the CPC. An over-simplified version of this calculus, apart
from CPC symbols, also comprises:

nominal variables: x, t/, z, χχ, j/1? ζχ , . . .

predicates: P, Q , Λ, Pi, Qi, P i, · · ·
quantifiers: V (universal or general), 3 (existential).

Nominal variables stand for individual names and predicates for name-
argument propositional functors1. Each predicate is uniquely ascribed
with a natural number indicating its syntactic category: one-argument
predicates refer to (objects’ ) properties; all the others represent relations.
Where P is an n-argument predicate, any formula P (x 1, X2, . . . , x n) is said
to be atomic. Quantifiers are operators corresponding to everyday phrases
like ‘ every’ , ‘for every’ (universal) and ‘some’ , ‘for some’ (existential); they
always appear together with single (name) variables in the sequences Vx
(for every x) and 3x (for some x).
The set o f formulae For* o f the predicate calculus is defined as For in
CPC where atomic formulae are treated similarly as propositional variables
and the following condition is added:

(*) if a £ For* and a; is a variable, then V x(a), 3x(a) £ For*.

The bracketed or atomic formula following a quantifier is referred to as a

scope o f the quantifier. Every occurrence o f a variable x in a scope o f Vx or
3x is bound. Generally speaking, x is bound in Φ £ For* whenever every
occurrence o f x in Φ is bound. A variable that is not bound in a given
formula (i.e. not standing at least once in the scope o f a quantifier which
bounds it) will be called free in this formula.
To get a description o f the predicate calculus in terms o f ‘ truth’ one uses
interpretations to extend the machinery o f truth-valuations of propositional
calculus. An interpretation / ( / d ) of For* in a non-empty domain D is a
function associating a free variable with an element o f D , every n-argument
predicate with an n-place relation in D and, ultimately, any formula with
a logical value (0 or 1). It is assumed that:

(1) f ( P ( x 1, « 2, . . . , Xn)) = f P ( f ( x 1) , . . . , f ( x n)) for any n-argument pre­

dicate P .
(2) / is a logical truth-valuation of the set For*.
(3) /(V x a ) = 1 iff f a = 1 for any interpretation / ' differing from /
at most on x.

1Predicates are, cleaiiy, variables o f another kind.

12 1. The classical logic

f( 3 x α ) = 1 iff f a = 1 for some interpretation f differing from /

at most on x.
A formula a £ For * is a tautology o f the predicate calculus provided that
f a = 1 for any / ( / d ).
The predicate calculus conceived as a theory o f all the tautologies is
undecidable (see Church (1956)), i.e.no effective method exists which, when
applied to any formula, would be able to decide whether it is a tautology
or not. An axiomatization avails all the more. For the predicate calculus
one may receive it by adding the schemes o f axioms and rules o f inference
which characterize the quantifiers to the set o f axiom schemes o f CPC with
(M P). The property o f mutual definability o f the quantifiers

(D ) 3x — a -iV x(-ia ), Vxa = ->Ξχ(-·α)

simplifies the task. Therefore, in order to axiomatize the predicate calculus,

it suffices to enrich the set o f axiom schemes o f CPC (see 1.3) with

(A16) V x(a —►/?)—►(a —►Vx/?) under the proviso that x is not free
in a,
(A17) V xa(x) —►a(y) provided that the substitution y/x
is permissible (i.e. y will not be
bound in any place in which
x is free in a (x )),

and to take (M P) and the so-called generalization rule

(GN) a( x) / Vya(y),

as the rules o f inference.

The earliest proof o f completeness o f the classical predicate calculus
is due to Godel (1930). The investigation o f the predicate calculus has
a celebrated tradition and the theory o f its interpretation, i.e. the theory
of models, has become an autonomous scientific discipline (see Chang and
Keisler (1973)).

1.6 Algebraization
The idea o f treating the formulae o f logical language as algebraic expres­
sions was originated by Boole (1815-1864). The basic conceptions o f an
algebraic approach to logic begun by Boole and his followers were mainly
inspired by Aristotle’s syllogism, the theory o f inference o f propositions
concerning simple relations between classes (i.e. sets) o f objects; see e.g.
Kotarbmski (1964).
To articulate Boole’s intuitions more specifically, we have to stick to
set theory. Let U be a universal set (space, universe, etc.) o f all actually
considered objects and let 0 be the empty set, i.e. containing no elements.
 Algebraizaiion 13

Let, moreover, P(U) denote the power set o f 17, i.e.the family o f all subsets
o f U:

P(U) = { X : X C U}.

Then any set X £ P(U) may be identified with a function from U to the
set o f classical logical values: X : U —> {0 ,1 } defined as

1 if a£X
X( a) =
0 if a£X.

The equations X( a) = 1 (0) read 6a £ X is true (is false)’ and the above-
mentioned function is often referred to as a characteristic function o f X
(in U). Obviously, U and 0 behave as constant functions with values 1 and
0, respectively.
Given the function set thus defined, we determine the operations o f the
complement, - , o f the union U and o f the intersection Π assuming that for
any a £ U :

—X( a) = 1 - X(a)
X U Y(a) = max{ X{ a) , Y (a )}
X Π Y (a) = m m {X (a ), Y(a)}.

Next, if we introduce a partial ordering relation C putting2

i c y iff X( a) < Y(a) for any a £ U ,

then l u y will occur as the least upper bound o f X and y , and X C\Y
as the greatest lower bound. Hence the family P(U) is a lattice with the
greatest element U and the smallest element 0; see e.g. Gratzer (1968) for
details. Furthermore, the lattice is distributive: for any Χ , Υ , Ζ £ P(U)

x n {Y uz) = (x n Y )u (x n z)
x u (Y η z ) = ( x u y ) η ( x u z )

and complemented since

x u - x = u, i n - i = 0.

Nowadays, one defines Boolean algebra as just a complemented dis­

tributive lattice having a greatest 1 (unit) and a smallest 0 (zero) element.
Notably, the structure (L , U, Π, —, 1, 0) is called a Boolean algebra
whenever the following conditions are satisfied:

2111 the following, at any place where it is convenient, 0 and 1 will be treated as
numbers.
14 1. The classical logic

(L I) x U y = y U x , xC\y = yC\x
(L2) x U (y U z) = (x Uy) U z, x ΙΊ (y Π z) = (x ΙΊy)ΙΊ 2
(L3) x U x = ai, x Γ\x — x
(L4) x Π (x U y) = x, x U (x Π y) = x
(L5) x Π (y U 2) = (x Πy) U (x Π z), x U (y Π 2) = (x U y) Π (x U 2)
(L6) xU O = x, x π 1 = x
(L7) x U —x = 1, x Π —x = 0

(see Traczyk (1970)); (L 1)-(L 4) define lattice, (L5) constitutes distribution

conditions, (L6) appoints 0 as zero and 1 as unit o f the algebra, and finally
(L7) conveys the operation - as a complement.
Boolean algebras can, in general, be formed o f any unspecified elements.
The representation theory provides several methods for describing them as
some ‘ concrete’ algebraic structures. In this connection, o f special impor­
tance is the representation of Boolean algebras by means o f fields o f sets;
a field of sets is a non-empty family o f subsets o f any set closed under the
operations o f union, intersection and complementation. One may readily
remark that, for instance, the above family P(U) is a case. Each Boolean
algebra is (isomorphic to) a field of sets. The two-element structure

S 2 = ({0 ,1 }, V, Λ, 1, 0),

with the operations defined by truth-tables (see also 3.1) is the simplest
Boolean algebra. Another, easily applicable representation theorem states
that each Boolean algebra is (isomorphic to) a subalgebra o f a Cartesian
power o f a number o f copies o f f?2·
Now, consider a binary relation « defined on the set For o f the formulae
of CPC as follows: for any α , β E For

a &β iff a = β is a tautology

(a « β iff !>(<*) = ν(β) for any valuation v). Subsequently, in the

set For/ « o f abstraction classes o f « introduce the operations —, U, Π
putting3:

—H = h « l
|«|U|/?| = \ανβ\
Μ η \β\ = | « λ /3|.

Let, moreover, 0 = \a Λ -<α| and 1 = \a V -ια|. Then it can be easily

verified that the introduced operations and constants satisfy (L 1)-(L 7).
The Boolean algebra

L_ = (For/ « , U, Π, - , 1, 0)

3The correctness o f the construction follows from the algebraic properties o f S . ~ is

a congruence o f the language of CPC (compare 3.1).
 Algebraization 15

will be called a Lindenbaum algebra (o f the classical logic). Prom the con­
struction presented, it immediately follows that any equivalence tautology
determines some identity o f and, conversely, any identity o f the algebra
corresponds to a tautological schema (a class of tautologies). This relation­
ship holds, for example, between formula α Λ (/? V 7 ) = ( a A β ) V ( a A j )
and the first identity in (L5). Since is a free algebra (in the class o f all
Boolean algebras; see Gratzer (1968)), every identity which is true in this
algebra is also satisfied in any Boolean algebra. It is in order to mention
that similar connections hold between the predicate calculus and (special)
Boolean algebras; see Rasiowa and Sikorski (1963). Thus, the theory o f
Boolean algebras is, in a sense, an algebraic version o f the classical logic.
2. The third logical value of
Lukasiewicz

The actual introduction of a third logical value by Lukasiewicz (1920),

next to truth and falsity, was preceded by thorough philosophical stud­
ies. Their crowning achievement was a three-valued propositional calculus
which, from the point o f view o f logic, represented a standard line o f ap­
proach. However, in view o f the surprisingly rich motivation substantiating
the new logic and also the hopes it produced, its exceptional position has
been maintained.
The present chapter is an overview o f the origin and basic properties o f
the three-valued logic o f Lukasiewicz.

2.1 Lukasiewicz and the Lvov-W arsaw school

Lukasiewicz’s work, especially that done in 1906-1922, resulted in a signifi­
cant philosophical context. The first and second decade o f the 1920s wit­
nessed the development o f the so-called Lvov-Warsaw philosophical school
of which Lukasiewicz was a co-originator and a prominent representative
(see Woleiiski (1985)).
It is an unenviable task to decide which o f the debates carried out
in the Lvov-Warsaw school directly contributed to the chief logical dis­
covery o f Lukasiewicz. According to several notes in his works one may
maintain that the main source of Lukasiewicz’s views on the logic may be
subscribed to a discussion concerning the Brentano-Twardowski-Meinong
general theory o f objects. Created by Meinong towards the end o f the
19th century, the theory o f contradictory objects postulated the existence
of objects having contradictory properties, such as, for example, squaring
the circle. Meinong claimed that the non-existence o f contradictory ob­
jects would result in the inability to utter any true proposition, notably a
proposition stating that they are not objects. Lukasiewicz advocated the
theory o f contradictory objects. What is more, he even tended to think
that non-contradictory objects do not exist. It is noteworthy that the turn
o f the 19th century witnessed Russell’s discovery o f a paradox in set theory
with the comprehension axiom. This fact was acknowledged by Lukasiewicz
who, consequently, attacked the logical principle o f contradiction (see 1.2)
 Three-valued logic 17

(Lukasiewicz (1910)).
The important part o f Lukasiewicz’s studies was concerned with the
problems o f induction and the theory o f probability. Especially, while deal­
ing with the latter, he extricated himself from the ‘ embarrassing’ principle
o f contradiction (see Lukasiewicz (1913)) and classified as undefinite the
propositions with free nominal variables. Subsequently, he assigned frac­
tional ‘ logical’ values to undefinite propositions indicating the proportions
between the number o f actual variable values verifying a proposition and
the number o f all possible values o f that variable. Under this conception,
‘ logical values’ exhibit the feature o f relativity and depend on the set o f
individuals actually evaluated. So, for example, the value o f the propo­
sition ‘ cc2 = 1’ amounts to 2 /3 in the set { —1 ,0 ,1 } and to 2 /5 in the
set { —2, —1,0 ,1 ,2}. Obviously enough, infinite sets o f individuals are not
admitted and this immediately implies that Lukasiewicz’s suggestion can­
not be taken seriously within the theory o f probability (some attempts to
improve the idea are due to Zawirski (1934a)). However, independently
o f the justification for combining the problems o f many-valuedness with
probability - we will return to this question later - the crucial fact is that
already in 1913 Lukasiewicz had employed the concept o f logical value in
an unorthodox manner.
The studies that finally led Lukasiewicz to the construction o f three-
valued logic touched upon determinism, indeterminism and some related
problems like the causality principle and modality (i.e. possibility and ne­
cessity); see Lukasiewicz (1906, 1910). Some historians o f logic suspect
Lukasiewicz o f being influenced by the discourse (in the Lvov-Warsaw
school) about freedom and creativity. Its main thesis was formulated by
Kotarbmski (1913) who suggested the need for revising the two-valued logic
that seemed to interfere with the freedom o f human thinking. Lukasiewicz
was a fierce follower o f indeterminism, which found its expression, among
others, in the introduction o f the third logical value, next to truth and fal­
sity, to be assigned to non-determined propositions; specifically, to propo­
sitions describing casual future events, i.e. future contingents (see also the
Introduction).

2.2 Three-valued logic

The very earliest remarks about the three-valued propositional calculus
can be found in the Farewell Lecture given by Lukasiewicz in the Assembly
Hall o f Warsaw University on the 7th o f March, 1918. Next came the paper
‘ O logice trojwartosciowej’ (Lukasiewicz (1920)); it contains a brief moti­
vation for the new logical construction casting off the principle o f bivalence
and, further, it provides an outline o f the three-valued logic. Lukasiewicz
(1930) analyses the sentence T shall be in Warsaw in a year’ and states that
at the moment it is being uttered, its value (truth or falsity) is not settled.
18 2. The third logical value o f Lukasiewicz

Hence, he suggested that sentences o f this kind pertain to a ‘ third’ logical

category. Consequently, to the two classical values 0 and 1, he added an
intermediate logical value 1/2 interpreted as ‘ possibility’ or ‘ indeterminacy’
- the former o f these options was subsequently repudiated by Lukasiewicz
under the influence o f the studies on modality (see 2.3). Using his intu­
ition, Lukasiewicz extended the classical interpretation o f implication (—►)
and negation (-■) connectives, which resulted in the putting forward o f the
following tables4:

CL - > 0 1 /2 1
0 1 0 1 1 1
1/2 1/2 1 /2 1 /2 1 1
1 0 1 0 1 /2 1

The other relevant connectives o f disjunction (V), conjunction (A) and

equivalence (= ) were introduced through the following definitions:

a A β = -«(-ία V -■/?)
a = β = (α —►β ) Λ (β —►α ).

Their tables are as follows:

V 0 1/2 1 Λ 0 1/2 1
0 0 1/2 1 0 0 0 0
1/2 1/2 1/2 1 1/2 0 1/2 1/2
1 1 1 1 1 0 1/2 1

= 0 1/2 1
0 1 1 /2 0
1 /2 1 /2 1 1 /2
1 0 1 /2 1

A valuation o f the set For in Lukasiewicz three-valued logic is any function

v : For —►{0 ,1 /2 ,1 } ‘ compatible’ with the above tables (compare 1.1, 3.3);
a tautology is a formula which under any valuation v takes on the designated
value 1.
System L3 o f the three-valued propositional calculus differs radically
from CPC. On the one hand some important laws of the classical logic are

4Prom now on, the truth-tables o f binary connectives * are viewed as follows: the
(logical) value o f a is placed in the first vertical hue, the value o f β in the first horizontal
line and the value o f α * β at the intersection o f the two hues.
 Modality and three-valuedness 19

not tautologies o f L3, but on the other hand there are classically contra­
dictory formulae (countertautologies)5 that are consistent in the logic o f
Lukasiewicz. Among the formulae of the first kind appear:

(1) p V -ip (law o f the excluded middle)

(2) -i (p Λ - φ ) (principle o f contradiction).

The second case is illustrated by the following important formula:

(3) p = ->p.

The facts presented can be checked with the help o f any valuation which
yields 1/2 for p. Such a valuation associates 1/2 with (1), (2) and 1 with
(3).
The thorough-going refutation o f the law of the excluded middle and the
principle o f contradiction was intended, in Lukasiewicz’s opinion, to codify
the principles o f indeterminism: he says that ‘ both disjunction and con­
junction o f two possible propositions are possible propositions (and nothing
else)’ . Whereas the consistency o f (3) supports the claim that Lukasiewicz
logic is adjusted to the formalization o f reasonings about contradictory ob­
jects, peculiarly, as Lukasiewicz remarks, Russell’s paradox o f ‘ the set o f all
sets that are not their own elements’ ceases to be an antinomy in L3. It may
not be amiss to remind readers that Russell’s set Z is defined by the equa­
tion

Z = {x : x £ a?},

and the resulting paradox

Z £ Z = Z £Z

is a substitution o f (3) (see also 11.2). The question as to what extent the
hopes are realized o f applying the three-valued as well as other many-valued
logics o f Lukasiewicz, shall be brought up later.

2.3 Modality and three-valuedness

W hat appears to be one o f Lukasiewicz’s intentions is working out a tool to
formalize the non-truth-functional functors o f possibility ( M ) and necessity
( L). Adopting the established relation of mutual definability

La = -iA f-ια,

he put forward a minimal postulate to preserve in the logic acquired the

consistency o f everything inherited from the earlier intuitive theorems on

5That is, formulae taking 0 at arbitrary logical valuation.

20 2 . The third logical value o f Lukasiewicz

modal propositions, i.e. that for every such theorem there would exist at
least one interpretation (a valuation) in L3 by which it would be true. All
these theorems were finally reduced (see Lukasiewicz (1930)) to the three
groups:

(M l) -i M ol —►—>a
(M2) — —> (-i« —> -i M ol)
(M3) Μ α Λ M->a for some a.

Let us add that ->M corresponds to ‘ it is impossible that’ , -*M - ■can be read
as ‘it is necessary that’ and Af-· as ‘ it is contingent that’ . It is also worth
while mentioning that to formulate (M3), an (existential) propositional
quantifier was adopted; this condition can equivalently be interpreted as:
‘ there exist(s) a such that Μ α Λ M-*a\
The following definition o f possibility connective satisfying the above
conditions was given by Tarski in 1921:6

M ol = —>a —>a.

Note that the table o f M and the functor o f necessity L associated with M
is as follows:

a Ma La
0 0 0
1/2 1 0
1 1 1

In turn, putting

Ι α = Μ α Λ -i La

we get a third modal connective ‘it is contingent, that’ (or, ‘it is modally
indifferent’ ) distinguishing the intermediate logical value:

a la
0 0
1/2 1
1 0

Applying I allows the formulation, within L3, o f counterparts o f the law of

the excluded middle and the principle o f contradiction

6According to Lukasiewicz (1930).

 Interpretation difficulties 21

(13) p V I p V -i p
(23) - i ( p Λ -i I p Λ -i p)

altogether confirming that Lukasiewicz logic is three-valued.

In spite o f the promising combination o f trivalence and modality the
full elaboration o f modal logic on the basis o f the three-valued logic never
succeeded (with the single exception o f the algebraic constructions o f Moisil
- see 5.4), which was the result o f Lukasiewicz’s further investigations on
modal sentences (see Lukasiewicz (1930)). Many years later Lukasiewicz
returned to the idea of construction o f a pluri-valued modal system and he
proposed a four-valued logic o f possibility and necessity; see Lukasiewicz
(1953).

2.4 Interpretation difficulties

The problem o f how to interpret Lukasiewicz logic intuitively has been
repeatedly explored. The most frequent aim was to relate the author’s
motivations more explicitly to the logico-formal construction.
Relatively early, i.e. in 1938, a serious blow was inflicted upon Lukasie­
wicz’s conception. At the time Gonseth noticed (see Gonseth (1941)) that
the formal characterization o f the connectives in Lukasiewicz logic is en­
tirely incompatible with the suggested ways of interpreting the third logical
value 1/2 either as "possibility’ or as undetermination. The argumentation
o f Gonseth is extremely sound and straightforward. Consider two proposi­
tions a and -■<*. Whenever a is undetermined, so is ->α, and then, according
to the table o f conjunction, a Λ ->a is undetermined, which contradicts the
intuition since, independently o f a ’s content, α Λ -«a is false. Similarly,
the unintuitiveness o f disjunction can be "explained’ : this time instead o f
a Λ -i« , the formula a V -*a is taken into consideration. As can be easily
observed, the definitiveness o f the original interpretations o f Lukasiewicz is
caused by the fact that they neglect the mutual dependence o f ‘ possible’
propositions.
An interesting attempt to interpret Lukasiewicz logic intuitively was
made by Slupecki (1964). By an "interpretation’ Slupecki means pointing
out some definite language which describes the property o f the determi­
nation o f events in a three-valued manner reconciling Lukasiewicz’s truth-
tables.
The language considered by Slupecki is a set-theoretical union o f a set
S comprising propositions about events and a set S* o f propositions that
do not state events. Both sets contain simple (atomic) propositions and
compound ones formed by means o f disjunction (V ), conjunction (A ) and
negation (-■) connectives. Since S* does not play a significant role in what
follows, it is omitted. Slupecki supposes the set o f events Z described by
propositions o f S to be closed under the operations o f union (U), meet (or
intersection) (Π) and complementation ( —) and, furthermore, the structure
22 2. The third logical value o f Lukasiewicz

Z = ( Z , U, Π, —) to be a Boolean algebra. There is a causality relation »-►

on Z (‘ / i »-►/ 2’ means ‘ the event / x is a cause of the event / 2’ ) providing
the assumption that

(P I) / *-*· / i U / 2 iff f f i or / ·—> / 2

(P2) / h / j H /2 iff f ^ f i and / »-> / 2
(P3) If / »-> / i for some / , then f + »-> —/ x for no f +
(P4) If / i i—►/ , then / i Π / 2 ·—►/

for any / , / i , / 2 E For the purpose o f defining the property o f deter­

mination, he then singles out a set of past and present events hereafter
denoted by symbols <7, <71, <72, . . and puts

D ( f ) = df there is a g £ Z such that g »-> / ,

D(f) = d/ not £>(/) and not D ( —f ) .

The above expressions can be read as:

D(f) — f is (at the present moment) determined,

D(f) — f is not determined (at the present moment).

The relationship between expressions of the language S and events is

established by relation * of describability (‘p * / ’ means: (proposition) p
describes (event) / ) satisfying the following conditions

P V pi * / U fi whenever p * / and p\ * fi
(*) ί > Λ ρ ι * / η / ι whenever p*f and pi * fi
when p * / , then -*p* —f

for any ρ ,ρ ι E S. Moreover, Slupecki assumes that every proposition o f S

refers to some event o f Z and he then defines the properties l(p ),0 (p ) and
l / 2(p) : 1(p) —p is true, 0(p) = p is false, l / 2(p) = p has the ‘ third’ logical
value, as follows:

if p * / , then { l(p ) iff D ( f ) }

(D T ) if p * / , then {0(p) iff D ( - f ) }
if p * / , then { l/2 ( p ) iff D ( f ) } .

Using (P 1 )-(P 4 ) and (*), it is easy to check that for x E {0, 1/2, 1}

x(p V q) = x(p) V x(q)

x ( p Aq ) = x ( p) A( q)
x(->p) = ~>x(p),

where V, Λ and -> appearing on the right-hand side are the connectives
of the Lukasiewicz three-valued logic. Therefore, what Lukasiewicz pro­
posed, referring to the property of determining propositions as a manner
 Interpretation difficulties 23

of interpretation of logical values, is, by (D T ), to some extent justified.

Slupecki’s interpretation, however, is not free o f faults. First, it is par­
tial, lacking the implication connective. Admittedly, Slupecki extends it
onto the language with modal connectives M and L, and in this enriched
language the implication o f Lukasiewicz is definable. However, it certainly
does not change anything since the interpretation o f the implication ob­
tained is fairly unintuitive. Secondly, more profound analysis of the whole
construction reveals that the assumption concerning Z has to be modi­
fied: Nowak (1988) proved the formal correctness o f the interpretation
exclusively when Z is a de Morgan lattice (see 7.1) and not a Boolean
algebra. This result, nevertheless, does not depreciate Slupecki’s proposal
but, on the contrary, makes it still more noteworthy; three-valued logic
can thus be interpreted as a set o f propositions describing events which
form a non-classical algebra. If so, then (D T) implies that the third value
of Lukasiewicz, 1/2, is assigned to propositions concerning non-Boolean,
undetermined events.
3. Logic algebras and matrices

The set o f formulae of the classical propositional calculus, For , considered

in Chapter 1, is closed under the operations of forming formulae by means
of propositional variables and the connectives -·, — V, Λ and = . There­
fore, treating connectives as algebraic operations we obtain that

Ck = (For, -«, —►
, V, A, = )

is an abstract algebra o f the signature (1 ,2 ,2 ,2 ,2 ). The algebra Ck is an

exceptional example o f a propositional language - the general definition
of a language given below will serve as a starting point for the semantic
considerations o f logical many-valuedness.
First, algebraic interpretation structures for propositional languages
within the classical Fregean semantic framework are derived. Next, func­
tional completeness, the property o f finite logic algebras which warrants
the biggest expressive power of the corresponding bunch o f connectives, is
discussed. In the end, matrices, i.e. algebras with distinguished elements
serving as models for propositional calculi, are introduced.

3.1 Language and logic algebras

Let λ/ar = { p , q, r , . . . } be a denumerable set o f propositional variables and
F = { jFi , . . . , jFm} a finite set o f sentential connectives - with each con­
nective Fi there is an associated natural number a(Fi) that describes its
syntactical category (i.e. number o f arguments). Assume, moreover, that
a(Fj) φ 0 for some i £ { l , . . . , m } . Finally, define inductively the set o f
formulae, For , putting

(i) Var C For

(ii) For any Fi £ F such that a(Fi) — k, F i(a i , . . . , a *) £ For
whenever a 1?. . . , a k £ For.

The algebras o f formulae

£ = (For, F i , . . . , Fm)
 Language and logic algebras 25

constructed in such a way will be called a propositional language.

For interpreting a language each formula Φ is provided with a meaning
which is its semantic correlate. At this stage two conditions are required
as far as a mapping r o f For into the range A of all semantic correlates is
concerned:
(*) W ith each Φ £ For exactly one semantic correlate is associated, i.e.
r is a function;
(**) Two formulae a ,/? £ For are interchangeable in any propositional
context Φ £ For whenever r(a) = r(/?), i.e.for any Φ £ For, p £ Var:

7*(Φ (α/ρ)) = τ*(Φ(β/ρ)) if and only if r(a) = r(/?),

where Φ(ct/p) and Φ(β/ρ) stand for the formulae resulting from Φ after
the substitution «(/?) instead o f p.

Conditions (*), (**) were assumed by Frege (1892). Since Frege had
identified semantic correlates of propositions with logical values o f truth
or falsity he inevitably leaned towards the classical logic. The second of
the two conditions is an exemplification o f the Leibniz principle of exten-
sionality and states that denotation of a proposition is a function o f the
denotations o f its components. So if we assume that the semantic corre­
lates are (logical) values o f new logics, we may conclude that connectives
of these logics are in the above sense extensional

3.1.1. (Suszko (1957)). When A is the set o f all semantic correlates of

the language C = ( For, F i , . . . , Fm), then for each t = 1 , . . . , m the formula

= r(Fi(au . . . , a n))

defines uniquely a function /,· on A o f the same arity as F,.

From 3.1.1, it immediately follows that each interpretation struc­

ture

A = ( A , h , . . . , f n)

is an algebra similar to £ . In turn, any mapping s : Var —►A may be

extended uniquely to the homomorphism hs : C —>A (hs £ Hom(C, A ))
and, therefore, C is an absolutely free algebra in its similarity class - it
is generated by the set Var. Furthermore, interpretations o f propositional
languages are determined by algebraic substitutions o f propositional vari­
ables. For their part functions o f the logical algebras A may be described
in an arbitrary way accepted in algebra, including tables.
26 3. Logic algebras and matrices

3.I.2.E. The structure B2 = ( {0 ,1 }, V, A, = ) with the opera­

tions defined by truth-tables (see 1.1) is the algebra o f the classical logic.

Propositional languages may also serve as special interpretation struc­

tures. C as well as any language similar to it are such structures. In the
former case the endomorphisms e : C —> C referred to below as substitu­
tions take the role o f interpretation mappings.

3.2 Functional completeness o f finite algebras

Where n > 2 is a given natural number, let us put En = { 1 , 2 , . . . , n } and
by Un denote any algebra o f the form:

= (•β'η? /l? ··· » /m )·

Among many-valued algebras Un there are some that have a special prop­
erty, namely every finitary mapping / : E* —> En (k > 0, k finite) can be
represented as a composition of the operations f i , . . . , f m. Such algebras
will be called functionally complete (Post (1921)).

3.2.1. E. The algebra B2 is functionally complete (see 1.3).

It is also possible to weaken that condition requiring only that for some
finite m any ^-argument operation on En, where k < m, is definable in the
above sense. In the latter case we shall say that Un is functionally com­
plete for m variables. The following theorem shows how the two notions of
functional completeness are related:

3.2.2. (Post (1921)). If Un is functionally complete for m variables,

where m > 2, then it is also functionally complete for m + 1 variables and
hence also functionally complete.

Notice that 3.2.2 reduces the problem o f functional completeness of

many-valued logical algebras to the problem o f definability o f all unary and
binary connectives. Given n, the number of these connectives equals nn
(unary) and (nn)12 (binary). For example, 4 and 16 when n = 2 (compare

1 · 3 ) ·
The first functionally complete n-valued algebra of logic (given finite
n > 2) is due to Post, who showed that in order to establish functional com­
pleteness o f Un it is sufficient to generate two functions: the one-argument
cyclic rotation function ( negation) and the two-argument maximum func­
tion V ( disjunction):
 Functional completeness o f finite algebras 27

1 if χ = η
“ •η® = x V y = m ax(x, y).
ί+ 1 if χ = ζ ^ η

Consequently, each n-valued Post algebra (see also 4.3)

r n = (E n, i n, v)

is functionally complete (Post (1921)). It is of interest to notice that V 2 =

( jE?2j ” ·25 V) is the (->, V)-reduct o f the algebra B2 .
Since functional completeness is a scarcely frequent property o f the
known logic algebras several scholars aimed at formulating handy crite­
ria which might determine whether or not a given algebra has the property
in question. Prom among them the following two are of special importance:

3.2.3. (Slupecki (1939a)). An n-valued algebra Un (n > 2,n finite) is

functionally complete if and only if in Un there are definable:
1°. All one-argument operations on En
2°. At least one two-argument operation / ( x , y) whose range
consists o f all the values i for 1 < %< n.

3.2.4. (Picard (1935)). In Un all one-argument operations are definable

whenever the functions if, if, S in 6.4 are definable in it.

3.2.5.E. The three-element Lukasiewicz algebra (see 5.3)

£3 — ( { 0, 1/2, 1}, V, A, = )

with the operations determined by tables in 2.2 is an example o f an in-

complete algebra. To see this it suffices to remark that the one-argument
constant function T : Tx = 1/2 for any x = {0, 1/2, 1} is not definable
in terms o f the basic operations. Notice, however, that 3.2.3 implies that
adding T to the stock of functions of leads to the (three-element) func­
tionally complete algebra (Slupecki (1936)).

The functional completeness o f the n-valued logic algebras is a mat­

ter of consequence since the propositional logics founded on such algebras
are logics o f all possible extensional n-valued connectives (truth-functional
when n = 2). The idea has also been elaborated in the most extremal di­
rection, namely by looking for the single operation generating all possible
operations on Un. It was Webb who, in 1935, found the so-called Sheffer
connective o f this algebra (Webb (1935)). This, as well as other minimaliza-
tion problems, attracted the attention o f many scholars especially because
o f their applicational aspects; compare 14.3.4. To complete the section
let us also mention that infinite logic algebras in principle are functionally
28 3 . Logic algebras and matrices

incomplete. This is due to the fact that the set o f possible functions of
any algebra o f this kind is uncountably infinite, while using a finite number
o f original operations we are able to define at most a countable family of
functions. Nevertheless, in the literature of the subject several notions of
local completeness have been considered.

3.3 Logical matrices

Interpretation structures equipped with a distinguished subset o f the set of
semantic correlates corresponding to propositions o f a specified kind (e.g.
true propositions) are called logical matrices. More specifically, a pair

M = ( A D),

with A being an algebra similar to a propositional language C and D C A

a non-empty subset o f the universe o f A , will be referred to as the matrix
for C. Elements of D will be called designated elements o f M . W ith each
matrix M for C there is associated a set o f formulae which take designated
values only:

E ( M ) = { a G For : ha £ D for any h £ H om (C ,A )}

is called its content

3.3.I.E. M 2 = (#2 > {1}) is the matrix for CPC (in the language £ * ).
Clearly, its content coincides with the set of all tautologies, E ( M 2) =
TAUT.

The notion o f matrix consequence being a natural generalization o f the

classical consequence is defined as follows: relation \=m Q %For x For is said
to be a matrix consequence of M provided that for any X C For, a £ For

X \=M if and only if for every h £ Hom (C,A)

(ha £ D whenever hX C D ) .
Notice that

E ( M ) = { a : 0 \=M <*}.

In turn, with every \=m there may be uniquely associated an operation

C nM : 2For — 2For such that

a £ C tim( X ) if and only if X m ct.

 Logical matrices 29

Where K is a class (a set) of matrices for a given language £ , the rela­

tion \=k is to be identified with the set-theoretical meet o f {\=m : Af £ K }.
Consequently, C tik = f]{CriM : Μ £ K }, i.e.for any X C For

CnK ( X ) = Π { CnM( X) : Μ £ K } .

The theory o f logical matrices is a theory of models o f propositional

calculi and, from a certain point o f view, may be treated as a fragment of
the algebraic systems theory; see Czelakowski (1980). The operations of
taking subalgebras, direct products, o f forming quotient algebras, etc., are
conveyed in an obvious way from algebras onto matrices: N = (Λ\ Djy) is a
submatrix o f Μ = (ΛΊ, D m ) if λ ί is a subalgebra of Λ4 and jDjy = NC\Dm ,
the direct product o f matrices Af and IV, is defined as Μ x N = (Λ ί x -A£,
x jDjv), a binary relation ~ is a matrix congruence of Af if ~ is a

congruence o f Λ ί and a ~ b (a,b £ M) implies [a € D m if and only if

b £ D m ] and, finally, Μ /^ = (Λ ί/^ , D m /~) is the quotient matrix of
Af modulo
4. Many-valuedness

The application of specialized tools of the methodology o f propositional

calculi (see W ojcicki (1988)) affords possibilities for a profound investiga­
tion o f many-valued constructions. The present chapter aims to present
some important methods o f that methodology, especially those revealing
the complexity o f the problem connected with a general characterization
o f logical many-valuedness.

4.1 Tw o criteria
The use o f logical matrices is undoubtedly the most natural way of achiev­
ing many-valuedness (see the Introduction). The very construction starts
with the choice o f the propositional language C - in most cases either Ck
itself or some o f its reducts serves for that purpose. Subsequently, one
defines a multiple-element algebra A similar to C (i.e. an algebra o f
a power > 2) and chooses a set D C A o f distinguished values. The
resulting matrix

M = ( A D)

may define a many-valued logic: we say that it is so whenever the content

o f M or the consequence determined by M cannot be described by any
two-element (two-valued) matrix. Putting the problem in such a way we
get the two following criteria:

(K l) M determines a many-valued logic if and only if for no two-element

matrix N for £ , E ( M) = E(N).
(K2) M determines a many-valued logic if and only if for no two-element
matrix N for £ , \=m = \=n (or? equivalently, C um = Cnjv).

It is in order to remark that two-valued logics need not necessarily be

determined by two-element matrices being directly related with the basic
matrix for the classical logic, i.e. having { 0, 1} as the set o f elements and,
at the same time, 1 as the designated value. To get the idea let us look at
the two-element matrix

dM 2 = ({0, 1}, V, Λ, = , {0 })
 Two criteria 31

for Ck (with the connectives exactly as in the standard case) dual for M 2.
Though very close to M 2, dM 2 differs from the former in two respects: the
content, E{dM 2 ) = {-·α : a G TAUT} φ E (M 2), and the consequence,
(=m 2 Φ \-dM-2 ·
Since \=m = \=n imphes E (M ) = E (N ), every matrix satisfying (K l)
also meets the requirements o f (K2). At the same time, however, there
exist matrices meeting (K2) but not (K l). Examples o f matrices having
an empty content are also known, which determine interesting consequence
relations (see Chapter 7). This obviously justifies taking into account the
relations o f consequence as well as the two criteria o f many-valuedness.
The examples given below illustrate the matter somewhat better.

4.I.I.E . Consider the matrix

M3 = ({0 ,M }, V, Λ, = , {t, 1})

for Ck with the operations defined by the following tables:

X —>x 0 t 1 V 0 t 1
i-H
O

1 1 t 1
"Si

0 0 0
t 0 t 0 t t t i t 1
1 0 1 0 t 1 1 1 1 1

Λ 0 t 1 = 0 t 1
0 0 0 0 0 1 0 0
i 0 t t i 0 t t
1 0 t 1 1 0 t 1

One may easily verify that this three-element matrix does not determine
many-valued (i.e. three-valued) logic even according to the condition (K2).
Due to the choice o f the set D , with each h G Hom(Ck, M 3) the valuation
h* G Hom(Ck,M 2 ) corresponds in a one-to-one way such that ha G {t, 1}
iff h*a = 1. Therefore \=m * = \=m 2-

4.1.2.E. Consider the matrix

J<3 = ( { 0 , U } , -S V, A, = ,

for £jfc with the operations defined by the following tables:

X —>a —► 0 t 1 V 0 t 1
0 1 0 1 1 1 0 0 0 1
f 1 t 1 1 1 f 0 0 1
1 0 1 0 0 1 1 1 1 1
32 4 . Many-valuedness

Λ 0 t 1 = 0 t 1
0 0 0 0 0 1 1 0
t 0 0 0 t 1 1 0
1 0 0 1 1 0 0 1

One may show that E (M 3) = E (M 2) and, accordingly, that M 2 is the

only two-element matrix which might determine \=m *· Simultaneously,
\=Mt Φ |= m 2 , since, for example,

{p q, P} l=M2 q while not {p - » q, p} |=Ms q.

To verify this it simply suffices to turn over a valuation h such that hp = t

and hq = 0. Summing up, M 3 satisfies the second condition o f many-
valuedness but fails to satisfy the first. From the point o f view o f (K2) it
determines a many-valued logic and in view o f (K l) a two-valued logic.

4.2 Stm cturality and many-valuedness

C um considered in 3.3 are special examples o f the operations considered
in the general theory o f deductive systems originated by Tarski (1936). A
mapping C : 2For —►2Fot will be referred to as a consequence operation o f
the language C if and only if for any 1 , 7 C For

(TO) X C C(X)
( T l) C(C(X)) = C(X)
(T 2) C ( X ) C C ( Y ) whenever X C Y.

If, moreover, for any substitution e E End(C)

(S) e C ( X ) C C( eX) ,

we shall say that C is structural

It is easy to prove that each matrix consequence operation C um is
structural. Conversely, each structural consequence C o f C and any set o f
formulae X together determine a matrix

Lx = (£ , C ( X ) )

called a Lindenbaum matrix.

4.2.1. (Lindenbaum). The content o f (£ ,C (0 )) equals C (0).

The class o f all Lindenbaum matrices o f a given consequence C

of £,

LC = { ( £ , C ( X ) ) : X C For },
 Finiteness and deduction 33

will be referred to as the Lindenbaum bundle. Notice that for Lindenbaum

matrices the substitutions (i.e. endomorphisms) o f the language C take the
role o f valuations. Taking this into account it is easy to verify:

4.2.2. (W ojcicki (1970)). Each structural consequence operation C is

uniquely determined by its Lindenbaum bundle: C = C tilc·

4.2.2'. For every structural consequence operation there is a class K of

matrices such that C = Cnjc-

When a class K o f matrices for a given language C determines the

structural consequence C, i.e. if C uk = C, we shall say that K is adequate
for C . Such C is referred to as strongly finite provided that there is a finite
class o f finite matrices adequate for it. Some criteria o f strong finiteness
using congruences o f the language compatible with C (a ~ β implies
C( a) = C (/?)) can be found in W ojcicki (1977).
A sufficient and necessary condition for a given structural consequence
to be determined by a single matrix formulated in terms o f absolute uni­
formity is due to W ojcicki (1970). Such a C is uniform if for any X , Y , a
such that Var{X,a) Π Var(Y) = 0 and C( Y) φ For , α E C ( X U Y) im­
plies a: E C( X) . In turn, C is absolutely separable if for any family X
o f sets o f formulae such that for arbitrary Χ ,Υ E X , X φ Y implies
Var{X ) Π \hr(Y) = 0, and for any propositional variable p E Var{(J^O

(as) p E C^U^O implies p E C (X ), for some X E X-

C which is both uniform and absolutely separable is referred to as absolutely

uniform.

4.2.3. (W ojcicki (1970)). For a structural consequence operation C an

adequate single matrix exists if and only if C is absolutely uniform.

4.2.4. E. Consider the (structural) consequence Co determined by the

two-element class o f matrices K = {A f2,dA f2} on £ *, i.e. Co = C n # .
Notice that Co(0) = 0. Subsequently, take any two formulae having no
variables in common, e.g. ο: = p V -*p and β = q V ->q. Then we have
a E Co(/3) = For and a φ Co(0) = 0. This proves that Co is not uniform
and hence absolutely uniform and according to 4.2.3 there is no single
matrix adequate for it.

4.3 Finiteness and deduction

Syntactical characterization o f consequence operations refers to inference
(associated with each C is a relation f=c such that: X ^=c a iff α E
C (X ) ) . Any pair (X, a) will be referred to as a sequent and a set o f
sequents as a rule (of inference). Sequents o f the form (0, a ) will be called
34 4- Many-valuedness

axiomatic and a set consisting o f only such sequents an axiomatic rule.

It is straightforward that an arbitrary consequence C may be conceived
as a rule composed o f all the sequents ( X , a ) where a E C( X) . A rule
R is structural provided that ( X , a) £ R implies ( e X , ea) E R for any
substitution e E End(C). For their part structural rules may be represented
through a number o f sequents, one in particular:

4.3.1. Modus ponens (the detachment rule), MP, is a structural rule

determined by the sequent {p —►q, p}/q and represented through the fol­
lowing schema: Φ —►Φ ,Φ /Φ .

Let X be a set o f formulae and R a rule o f inference (or, equivalently,

a set o f rules o f inference). We shall say that X is R-closed iff for every
a E For , Y C X, (Y,ot) E R implies a £ X . The operation C ur defined
for every X C For by

CnR( X ) = f|{ Y C For : X C Y, Y being R-closed}

proves to be a consequence (structural if R is structural). Every R such

that C = CnR is referred to as a base o f C. In the case when R contains
axiomatic sequents, i.e. when R = RA U R1, we may, by putting A = {a :
(0, α ) E represent C ur as C n ^ R i ) defined for any X C For as
follows:

Cn(A, Ri)( X ) = f ) { Y : A U X C Y and Y is ^ - c l o s e d } .

Then, the elements o f A will be called axioms.

A binary connective —►o f the language C o f C, for which the following
condition (later referred to as the deduction theorem)

(Ded) β E C{ X, a) if and only if a —►/? E C ( X )

is satisfied, is called implication (with respect to C). The widely accepted

way o f formalizing propositional logics (£ , C) with implication consists in
defining C as C n ^ MP) or ^ n(A, {SUB,MP})> where SUB is the substitution
rule, SUB = {Φ /βΦ : e E End(C)}. From the perspective o f the present
chapter, such a formalization is very handy. Mostly, when the consequence
operation o f a logic is finite (see below) the question o f many-valuedness
reduces to the satisfiability of the condition (K l).
We say that a consequence C is finite if for every X C For , a E For:

(Fin) ο: E C ( X ) if and only if α E C ( X f ) for some finite X f C X .

4.3.2. (W ojcicki (1977)). Every strongly finite consequence operation

is finite.
 Finiteness and deduction 35

4.3.3. If —> is an implication with respect to a finite and structural

consequence C, then C = Cn(C(ib),MPy

P roof . The inclusion C n (c(0), m p ) (X ) Q C7(-X") is obvious. Assume then

that a G C( X) . From the assumption it follows that there is a finite set o f
formulae X f C X such that a G C ( X f ) . Suppose that X f — { /? i ,. . . ,/?*}.
Since a G C (/? i,. . . ,/?*), applying the deduction theorem fc-times (Ded),
we get βι (fo -*■ . . . -*· (/3*-i -► (/3fc -> a ))) G C (0). Therefore,
a € C « ( c ( 0), m p )(-X /) C C n (c(0), m p )(-^)·

4.3.4. E. Every consequence having a finite base consisting o f finite rules

only (i.e. rules with finite sets of premises) is finite. An important example
o f such an operation is the inferential consequence o f the classical logic:
C2 = ^ ( { Λ ΐ - Λ ΐ 5},ΜΡ)ί where A1-A15 are schemes o f the axioms in 1.4.
Since C2 = C um 2, C2 is strongly finite as well. That 4.3.3 implies C2 =
Cn^TAUT, MP) may also be o f some interest (compare 3.3.I.E).
5. Lukasiewicz logics

The generalization o f his three-valued construction (see Chapter 2) led

Lukasiewicz in 1922 to the family o f many-valued matrices, both finite and
infinite valued (countable and o f the power o f continuum); see Lukasiewicz
(1961 p.135). The propositional calculi o f Lukasiewicz and the related
algebras have undoubtedly an important place in the history o f logical
many-valuedness. An account o f the work o f Lukasiewicz and his followers
is given below.

5.1 Fundamental properties

A matrix o f the form

— (Ln? V, A, = , {1 })

is called a Lukasiewicz n-valued matrix (n £ N , n > 2 ) ( n = N0 or η = Νχ)

provided that

f { 0, l / n —1, 2/ n —1, , 1} if n G A", n > 2

Ln = \ {s/w : 0 < s < w, s, w G N and w φ 0} if n — No
l [0, 1] if η = Νχ.

-· ( negation) is a unary function and —►

, V, A, = are binary functions (o f
implication, disjunction, conjunction and equivalence, respectively) defined
on the set Ln as follows:

(a) -I x = 1 — x
x —►y = ra m (l, 1 — x + y)
(b) x V y = ( x ^ y ) ^ y = m ax(x , y)
x A y =■ —'(—'X V ~*y) — m in(x , y)
x = y = (x —►y) A (y —►x) = 1 - \x - y\.

In introducing many-valued logics Lukasiewicz did not give any specific

reason for his choice o f ‘ truth-values’ . One may guess that he was influ­
enced by previous considerations o f the probability calculus and, especially,
by his attempts to define the logical probability; compare Chapter 9. As
for the connectives, Lukasiewicz did not give any explanation either. All
 Fundamental properties 37

Lukasiewicz offered is an explicit underlining o f the relations between new

matrices, classical truth-tables and the three-valued matrix M 3. Taking
into account, however, the nice algebraic properties that functions cor­
responding to these connectives have7, one may admire the ingenuity o f
Lukasiewicz’s design and intuition.

5.1.1. Every finite algebra (L n ,-», —►) is one-generated by the element

1/n — 1 (or, equivalently, by η — 2/n — 1 = -* 1/n — 1).

Proof, n —2/n—l —►1 /n —1 = 2 /n —1; - « 2 /n —1 = η —3 /n —1;

η —3 /n —1 —►2/n—1 = 4 /n —1 etc.

One may easily notice that M 2 (isomorphic with the matrix o f classical
logic) is a submatrix o f any Lukasiewicz matrix and therefore

E ( M n) C E ( M 2).

Relations between the contents o f finite matrices are established by the

well-known Lindenbaum condition:

5.1.2. (Lukasiewicz and Tarski (1930)). For finite n, m £ N, the follow­

ing conditions are equivalent:

(i) E( Mn) C E(Mm)

(ii) m — l|n — 1 (means: m — 1 is a divisor o f η — 1).

The most interesting fact is, however, that the infinite Lukasiewicz ma­
trices have a common content:

5.1.3. £ ( M Kl) = £ ( M Ko).

Proof. Since L n0 is a subalgebra o f L k1? each valuation in Μ #0 is at

the same time a valuation in M#1. Thus Ε( Μ^λ) C E(M#0).
Assume in turn that a £ E ( M ^ ) . Let Var{a) = {/>i,...,/>*} be the
set o f all propositional variables in a and let Λ be a valuation such that
ha φ 1. Now let h i, h2, . . . , h*,. . . be an infinite sequence o f valuations
in M n0 such that ϋπι*_οο htpi = hpi for each i = 1 , 2 , . . . , & . Using an
inductive argument one may show that the sequence hia, h2a , . . . , hta , . . .
converges at ha. Since ha φ 1, it follows that, for some ht, hta φ 1. This
obviously yields that a φ Ε( Μχ0).

5.1.4. E( M* 0) = Π { Ε ( Μ η) : η > 2, η Ε Ν} .

Proof. The inclusion (C ) is easy: every tautology o f M n0 is a tautology

o f any finite Lukasiewicz matrix as well.

7See e.g.5.2, 5.4, 9.2 and 10.2.

38 5 . Lukasiewicz logics

Conversely, if a £ Ε( Μχ0), then ha φ 1 for some valuation h in

L n0. Where Var(a) is the set o f propositional variables in a , let us put
Gerih(a) = {1 /k — 1 : 1/k — 1 E hVar(a) or k — 2/k — 1 E hVar(a)}
and s — 1 = L C M { k — 1 : 1/k — 1 E Genh(a)} ( LCM being ‘ the least
common multiple’ ). According to 5.1.1, the fraction 1/s — 1 generates the
Lukasiewicz matrix M s. Finally, it immediately follows from the construc­
tion that Λ is a valuation in Ms for which ha φ 1. Thus a ^ E ( M S).

5.2 Definability in Lukasiewicz matrices

As for the three-valued matrix (see 3.2.5.E), all other Lukasiewicz matri­
ces are not functionally complete; this may be proved by exactly the same
argument (non-definability o f constants). Subsequently then, enriched by
all unary constant functions, by virtue o f Slupecki’s criterion, finite ma­
trices became complete. Independently o f this, however, the knowledge o f
whether or not a given function is definable in a matrix under consideration
is a matter o f importance for the purpose o f investigation o f Lukasiewicz
logic. The formulation o f a general definability test for Lukasiewicz matri­
ces turned out to be a hard task. Finally, only in 1951 did McNaughton
succeed in getting satisfactory results in this direction.

5.2.1. (McNaughton (1951)). MN1. A function f ( x i , . . . , x k) is defin­

able in the matrix Μχ0 (by the functions -i and —►) if and only if:

1°. / is a continuous function in [0,1].

2°. There exists a finite number o f distinct polynomials P i , . . . , Ps o f
the form

Pj = + πΐχχι + . . . + rajfcZfc,

where bj and m i , . . . , mk are integers such that for every sequence z i , . . . , ζ*

E [0,1] there is a 3, 1 < 3 < s , such that

f ( x l , . . . , *fc) = Pj ( x i, · . . , Xk)·

3°. For arbitrary ® i , . . . , Xk E [0,1] the inequalities 0 < f ( x i , . . . , Xk) <

1 hold.

MN2. A function f ( x i , . . . , zjb) is definable in Mn (by -· and —►) if and

only if for every ( S\/n — 1, . . . , Sk/n — 1), where 0 < Si < η — 1, if d is the
greatest common divisor o f s i , . . . , s*, η — 1, then d is a divisor o f s, where
s/n - 1 = f ( si /n - 1, . . . , sk/n - 1).

The proof by McNaughton in the infinite case (i.e. the proof o f MN1)
is very complicated; the author makes use o f the methods o f real function
 A xiomatization 39

analysis and not simple set-theoretical constructions. Moreover, the appli­

cation o f the axiom o f choice makes it ineffective. McNaughton, however,
gave an effective proof o f the following partial criterion:

MN1*. Any function o f the form

f ( x i , . . . , Xk) = m in(max( 0, b + rriixi + . . . + ra^z*), 1),
where b and ra* are integers, is definable in the matrix Μχ0.

5.2.2.E. Applying MN2 one may easily show that, given finite n > 2,
the family o f one-argument functions jo , . . . , j n- i

•/ \ _ J 1 for x = i/n — 1
3i\x ) — | o for x φ i/n — 1

and

x => y = min{ 1, (n — 1)(1 — x) + y) = x —►(x —►. . . —►(x —►y))

(η —1) times

are definable in the Lukasiewicz matrix Mn.

5.3 Axiom atization

It can be easily verified that the content o f any Lukasiewicz matrix is closed
under the detachment rule (for the original implication o f the matrix in
question). A proof that finite matrices are axiomatizable was given in
Lukasiewicz and Tarski (1930). Nevertheless, the problem o f formulation
o f a concrete finite axiomatization o f E( Mn) (with the exception o f the
case n = 3) remained open till 1952; see Rosser and Turquette (1952).
Wajsberg (1931) showed that the ( i , —►J-fragment o f Lukasiewicz’s three­
valued propositional calculus may be axiomatized, accepting the rules MP
and SUB, in the following way:

W l. £>—►(<?—►£>)
W 2. ((q - * r ) - . ( p - » r))
W 3. (~>p -> ~<q) —* (q —* p)
W 4. ((p -► -<p) -► p) -> p.

The result extends naturally onto the whole three-valued propositional cal­
culus since the remaining Lukasiewicz connectives are definable by those
o f negation and implication. Slupecki (1936) enriched the set o f primitives
by T (see 3.2.5.E) and adding to W 1-W 4:

W5. Tp — -»Tp
W6. ->Tp -+ Tp

got an axiom system for the functionally complete three-valued logic.

40 5. Lukasiewicz logics

It was Lukasiewicz’s hypothesis that the set Ε(Μ#0) was also axiomat-
izable (Lukasiewicz and Tarski (1930)) and that the axiomatization o f the
infinite-valued propositional calculus together with MP and SUB was as
follows:

LI. p -► («-► p)
L2. (P - » q) -*·((9 -► r) (p -► r))
L3. ((p -*■ 9) 9) ((9 ~^ p ) ^ p )
L4. {-'P - » _l9) -*· (9 -*· P)
L5. ((p —►9) —►(9 —*■p )) -► (9 - * p)·
Lukasiewicz’s hypothesis about the axiomatizability o f the No-valued logic
with axioms L1-L5 was confirmed by Wajsberg in 1931; it is a generally
known fact, though there is no publication by Wajsberg concerning the
matter (see e.g.the footnote on p.137 in Lukasiewicz (1961)). The accessible
proofs o f completeness o f the Lukasiewicz axiom system are credited to
Rose and Rosser (1958) and Chang (1959). It may be o f interest to note
that Meredith (1958) and Chang (1958a) independently and in slightly
different ways showed that the axiom L5 is dependent on the remaining
ones. Thus, L1-L4 together with MP and SUB create the axiomatization
of Lukasiewicz’s Nq-valued propositional calculus.
Several axiomatizations o f finite-valued Lukasiewicz logics (n > 3) were
obtained by way o f extension o f L1-L4 (MP and SUB being assumed) with
the sets o f ‘ specific axioms’ depending on n; to get the rationale consult
5.1.4. Thus, for instance, Grigolia (1977) for the purpose o f axiomatization
o f n-valued Lukasiewicz logics uses two additional binary connectives +
and · defined by a + β = —►/?, a · β = -«(a —►-*β) (in what follows we
write ka as the replacement o f the formula α + α + . . . + α (k times) and
a k for the formula α · a · . . . · a (k times)). Given a finite n > 3, Grigolia’s
axiom system for Ln consists o f the schemes o f L1-L4 and, additionally,
the axioms

Ln5. na —►(η — l ) a
L »6. (n - 1)((-O !y + (a · (j - l)a )),

where 1 < j < η — 1 and j does not divide η — 1.

Tokarz (1974) worked out a method o f axiomatization o f finite-valued
Lukasiewicz logics based on the characteristic functions o f the set Ln (in
[0,1]) and the properties o f the matrix consequence o f Μχ0. In 8.2 we return
to the problem o f axiomatization o f Lukasiewicz logic when discussing a
general Rosser and Turquette construction.

5.4 Algebraic interpretations

The attempts to obtain algebras which would play the same role for Lukasie­
wicz calculi as Boolean algebras do for the classical logic bore at least two
 Algebraic interpretations 41

important constructions: Moisil and Chang algebras.

Moisil’s constructions, dating back to the 1940s, are best adapted to
the finite case; see Moisil (1972). These are algebras with operations cor­
responding to ‘m odal’ connectives o f Lukasiewicz logics8; more specifically,
distributive lattices with Boolean-valued endomorphisms.
A structure

(L, u, n, N , .$1, ^2, . . . , s n_ i , 0, 1)

is an n-valued Moisil algebra if the following conditions are satisfied:

M l. (Z , U, Π) is a distributive lattice with 0 and 1.
M2. N is an involution o f Z (i.e. N N x = x) also satisfying:
N (x U y) = N( x) Π IV(y), N( x Oy) = N( x) U N(y).
M3. The elements o f {$ * }ι< * < η are endomorphisms o f (Z, U, Π):
Sk{ xVy) = Sk( x) Vs k(y) and sk(x fly ) = sk(x) Π sk(y), such that
(i) sk(x) < s* + i(z )
(ii) sk(st(x)) = st(x)
(iii) *k(N(x)) = N s n- k(x)
(iv) N s k(x) Us k(x) = 1
N s k(x) Π sk(x) = 0
(v) If 5fc(a:) = sk(y) for every fc, then x = y.
The simplest example o f n- valued Moisil algebra (given finite n) is the
linearly ordered n-element algebra constructed on the base o f Lukasiewicz
matrix M n:

A A n — (Znj C, Π, 1ΝΓ, 5i, 52, . .. ,Sn_i, 0, 1),

where x U y = x V y, x Π y = x Λ y, N( x) = ->x and

0 when 1 < k < n —j —1

Sk(j/n ~ l ) = 1 when n —j —1 < k < η —1.

Using MN2 in 5.2, it is not hard to verify that the operations sk are de­
finable in Mn. The formulae definining them by means o f -· and —►were
given by Suchofi (1974). Incidentally, the sk are easily described through
Rosser and Turquette’s j (compare 5.2.2.E and 8.1).
Moisil called his structures Lukasiewicz algebras. In the course o f time
the term Moisil algebras came to be used because, for n > 5, Lukasiewicz
implication was not definable in n- valued structures o f this kind. Despite
that fault, however, Moisil algebras proved to be the most fertile source of
applications o f Lukasiewicz logics (see 14.3).
The concept o f MV algebra, tailored by Chang (1958b), resorts to math­
ematical group theory. Consider an algebra

8Compare 2.3 for tlie discussion n = 3.

42 5. Lukasiewicz logics

Λ — (-4, + , ·, —, Ο, 1),

where + and · are binary operations, — is a unary operation and, finally, 0

and 1 are two different constants in A, 0 φ 1'. Let us introduce, moreover,
the following notations:

xU y = x - y ~ + y x D y = (x + y~) · y.

The algebra A so defined is called an M V algebra if the following conditions

are satisfied:

C l. ® + j/ = y + ® C l* . X -y = y - x
C2. * + (j/ + z) = (* + y) + z C2*.

II
C3. x + x~ = 1 C3*. x · x~ = 0
C4. x+ 1= 1 C4*. x ·0 = 0
C5. x+ 0= X C5*. x ·1 = x
C6. (x + y)~ = x~ -y~ C 6*.

1
+
II
C7. (x- )- = X C 8. 0- = 1
C9. x uy = y ux C9*. x Πy = y Πx
CIO. X U (y U z) = (x U y) U z CIO*. x Π (y Π z) = (x Π y) Π z
C ll. x + (y Π z) = (x + y) Π (x + z)
C ll* . x · (y U z) = (x · y) U (x · z).

The simplest example of the MV algebra is an arbitrary (finite or not)

Lukasiewicz matrix - in this case x + y = ->x y, x · y = -i(z —►-iy),
and U, Π, — are the functions o f disjunction, conjunction and negation,
respectively. The problem of adapting MV algebras to finite cases (i.e. of
introducing such modifications which would allow algebras characteristic
for particular cases o f Lukasiewicz finite-valued logics to be obtained) was
taken up in 1973 by Grigolia (see Grigolia (1977)) who introduced the
notion o f M Vn algebra: an MV algebra A is an MVn algebra provided that
the following additional conditions are satisfied:

C 12. (n — l)x + x = (n — l)x C 12*. χη~λ · x = z n-1

and for n > 3:

C13. { ( j x ) · ( * - + ((j - I ) * ) ” } " ” 1 = 0

C13*. (n - 1){*J + ( * - - (®i " 1) · ) } = 1

where 1 < j < η — 1 and j does not divide η — 1 (see 5.3 for missing
notations).
The common property o f the two algebraic interpretations o f Lukasie­
wicz logics is the way in which they are related to Boolean algebras. Every
Moisil algebra as well as each MV algebra contains some Boolean subalge­
bra. The set o f Boolean elements of an arbitrary n-valued Moisil algebra
 Algebraic interpretations 43

is equal to

{$*(:ε) : x £ L, 1 < k < n } ,

and the set o f Boolean elements of an MV algebra Λ to

{ x E A : x + x = x} (or, equivalently, to { x E A : x · x = a?}).

From among the purely algebraic results for Moisil and Chang algebras
those o f the greatest importance for logic are the representation theorems
which state that any algebra is (isomorphic to) a subdirect product o f a
class o f linear algebras, i.e. definitional variants of Lukasiewicz matrices.
Employing MV algebras Chang (1959) gave a purely algebraic proof of
completeness o f the No-valued Lukasiewicz logic. The method was also
transferred to the case o f all finite (Lukasiewicz) logics by Grigolia (1977)
(compare 5.3).
6. Post logics

Post (1920, 1921) is the author o f a class of finite-valued, functionally com­

plete propositional logics. In his constructions Post was inspired by the
well-known formalization of the classical logic presented in Principle, Math-
emaiica (Whitehead and Russell (1910)) and by the truth-table method o f
verification o f logical laws.
In this chapter we survey the work o f Post and his followers. Given
the history o f the subject, special attention will be put on recent algebraic
interpretations and formalizations o f Post logics which, as one may see in
14.3 and 14.4, are oriented almost entirely towards applications.

6.1 Post matrices

The fundamental, many-valued constructions of Post are connected with
two propositional connectives primitives in the Principia Mathematical Post
logics negation (-«) and disjunction (V). For any natural n > 2 Post builds
an n-valued logical algebra on the linearly ordered set o f objects

P n = {f 1> i ^n}

(ti < tj iif i < j ) equipped with two operations: unary cyclic rotation
( cyclic negation) and binary disjunction V, defined in the following way:

t {+1 if i φ n
->ti = t i V t j —t m a x ( i j ) ·
ti if i — n

The disjunction function fixes on a natural and entirely intuitive mean­

ing o f the disjunction connective, typical at least for most known many­
valued (including infinite) constructions. In plain
terms, the logical value o f disjunctive proposition
a —■a
equals the greater o f the values of its components. The
function o f cyclic rotation permutes, in some specified fi <2
manner, the set Pn and the negation corresponding to t2 *3
it, the case n = 2 being excluded, is quite special - • •
compare the table here. It is just the fact o f com­ tn h
bining the latter with an appropriate binary function
 Post matrices 45

of algebra on Pn that warrants the functional completeness o f that alge­

bra, i.e. it ensures that by means of the primitive functions, every finite-
argument function can be defined on Pn, including constant functions and
hence the objects ti, t^ . . . , tn. For a given finite n > 2 the algebraic
structure:

T n = ({<1,<2>· · · , < « } , “ ■) V)

will be called an n-valued Post algebra.

The matrix Pn naturally associated with the algebra Vn·

* » = ({ *!, *2, . . · , * » } , V, {* „})

will, in what follows, be referred to as the (basic) n-valued Post matrix. It is

easily seen that the two-valued Post matrix is isomorphic to the negation-
disjunction matrix for the classical propositional calculus. To check it, one
must replace t\ in P 2 by the falsity (0) and tz by the symbol o f truth ( 1). Si­
multaneously, however, the matrices Pn for n > 2 are totally incompatible
with the mentioned classical matrix, which is the result o f the non-standard
mode o f the negation connective. Hence for n = 3, for instance, £3 could
be the only counterpart o f ‘ truth’ with respect to the adopted interpre­
tation o f disjunction but then ti would have to correspond to ‘falsity’ as
-i<3 = t i , which should not take place because — «^3 = ->ti = Φ ^3· A
contradiction.
It is remarkable that among the laws o f n-valued logics determined by
Post matrices Pn there are many-valued counterparts o f some significant
classical tautologies expressed in terms o f negation and disjunction connec­
tives, including the ‘generalized law o f the excluded middle’ :

p V —1p V ~~*~~*p V . . . V —1 · .. . —1 p.
(n —1) times

On the other hand, however, the connectives of implication and conjunction

introduced up to the two-valued pattern (1.3) evidently stray away from
their classical counterparts. Thus, putting

<x —>β = ->a V /3,

α Λ β = ~■(—■(* V —1/3) and a = β = (a —>β ) Λ (/3 —►a ),

we get functions bearing some unexpected properties: for example, it can

be shown that the formula

—>(p Λ —*p Λ —■— Λ — Λ . -ι ρ)

(η—1) times
does not belong to E(Pn).
46 6. Post logics

6.2 Interpretation
Post did manage to present a semantical interpretation for his non-standard
matrices Pn providing the following, Euclidean in spirit, construction o f the
‘spaces’ E n~1:

(1) Elements P £ E n_1 are (n —l)-element tuples o f ordinary two-valued

propositions (represented by lower case letters), P = (pi, · · · >Pn-i)i
subject to the condition that the true propositions are listed before
the false.
(2) -i P is formed by replacing the first false element o f P by its denial,
-.P = (-*(piAp2A .. .A p „_i), --(piA paA .. .A p „_i)A (p iV p 2), · · . , ” ·(ΡιΛ
P2 A . . . A pn—l) A (pn-2 V P n-i))? the connectives on the right-hand
side being the usual (classical) connectives; but if there is no false
element in P , then all o f them are to be denied, in which case ->P is
a sequence o f false propositions.
(3) W hen P = (p i,p 2, . . . ,p n- i ) and Q = (qu q2, . . . ,gn- i ) , then P V Q =
(P iV g i,p 2Vg2, · · · ,Pn-iV<ln-i) (with the right-hand side V as above).

The mapping i : P n_1 —►Pn

i(P) = t i iff P contains exactly (i — 1) true propositions

establishes an isomorphism o f (P n_1, V,-«) onto the Post algebra P „.

The interpretation discussed shows, among other things, that similarly
stated logical values appearing in diverse matrices Pn are, regarding the
author’ s intention, objects different from each other. Therefore, for the
sake o f precision the symbols o f logical values U should always be indexed
by a parameter assigning them to a given matrix.
Apart from the basic matrices Pn, Post also defined a family o f func­
tionally incomplete n- valued implicative matrices with k designated values
(1 < k < n ):

Pnk — ({<1, ^2? · · · » }i *nfci {^n—fc+li · · · » })i

where

' tn if i< j
ti >nJs tj — < tj if i> j and i > n —k + 1
, tn-i+j if i> j and i < n — k + 1.

The matrices o f that family can serve as a tool for describing the implication
connectives o f other, known, many-valued logics. So, for instance, —>ni and
- η>n n_ ! are the n-valued implications of Lukasiewicz and Godel respectively
(to obtain the implicative Godel matrix one ought to reduce the set of
distinguished values to {<n}? having previously compiled the truth-table
for implication).
 Algebraic form o f Post logics 47

The fact that Post designated many (at a time) logical values induced
a significant impulse in the 1920s, for the emerging theory o f logical ma­
trices. It seems that other originators of many-valued logics ignored that
possibility, or did not attach great importance to it.

6.3 Algebraic form of Post logics

The functional completeness o f Post matrices makes logics based on them
easily characterizable by means of algebras; there is total freedom o f choice
o f primitive operations. A generalization o f the concept of characteristic
function leads to further precision of intuition concerning algebras that play
for Post logics a role similar to that o f Boolean algebras in the classical logic
(see 1.6). For a given U and n > 2, let P(U, n) be a family o f ‘ n-valued
sets’ X identified with functions X : U —► X( a) = ti means
6a is an element o f X ’ has logical value f,·. The relation C defined for any
X , Y e P ( U , n ) by

X C Y iff X( a) < Y(a) for each a £ U

orders partially the family P(U, n). Taking this into consideration, we can
introduce the operations U and (Ί on Ρ(Ϊ7, n) as in 1.6. As everyone is quick
to point out, the algebra (Ρ(Ϊ7, n), U, Π ) is a distributive lattice with unit
U and zero 0. Special elements o f that family are the constant functions
βο, · · · j ^n —li

e,(z) = ti+ 1 for any x E U (i E {0, 1 , . . . ,n - 1})

forming a chain, 0 = eo < e\ < . . . < e „ _ i = U.

An important subfamily o f P(U,n) is the class C(17, n) o f ‘ two-valued’
sets, i.e.such that X ( a ) E ( { 0 , 1 } ) for any a. That class contains all
the complemented elements o f the family P(U, η), X E C(U, n), iif there
exists —X E P{U, n) such that X U —X = U and X Π —X = 0, and is
closed under lattice operations. Inconsequence, (C(U, n), U, Π , —, U, 0),
with — being the operation o f complement, is a Boolean algebra.
Consider, subsequently, the set o f unary operations Dk (k = 1, . . . , η —1)
on P(U, n) defined as follows:

η / tn when X( a) > tk
Dk(X)( a) = | ti when χ ( α ) < ^

That each of the sets Djg(X) belongs to C(U, n) is obvious. Moreover, it

can be easily checked that for every X E Ρ(Ϊ7, n)

(D ) X = ( D1( X ) Π e x) U ( D2( X) Π e2) U . . . U ( D n_ i ( X ) Π βη_χ).

The construction just presented elucidates basic intuitions underlying

the abstract theory o f Post algebras. Besides, the algebras o f many-valued
48 6. Post logics

sets play an important role in this theory.

The concept o f a Post algebra o f order n (n > 2) was introduced by
Rosenbloom (1942) who defined Post algebras by means o f the rotation
the disjunction V and some auxiliary functions. Subsequently, it has un­
dergone several modifications resulting from both theoretical and practical
studies (see e.g. Dwinger (1977)). O f particular importance among them
is the lattice-theoretical characterization (Epstein (I960)) fixing a creative
direction for the studies. The equation defining these algebras is due to
Traczyk (1964): a Post algebra of order n (n > 2) can be presented as an
algebra having two binary U, Π, n unary operations —, D\, . . . , Α ι - i and
constants eo, . . . , en_ i

C, — (Z, U, Π, , Di, · · · j Dn—h €0, · · · j ^n—l)j

satisfying the conditions:

(1) (Z , U, Π) is a distributive lattice with 0 = eo, and 1 = en_ i

(2) —(x U y) = —x (Ί —y , -----x — x
(3) e,· (Ί ej = e,· if i < j
(4) Di(x U y) = Di(x) U Di(y), Di(x Dy) = Di(x) Π Di(y)
(5) Di(x) U —Di(x) = 1, Di(x) (Ί —Di(x) = 0
(6) Di(x) (Ί Dj(x) = Di(x) if i < j
(7) D i (- z) = - D n-i(x)
(8) Di(ej) = 1 for i < j, Di{ej) = 0 for j < i
(9) x = { Di {x) (Ί e i) U ( D 2 ( x ) (Ί e2) U . . . U ( Dn- i ( x ) (Ί en_ i) .

It can be proved that the set C ( L ) = {Di(x) : x G L , i G { 1 , . . . , η —1 }}

is closed under lattice operations and the structure (C (Z ), U, Π, —, 1, 0)
is a Boolean algebra. Apparently, each Post algebra o f order 2 is a Boolean
algebra as well.
The simplest Post algebra o f order n is the structure based on the set o f
logical values {tfi,. . . , £n} having the operations x U y = m a x { x , y }, x(~)y =
mi n { x , y } and

pi ( \ _f tn if a? > t{ _
-DiC*) — \ if x < t i * < -* « -< + 1·

In the end, every Post algebra o f order n is isomorphic to some field o f sets
P(U, n) (Wade (1945)).

6.4 Axiomatization of functionally complete sys­

tems of n-valued logic
The original (-«, V) systems o f Post logic have not yet been axiomatized.
However, the problem o f their axiomatizability has for years been taken
for granted, so Slupecki (1939b) constructed the largest possible class of
 Axiomatization o f functionally complete systems o f n-valued logic 49

functionally complete finite logics and gave a general method of their ax­
iomatization. From this it evidently follows that Post logics are also axiom-
atizable although the problem o f providing axioms for their original version
still remains open.
The Slupecki matrix Snk (n being a given natural number, 1 < k < n)
is of the form

Snk — ( { l j 2, — , n}, — R, 5, {1,2, — , &}),

where —►is a binary (implication), and iZ, 5 unary operations defined in

the following way:

y if 1 < x < k
x-^ y-
1 if k < x < n

x+ 1 if 1 < x < η — 1
R(x) =
1 if x = n

2 if x = 1
S(x) = 1 if x = 2
x if 3 < x < n.

Functional completeness o f each o f these matrices results from the criterion

3.2.4. R and 5 are two o f Picard’s functions, and in order to define the
third, it suffices to put:

1 if a? = 2
H x = (x —►iZ(a? —►a?)) —> 5a? for k = 1, then Hx —
a? if a? φ 2

1 if a? = fc
Hx = R(x —►x) —►x for fc > 1, then Hx =
a? if a? φ k.

Slupecki produced an effective proof o f axiomatizability o f E(Snk) (any

pair (n, k) as above) giving a long list o f axioms formulated in terms of
implication and special one-argument connectives defined through the su­
perpositions o f iZ, 5 , and H. The chief line o f approach here is to make
capital o f the character o f implication9, which can be classically axioma-
tized (1.4) using the detachment rule. Slupecki extends MP onto the whole
language, taking the Lukasiewicz formula ((p —►q) —> r) —►((r —►p)
—►(5 —►p)) as the only axiom for implication, and provides an inductive,
combinatorial proof o f completeness.
At present attention is generally focused on the formalization o f Post
logics based on Rousseau algebras. Rousseau (1969) noticed that any
Post algebra o f order n is a pseudo-Boolean algebra (see Rasiowa (1974)).

9 Compare 8.1 and 8.2.

50 6. Post logics

Consequently, he proposed a definition o f Post algebra (o f order n) which

turned out to be exceedingly important from the point o f view o f applica­
tions (see 14.4). A new operation appearing in this definitional version is a
binary operation o f relative pseudo-complement — which can be described
on the set o f constants { e o , . . . , en_ i } as follows:

e, - * ^ jf
= [ ej
en_ i i
when < j
otherwise.

The system o f n-valued propositional calculus corresponding to the Rous­

seau algebras (given n) is determined in the language with connectives
V, A, = , £>i, . . . , -Dn-ΐί and e0, en_ i (for the sake o f brevity
the algebraic symbols used here have a new meaning; e.g. e,· now refers
to logical constants, i.e. zero-argument connectives). Its axioms are the
schemes o f (H 1)-(H 10) in 12.1 and, for every i = 1 , . . . , η — 1,

(Pll) D i ( a V p ) = (Di<xV Dip)

(P12) Di(a Λ β) = ( A α Λ Ό φ)
(P13) D i ( a —* β ) = ((DxOt —* Ό φ ) A {D 2ol — Ό 2β ) Λ . . .
Λ ( Α « -► Ό φ ) )
(P14) = —*Dia
(P15) D iD ja Ξ D ja
(P16) Diej when i < j and “ A e j when i > j
(P17) a = ( D j a Λ βχ) V ( D 2a A e2) V . . . V ( D n_ xa A en_i )
(P18) D i a V ->Dia.

And, apart from MP, an extra inference rule is

(T>n) A,_i«

The predicate calculi for Post logics are built in a standard way on the
basis o f propositional calculi. The most systematic studies o f them carried
out so far are due to Rasiowa (1974).
7. Three-valuedness of Kleene and
Bochvar

Kleene (1938, 1952) and Bochvar (1938) are the founders o f the original
three-valued constructions motivated epistemologically by the indetermi­
nacy or absurdity o f some propositions at a certain stage o f scientific in­
vestigation. The assumed status of the third logical value is somewhat
different from the classical values o f truth and falsity. The resulting logics,
both propositional and predicate, are simulaneously allied to and distant
from the classical logic.
In what follows the fundamentals o f these epistemologically established
systems of logic are discussed. Further, we give an overall exposition of
two related constructions motivated by partiality in associating (classical)
truth-values to sentences: super valuations and the partial logic.

7.1 Logic of indeterminacy

Kleene (1938), being inspired by the studies o f the foundations o f mathe­
matics, creates a logic that renders the analysis o f partially defined predi­
cates (propositional functions) possible. A simple example o f such a predi­
cate is a mathematical property P given by the equivalence

P( x) if and only if 1 < 1 /x < 2,

where, x is a variable ranging over the set of real numbers. It seems apparent
that for some definite values o f x, the propositional function P ( x ) is true,
false or undetermined:

true if 1/2 < a < 1

Proposition P(a) is undetermined if a = 0
false otherwise.

The starting point of Kleene’s construction consists in considering the

propositions o f the ‘ third’ category, i.e.such as whose logical value (o f truth
or falsity) is undefined, undetermined by means of accessible algorithms,
or not essential for actual consideration. Besides the classical values of
truth (t) and falsity (f) he introduces the value o f undefiniieness (u) and
52 7. Three-valuedness o f Kleene and Bochvar

characterizes standard connectives with the help o f the following truth-

tables:

a -»a f u t V f u t
f t f t t t f f u t
u u 11 u u t u u u t
t f t f u t t t t t

Λ f u t = f u t
f f f f f t u f
u f u u u u u u
t f u t t f u t

Let us note that, similar to Lukasiewicz, the behaviour o f these connectives

towards t and f remains unchanged. Simultaneously, however, the classical
interdefinability relations do hold: each pair o f formulae: a —►β and ->aV/3,
o:V j3 and -»a —►β specifies the same truth-tables in Kleene’s logic.
Kleene distinguishes the value of truth (t). Therefore, the matrix

% 3 = ({f, U, t }, V, A, { t } )

with operations introduced by the above truth-tables will be called the

(three-valued) Kleene matrix. The relevant feature o f the propositional
logic obtained on the basis of the matrix K 3 is its non-tautological charac­
ter: E{ Kz ) = 0 since any valuation which assigns a u value to every propo­
sitional variable assigns u to any formula. Thus, in the case o f Kleene logic,
we face quite an unusual situation. Some intuitive objections are aroused
by the fact that a ‘ conservative’ extension o f two-valued logic leads to the
rejection o f all classical tautologies, with no exceptions, including even as
‘ obligatory’ as p —►p and ρ Ξ ρ .
The most accurate and compatible interpretation o f Kleene’s views of
the connectives o f K 3 was given by Korner (1966). Korner introduced a
non-standard abstract notion of an inexact class as an object characterized
by a non-definite classifying procedure determined by a partial definitional
algorithm. The mathematical description of the algebra o f inexact classes,
in a natural way, leads to Kleene’s logic.
An inexact class o f a given non-empty domain A is identified with a
three-valued ‘ characteristic function’ X p : A —►{ —1 , 0 , + 1 } correspond­
ing to the partition o f A generated by a partial definition D( P ) o f some
property P (o f elements of A):

—1 when P(a) according to D( P) is false

{ 0
+1
when P(a) is D(P)-undecidable
when P(a) according to D( P) is true.
 Logic o f indeterminacy 53

The operations o f union (U), meet (Π) and complement ( —) are defined as
follows:

( X U T )(a ) = m a x {X (a ), T (a )}
( X n y j ( a ) = m in {X (a ), Y (a)}
(-■ * )(«) = - * ( « ) ·

The family N { A } o f inexact classes of a given set A is closed under the

operations introduced, and the algebra (JV{ j4}, U, Π, —) is a distributive
lattice with an additional unary operation that meets the following:

(M ) - ( x u y ) = - x η -Y -(x η γ ) = -x u - y
(I) - ~ x = x.

In the literature, algebras o f this kind are called de Morgan lattices (see
e.g. Gratzer (1968)).
The link between inexact classes and Kleene logic is apparent: if —1,
0, +1 are interpreted as f, u, t, respectively, then the counterparts o f al­
gebraic operations U, Π, — are the connectives of disjunction, conjunction
and negation o f the logic of inexact predicates. Nevertheless, Cleave (1974)
showed that the ctranslation’ o f the inexact predicate logic into the algebra
o f inexact classes dictated by Lindenbaum construction (see 1.6) is feasible
only after a change in the notion o f a matrix consequence operation. Us­
ing the values —1, 0, +1 in place o f f, u, t, Cleave defines an appropriate
relation i—>c for this aim, putting:

X ·—►c if and only if πιίη{υ(β) : β £ X } < v(a) for any valuation

v : For —►{ —1, 0, + 1 } interpreting -i, V and Λ as the
Kleene connectives.

Let us note that the equivalence relation « necessary for getting Linden­
baum algebra may then be introduced as

a « β if and only if a ■—>c β and β ■—>c <*.

Several years later, in the monograph Introduction to Metamathemat­

ics (Kleene (1952)), Kleene refers to the above connectives as strong and
introduces the weak connectives o f implication, disjunction and conjunc­
tion (negation and equivalence remaining unchanged) characterized by the
truth-tables:

—> f u t V f u t Λ f u t
f t u t f f u t f f u f
u u u u u u u u u u u u
t f u t t t u t t f u t
54 7. Three-valuedness o f Kleene and Bochvar

The motivation stems from arithmetic. The novel truth-tables are to de­
scribe the employment o f logical connectives in respect o f those arithmetical
propositional functions whose decidability depends on the so-called effec­
tive recursive procedures. The ‘ weak connectives’ truth-tables are consti­
tuted according to the principle o f saying that the appearance o f the value
u in any place results in the whole propositional context taking u. Its
arithmetical equivalent states that indeterminacy occurring at any stage o f
computation makes the whole procedure undetermined.

7.2 Bochvar logic

Bochvar (1938) worked out his three-valued logical system intending to
solve paradoxes emerging within the classical logic and the set theory based
on it. His construction has two planes, one coinciding with the weak logic
of Kleene, and the other being just the classical logic.
Bochvar’s fundamental conception is built on the division o f proposi­
tions into sensible and senseless, and then on the ‘mapping’ o f this division
into a two-level formal language. A proposition is meaningful provided that
it is either true or false, all other sentences exempted from such evaluation
being meaningless or paradoxical The propositional language o f Bochvar
logic is equipped with two kinds of connectives: internal and external. Both
groups contain counterparts of the ordinary connectives o f negation, impli­
cation, conjunction, disjunction and equivalence. The internal connectives
are conservative generalizations o f the classical ones and are interpreted
similarly: in what follows they will be denoted also as -i, — V , A , = , re­
spectively. The external connectives are metalinguistic, and with their help
relations between the logical values (of truth and falsity) o f propositions
are described. They incorporate expressions o f the type: ‘ ... is true’ and
‘ ... is false’ . The specification mentioned below contains all the external
counterparts o f the standard connectives together with their notation and
the intuitive way o f understanding them:

external negation: ->*a 4a is false’

external implication: α —* β 4if a is true then β is true’
external disjunction: aV * β ca is true or β is true’
external conjunction: a A* β 4a is true and β is true’
external equivalence: a=* β ca is true iff β is true’ .

W ith the view o f characterizing both groups o f connectives, Bochvar

considers three-valued matrices with the values t (truth), f (falsity) and u
(meaningless) - here the symbolic representation of Kleene is applied to
expose the common features of the two constructions. The truth-tables of
all internal connectives have been compiled according to the rule which is
a rejoinder o f Kleene’s principle: ‘ every compound proposition including at
least one meaningless component is meaningless, in other cases its logical
 Bochvar logic 55

value is determined classically’ . It is not difficult to see that the internal

connectives o f Bochvar ‘ coincide’ with the weak connectives o f Kleene -
in the notation adopted the respective truth-tables are simply identical.
The clue to make Bochvar’s construction natural and compatible with the
internal logic matrix description o f external connectives is provided by the
introduction o f external assertion ‘a is true’ (A*) (see
the truth-table here). The definitions o f the intu­
itively characterized (external) connectives, so far, are a A*a
as follows: f f
u f
-»*a =df ” >A*a a —►* β = df A*a —►Α+β t t
otV* β = df A*a V Α*β a A* β = df A*a A Α*β
a =* β —df A*a = Α*β.

After some easy calculations we get the following truth-table description

o f the second collection o f Bochvar’s connectives:

*
|<*

a -»*a f u t f u t
f t f t t t f f f t
u t u t t t u f f t
t f t f f t t t t t

A* f u t —* f u t
f f f f f t t f
u f f f u t t f
t f f t t f f t

Since it is assumed that t is the sole distinguished element, the matrix

B3 = ({f,u ,t}, V, Λ, = , -» * , V*, Λ*, = *, { t } )

will be called the Bochvar matrix. Its two reducts, i.e. matrices

B 3i= ({f,u ,t}, V, A, = , { t } )

B3e = ({f,u ,t}, -» * , V*, A*, = * , { t } ) ,

‘ describe’ internal and external logics, respectively. The first o f these ma­
trices is isomorphic with the ‘ weak’ matrix o f Kleene, .K3*. And, in spite o f
a slightly distinct initial motivation, it portrays the same inferential logic
(i£(J33,·) = 0). In turn, the matrix B$e ‘identifies’ (does not distinguish)
u and f, whereas the behaviour o f its functions with regard to f and t is
classical. Consequently, the external logic o f Bochvar is the classical logic
(compare 4.1).
It is noteworthy that the set of rules of the consequence relation (=Bsi
56 7. Three-valuedness o f Kleene and Bochvar

(and thus ^=K3m) consists o f some special rules o f the classical logic. Namely,
for any consistent set o f formulae X C For (i.e.such that C 2 (X ) Φ For):

X [=Ba. a if and only if X |=m 2 a and Vhr(a) C Var{X).

7.3 Partial logics

The ideas o f Kleene and Bochvar have been reprised in some recent logical
constructions motivated by the partiality in associating truth and falsity to
sentences. Within these frameworks, sentences which are ascribed neither
truth nor falsity are supposed to form a third classification considered as a
truth-value gap. This type o f approach leaves a lot o f room for the possible
interpretation o f the new logical category. Therefore, the variety o f prob­
lems related to partiality becomes quite large: it contains topics concerning
presuppositions, conditional assertion, sortal incorrectness, referential fail­
ure and the like; see Blarney (1986).
Essentially, there are two approaches to truth-value gaps: ‘supervalu-
ation’ and the matrix approach. The first, due to van Fraassen (1966),
was intended as a tool to formalize linguistic presuppositions which are,
roughly, the conditions necessary for some (other) proposition to be at­
tributed with a logical value. The rationale o f the second approach is that
the truth-value gap is considered on a par with two classical values and,
consequently, three-element matrices are constructed.
The heart o f van Fraassen’s, non-matrix proposal is the method o f su­
pervaluations. A supervaluation assigns to a (com pound) formula, some
parts o f which lack truth-value, a value which all classical valuations would
assign, if there is such a unique value, and otherwise no value. Thus, a
super valuation is like a direct product o f matrices, or as we might say,
valuations in matrices, the difference concerning the assignment o f falsity
(compare 3.3).
It is not difficult to notice that all tautologies o f CPC are preserved
under any supervaluation and that, similarly, all contradictions are settled
as false. Some scholars consider this as an advantage o f van Fraassen’s
approach in comparison to Kleene’s and Bochvar’s original proposals. On
the other hand, no contingent formula is assigned a value by a supervalu-
ation. One consequence o f this is that van Fraassen’s systems differ from
the classical logic by inferences or, equivalently, by the accepted rules o f
inference.
The definition o f a supervaluation is relaxed if instead o f the whole class
o f classical valuations its subclass K is taken. In this way different systems,
or supervaluation schemes, are obtained. We then say that a supervaluation
is induced by K . In turn, when used for presuppositions, supervaluations
are relativized to sets o f formulae expressing them. Given a set o f formulae
X , a supervaluation induced by X is a partial valuation vs such that
 Partial logics 57

ν5(α ) = 1 iff v(a) = 1 for all valuations v such that v ( X ) C {1 } ,

v5( a ) = 0 iff v(a ) = 0 for all valuations v such that ν ( Χ ) C {1 },
v5( a ) is undefined otherwise.

The relativizations have the consequence that some classically contingent

formulae are true, i.e. K -tiue or X-true, and, accordingly, that the set of
rules o f inference is modified. In this way we get a variety o f super valuation
schemes.
Introducing matrices with an element representing truth-value gaps has
usually relied upon some specific ideas. Thus, for example, W oodruff (1970)
was attracted by connectives that yield compounds which are true or false
even when every constituent lacks value. Consequently, he considered a
family o f gaps admitting propositional connectives sufficient to define every
possible three-valued connective10. Belnap (1970) used such matrices in
the study o f conditional assertion - consequently, he interpreted the third
value as ‘ no assertion’ . Finally, Blarney (1986) stated that the justification
of the ordering values as well as the choice o f connectives rely upon the
monotonicity, a property which stems from the postulate o f persistence of
logical values o f truth and falsity. Since this approach is the most general
we shall focus our attention on it. In the following, we adopt the commonly
accepted partial logicians’ notation which labels J_, T, * as the falsity, the
truth and lack o f value, i.e. a gap, respectively.
A simple partial logic is semantically an extension o f the strong Kleene
system. The starting point consists in accepting the intuitive features o f
the classical connectives specified in 1.1, dropping simultaneously the as­
sumption about the universality of association o f _L and T to formulae.
This leads to the following translation of the original Kleene tables:

Ο! 1 * τ V. 1 * τ
1 τ _ι_ τ τ τ -L 1 * τ
* * * * * τ * * * τ
τ 1 τ 1 * τ τ τ τ τ
Λ 1 * τ = 1 * τ
1 1 1 1 1 τ * 1
* 1 * * * * * *
τ _ι_ * τ τ 1 * τ
The vocabulary o f the language o f propositional partial logic, in addi­
tion to the usual connectives, includes three constants T, _L, *, and other
connectives. To get a partial logic we impose first the following order □ on

10Thus, a functionally complete system o f three-valued logic.

58 7. Three-valuedness o f Kleene and Bochvar

_ .rr f both if x = T , then y = T

xQ v lff 1 and if z = _L, then y = ± .

Then, an n-argument function / : {_L,*, T } n —►{ ± , * , T } is monotonic

whenever the following condition is satisfied:

(*) if X{ □ yi for all i, 1 < i < n,

then / ( x i , . . . , x n) C f ( y i , . . . ,y n).

One may easily verify that functions corresponding to all Kleene’s strong
connectives are m onotonic11. Blarney (1986) completes the bunch o f con­
nectives by also considering two other binary functions on { _ L, * ,T }: in­
terjunction xx and transplication / . Their tables are as follows:

XX ± * T / _L * T
-L ± * * _L * * *
* * * * * * * *
τ * * T T _L * T

It turns out that -», V, A, = together with xx and / form an exhaustive

list o f all possible monotonic functions on {_ L, * ,T }. Thus, the ‘partial·
propositional logic can be seen as the total logic o f three-valued monotonic
modes o f sentence compositions; cf. Blarney (1986).

7.3.l.E . The following one-argument functions f i and f t

Λ ( ± ) = ± , Λ (*) = Λ (Τ ) = τ
M - l) = Μ * ) = ± , M V = τ

are non-monotonic. Note that f i is the function corresponding to Bochvar’s

assertion A *. It also turns out that monotonicity is not a property o f any o f
the ‘metalinguistic’ connectives defined on the set {_ L,* ,T }. Such are, for
example, W oodruff’s favourite connectives (including Bochvar’s external
connectives).

Monotonicity may be naturally related to valuations and consequence.

The usage o f □ is extended to define a relation between two admissible
valuations o f the language considered, say v and w , as follows:

v□ w iff v(p) □ w(p) for every propositional variable p.1

11 Constant functions ± , *, T are monotonic as well.

 Partial logics 59

In the latter case, mono tonicity leads to the concept o f two consequence
relations h T and I-1, respectively truth and falsity preserving. For any « ,/? :

(T ) a hT β iff for every valuation v, v (« ) = T implies ν (β ) = T,

(_L) a hj_ β iff for every valuation v, v(a) = _L implies ν (β ) = _L .

It is obvious that two (partial) valuations v and w such that v □ w and

w □ v coincide. On the other hand, the intersection o f the two relations
just defined, = = h T Π l·1 , is the consequence relation such that for any
a and β:

„ ~ iff / ®(a) = T implies ν(β ) = T

^ \ v(a) = J_ implies ν(β ) = J_.

The idea o f supervaluation may be explained in terms o f the relation □

between vacations. Given a (partial) v , a supervaluation evaluation v5(a )
o f a formula a is then defined by:

ν5(α ) = T iff ty(a) = T for all total w such that v□ w

vs(a ) = J_ iff w(a) = J_ for all total w such that v□ w .

It is obvious that any such vs is monotonic with respect to ‘classical’ formu­

lae a : v □ ν' implies v5(a ) □ ^£(<*)· On the other hand, in general there
is no monotonicity o f sentential modes 0 ( p i , . . . , p n) since we may have
v { 4>i) C ν(χ<), 1 < i < n, and not ν{φ{-φ1, . . . , φη)) C ν(φ (χ ι , . . . , χ „ ) ) .

The following example shows the non-monotonicity o f the excluded mid­

dle as well as the intensional character o f vs:

7.3.2.E. Consider any supervaluation vs on the propositional language

containing disjunction V and negation -> such that v5(p) = v(p) = * and
v5(g) = v(q) = T. It is easy to see that v5(pV-»p) = T, while vs(pV->q) = *.

The problem o f partiality is more sophisticated when considered for

the first-order languages. Then, all traditional queries on definite descrip­
tions and sentences whose truth-value attribution depends on existential
assumptions or presuppositions are in vogue; consult Blarney (1986). Fur­
thermore, a suitable theory o f interpretation o f the first-order language is
needed. To meet this expectation Langholm (1988) gave an extensive study
o f partial model theory.
8. Standard properties of
many-valued constructions

The interpretation principle of propositional languages stated in 3.1 is a

generalization o f the truth-functionality principle constituting the base of
matrix description o f the classical propositional logic. The link in question
is somehow manifested by the fact that most o f the many-valued construc­
tions known are "conservative’ extensions o f the classical logic matrices.
Rosser and Turquette (1952) determined the conditions that make finitely
valued propositional logics resemble more the CPC, and hence simplified
the problem o f axiomatization and also the question o f their extension to
predicate logics (compare 11.3).
The general conditions resolve themselves into the above-mentioned
principle o f the extensional interpretation o f propositional languages in
matrices o f the form

— (^n?-^fc)>

where lin = (f£n, / i , / 2, . . . , /m)> En — { 1, 2, . . . , n }, £)& = { 1, 2,.

π > 2 is a natural number and 1 < k < n. Adhering to the assumption of
Rosser and Turquette, the natural number ordering conveys a decreasing
degree o f truth: regardless o f the values o f n and fc, 1 refers to "truth’
and n takes the role of "falsity’ . It is noteworthy that the postulate o f
ordering deviates from the intentions o f some authors o f original systems
o f many-valued logics; see Chapter 7. Detailed conditions describe connec­
tives whose occurrence in many-valued logics is methodologically desirable,
and they are called standard. The counterparts of the classical functors ap­
pearing there are characterized by the division o f the set o f logical values
into two subsets o f designated and undesignated values.
The standard conditions for propositional connectives are listed below.
Then, an insight into Rosser and Turquette’s general method o f axiomati­
zation o f (standard) many-valued systems is given. The chapter closes with
a discussion o f the standard properties of matrix consequence operations.
 Standard conditions 61

8.1 Standard conditions

Let -i, — V, A, = be (primitive or definable) functions o f a given ma­
trix Mntk describing propositional connectives that in the logic constructed
represent negation, implication, disjunction, conjunction and equivalence,
respectively. Moreover, let { i i , j 2i · · · , i n } be a family o f one-argument
functions o f Un.

If for any x, y £ En and i £ { 1 , 2 , . . . , n )

~<x G Dk if and only if x & Dk

x —> y & D k if and only if x € Dk and y £ Dk
x V y € Dk if and only if x £ Dk or y £ D k
x A y £ Dk if and only if x £ Dk and y £ D*
x = y £ Dk if and only if either x ,y £ D k or x, y ^ Dk
ji(x) G Dk if and only if x = i

we will say that the respective connectives o f Μ η>* satisfy the standard
conditions. It is worth noticing that connectives corresponding to the ji
are logical value identifiers and that i i , . . . , j k perform as assertions while
ifc+i,. . . ,in are (many-valued) negations.
Let Q = V, A, = } and Jn = { j i , j 2 , - - - , j n } · The matrix Mn,k,
having a set o f definable functions T C Q U Jn which satisfy the corre­
sponding standard conditions, shall be called T-siandard, and standard if
it is T-standard for T = Q U J„.

8.1.1. E. All finitely valued Lukasiewicz and Post matrices are standard.
In the Post case this does not require any separate justification since those
matrices are built on the basis o f functionally complete algebras (see 3.2).
The isomorphic image o f Lukasiewicz matrix Mn obtained by the mapping
f ( x ) = n — (n — l)x o f the set {0, 1jn — 1, . . . , η — 2/n — 1, 1} onto En
is a matrix o f the form M n>i, in which disjunction and conjunction satisfy
standard conditions; the standard connective o f implication as well as the ji
are definable as shown in 5.2.2.E. The equivalence is obtained while putting
x = y = (x —> y) A (y —> x), and the negation by ->x = x => n.

8.1.2. Each Q-standard matrix K n^ = (E n, , V, A, = , D *) is epi-

—►
morphic to the classical matrix M 2.

Proof . The mapping h : En —> { 0 , 1 } defined by

f 1 when x £ Dk
h(x) = |
[ 0 when x g Dk

establishes the required epimorphism.

8.1.3. E ( K nik) = E ( M2) = TAUT .

62 8. Standard properties o f many-valued constructions

8.2 Axiomatization
The content E (M ntk) o f any { —►} U Jn-standard matrix M ni* is axiom-
atizable by means o f the rules MP, SUB and the following set o f axiom
schemes:

A l. (a —►(β —y a ))
A2. (a -► (/3 -► τ ) ) -► (β ->· (a ->■ 7 ))
A3. ( a —* β ) —* ((β - » 7 ) - * ( « - > 7 ))
A4. (ji(a ) - f - f /?)) — (i,( a ) — β )
A5. (* ,(« ) - 0 ) - ( ( * , _ ! ( « ) — j9) — — ( ( ji ( a ) - 0) - 0 )...))
A 6. # ( « ) —►a for z = l , 2, ...,&
A7. it(r)(«r) “ > (ji(r-l)(< * r-l) " ► ( . . . “ > (ji(l)(<*l) “ >
i / (^*(«1, . . . , a r) ) ) . . . ) ) where / = / ( i ( l ) , . . . , i(r));

Rosser and Turquette (1952). Symbols / and F used in A7 represent, re­

spectively, an arbitrary function of the matrix Μ η>* and a propositional
connective associated with it. To make the schemes A5 and A7 more read­
able a so-called ‘ chain symbol’ G can be employed defined recursively as
follows:

G}= i«i/3 = «1 -► β
= «m - G £ i W ) .

Then schemes equivalent to the indicated axioms are o f the form:

A 5'. G?= 1( * ( « ) - / 3 ) / 3
A7'· GJ= 1ii(, ) ( a , ) i / ([^ (« 1, . . . , Otr)).

Checking the soundness of the axiom system A 1-A 7 (with MP) elicits
no difficulties. Under any valuation of language formulae assigning values
from Dk to the standard implication and to its antecendent, the value of
the successor belongs to Dk· The justification that all formulae whose
schemes are the axioms adopted pertain to E {M n^ ) can be reduced in an
obvious way to the procedures known in the classical logic. Nevertheless,
the completeness proof o f that axiomatization requires much calculation
and, above all, involves quite a complicated induction. Further on, we
shall confine ourselves to the presentation only of the idea o f this proof. To
be perspicuous, the formulae which are theorems o f the deductive system
of Rosser and Turquette are preceded by K
Let W be a formula o f the propositional language considered; assume
that it is built o f τ variables (r > l , r finite) P i, P2, . . . , PT (in the present
consideration propositional variables will be denoted by capital letters (!)).
We assume that Pi,P2i---?Pr is a sequence o f elements o f En assigned
to the variables Pi, . . . , P r by any fixed valuation v in Wn, v(P*) = pi
for i 6 { 1 , 2 , . . . , r } . In turn, w is to denote the r-argument function
 Standard matrix consequence 63

w : E* —> En such that

W = W>(pi,p2, · · · , Ρ τ ) = vW (P u P2, . . . , Pr)·

As the result o f an appropriate substitution in A7' we get

(1) i- G U M P i)U w )
and using other axioms and theorems o f the system together with the
induction with respect to r we can show that

(2) h ( j w( W ) - W ) - (Gri= 1 j Pi(Pi)jv,(W ) - G U M P iW ).

Completeness proof. Assume that W G E (M njk)· Then w G { 1 , 2 , . . . , Λ;}

and by A 6 we obtain

(3) \~jw( W ) ^ W .

Then applying the detachment rule twice to (2) and also to (3) and ( 1)
leads to

(4) ^G\= ij Pi{Pi)W .

From the definition o f G we obtain, for any s G { 1 , 2 , . . . , n},

(5) h j,( P T) - (G 'Z ljPi(Pi)W )

and, consequently, by A5 and the detachment rule

(6) h G\ZljPi{Pi)W .

To copy the reasoning from (5) to (6) in relation to (6) and to successively
achieved formulae terminates at h W , thus concluding the proof.
The Rosser and Turquette result (positively) solves the problem o f ax-
iomatizability o f many known many-valued systems, including n-valued
Post and Lukasiewicz logics (compare 8.1.1.E).

8.3 Standard matrix consequence

Consider any T-standard matrix M and a propositional language £ asso­
ciated with it. The consequence operation C um determined by M is finite
(see 4.3.2) and, moreover,

8.3.1. If —* is in T, then C um = Cn^jej(m ),MP)·

Proof. Standard implication —►is an implication in respect o f C u m .

Thus our lemma is an immediate corollary to 4.3.3.

The consequences produced by T-standard matrices exhibit a great

64 8 . Standard properties o f many-valued constructions

resemblance to the classical logical consequence. For any a,/3 G For and
X C For :

( - 1) CriM(X, “ ·«) = For iff a G C tim(X )

(—0 β G CriM{oL,X) iff a —> β G C njif(A ’)
(V) C n A f ( « V / 3 , X ) = Cnjifiaj-X’) Π C u m 0 , X )
(Λ) Ο π μ ( ο£Λ β , Χ ) = Cnjvf(a, X ) U Cnjif (/?, Jif)
(= ) α = β G C tim(X ) iff Cn/if (α,-Χ") = Ο τΐΜ (β,Χ )
(j) i<(a) V . . . V j n(a) G CnM(0) = F(M )
(J) CnM(js(oi)Jt(a)) = For whenever s^ t

provided that a connective involved (or a group o f connectives in the case

of the j ) belongs to T.
The set o f conditions (“ ·)-(J) together with the axiomatics o f the con­
sequence operation adequately describe the connectives o f standard propo­
sitional logic; thus, for example, (—>) conveys both the deduction theorem
and the detachment rule. This way of expressing the connectives’ prop­
erties was inspired by Tarski (1930), who treated (—>) and some version
o f (—1) as conditions determining implication and negation o f the classical
logic.

8.3.2. The matrix with values { 1 ,2 } isomorphic to M 2 is { —«, — V,

A, =}-standard. Putting i i ( a ) = a and jzipt) = -«a it can be extended to
the standard one.

By virtue o f the foregoing remarks it can easily be seen that an ade­

quate characterization o f the standard propositional logics can, in many
cases, be simplified. So, certain conditions may be replaced by the clas­
sical definitions o f the connectives they concern; all classically permitted
possibilities are admitted, for instance the reduction to -« and —► , or to -i
and V (compare 1.3). Bloom and Brown (1973) studied finite consequence
operations that meet the requirements (-«), (V) and referred to them as
classical.

8.3.3. (Bloom and Brown (1973)). Each classical consequence operation

C is an axiomatic strengthening o f C2, i.e. for C there exists a set Y o f
formulae o f the language on which C is defined such that C (X ) = C 2 (X u y )
for any X C For.

This last result allows a far-reaching exemplification o f the message

given by the originators o f the standard many-valued logics; they state that
‘ It does not follow from this that ordinary two-valued logic is necessary for
the construction o f many-valued logic, but it does follow that it is sufficient
for such constructions’ (cf.Rosser and Turquette (1952 p.l)). According
to 8.3.3, the content o f any T-standard matrix Μ (T sufficient to define
 Standard matrix consequence 65

standard negation and disjunction) is a theory o f the classical logic, i.e.

E (M ) — CriM(0) = C 2 QO for some Y C For. In consequence, the many­
valuedness o f such constructions is entirely expressible in the axiomatics o f
the system: the pattern example o f an axiom is the generalized principle
of the excluded middle; compare (j).
9. Probability and
many-valuedness

Mathematical probability is a measure o f events occurring totally (or par­

tially) beyond human control, that is to say casual events. The probability
calculus in its simplest form resembles many-valued logic. Therefore, the
question o f a connection between probability and many-valuedness emerges
quite naturally.
A pioneering contribution to the probabilistic approach to logical values
was due to MacColl (1897) who reserved the category o f ‘ variability’ for
propositions that are true in some cases and false in others; here proposi­
tional functions with one variable are the case. Lukasiewicz (1913) invented
a theory o f logical probability where undetermined propositions (i.e. proposi­
tional functions) are associated with a fraction which is proportional to the
number o f variable values verifying the proposition and the number o f all
values o f a given finite domain. The differentiating feature o f this compre­
hended probability in comparison to mathematical probability is the fact
that the former refers to propositions and not events. The continuators of
Lukasiewicz’s conception, Reichenbach and Zawirski among them, exerted
much effort to create a many-valued logic within which logical probability
could find a satisfactory interpretation (see e.g. Zawirski (1934a), (1934b),
Reichenbach (1935)). It was not until the 1970s that Giles worked out an
operationalistic conception o f subjective probability interpreted unques­
tionably in denumerable Lukasiewicz logic (see Giles (1974)).

9.1 Logical probability

The foundations o f the Reichenbach-Zawirski conception are provided by
the assumption that there is a function P r ranging over the set o f proposi­
tions o f the considered propositional language, with values from the interval
[0,1]. The basic facts obtaining by virtue o f the postulated characteristics
of P r are embodied in the following rules:

(PI) 0 < Pr(p) < 1

(P2) P r(p V -ip) = 1
 Logical probability 67

(P3) P r(p V q) = Pr(p) + Pr(q) if p and q are mutually exclusive (i.e. if

pAq is an ‘ impossible’ proposition or, equivalently, if P r(pA q) = 0)
(P4) Pr(p) = Pr(q) when p and q are logically equivalent, i.e. if p A~>q
and -Ip A q a r e ‘ impossible’ .
Prom the above, other properties o f P r known in the probability calcu­
lus can be inferred. Among them are:

(1) P r(pA -*p) = 0

(2) P r(-ip ) = 1 — P r(p)
(3) P r(p V q) = P r(p) + P r(q) — Pr(p A q).

Therefore, if we identify the logical value w(p) o f p with the measure o f its
probability Pr(p)

w(p) = P r(p ),

then the function w should satisfy the conditions:

(~i) ty(-ip) = 1 — Pr(p)

(A) w(p A q) = Pr(p) + Pr(q) — P r(p V q)
(V) w(p V q) = Pr(p) + Pr(q) whenp and q are exclusive.

However, let us note that completion o f (V) through a direct reference to

P r is impossible. Whenever propositions p and q are not exclusive, w(p Vq)
is a certain number from [0,1] not less than P r(p ), Pr(q) and not greater
than Pr(p) + Pr(q). This number obviously depends on the ‘ content’ o f
both propositions and not only on their logical values. What follows is a
practical inability to satisfy the extensionality principle in the system whose
semantics resorts to w valuations; values o f some compound propositions
formed by means o f the disjunction connective are not uniquely determined
by the values o f the component propositions. As an illustration, consider
the case when w(p) = 1/ 2. Then we would have both

1/2 V 1/2 = Pr(p V -<p) = 1 (according to (P2))

and
1/2 V 1/2 = Pr(p V p) = 2Pr{p) - Pr(p) = 1/2 (due to (3)).

Consequently, logical probability must not be identified with logical values

in any ordinary extensional many-valued logic.
Emphasis should be placed on the fact that the hitherto presented con­
ception o f the many-valued logic o f (logical) probability does not comprise
the question o f specifying the P r function. Such considerations have been
undertaken in the logic o f induction. One o f the commonly accepted def­
initions identifies the logical probability with the so-called confirmation
function originated by Carnap (1950). Simultaneously, to each possible
probability distribution P r (Pr-specification) on the set o f formulae o f a
68 9. Probability and many-valuedness

given language, there corresponds a full valuation w* which is an extension

o f w. On account o f this we can get an insight into the properties o f the
non-extensional probability logic by considering the system consisting o f all
formulae that are true under any partial valuation w. It suffices to observe
that the conditions

(i) w(a) = 1
(ii) w* (a) = 1 for each w* extending w

are equivalent. Then, enriching the language by the implication and equiv­
alence connectives

p —*q = -ipV q
p = q = (p ^ q )A (q ^ p ),

we obtain a system o f (intensional) propositional logic

PL = {a £ For : w*(a) = 1 for any w* }

investigated by Rescher (1969). Rescher showed that P L = TAUT. The

p roof for the inclusion PL C TAUT is more straightforward; it suffices to
remark that for 0 and 1, w (w* alike) coincides with the classical valuation.
Justification o f the fact that each classical tautology is a PL-tautology
turned out to be more complicated: For this, Rescher considered a handy
( - 1, V)-axiomatics o f CPC by Hilbert and Ackermann and proved that its
axioms pertain to PL and that no application o f its inference rules, MP
and SUB, leads beyond PL.
\

9.2 Operationalistic conception of subjective prob­

ability
The researches of Giles directed towards finding a logic appropriate for
the formalization o f physical theories, including those with undetermined
propositions, resulted in a very convincing philosophical interpretation o f
countable Lukasiewicz logic L^0 combining the conceptions o f logical many­
valuedness and subjective probability (see Giles (1974)). The main point of
Giles’s approach consists in the so-called dispersive physical interpretation
o f standard logic language: it is assumed that each prime proposition in
a physical theory is associated through the rules o f interpretation with a
certain experimental procedure which ends in one o f two possible outcomes,
‘ yes’ or ‘ no’ . The tangible meaning o f a proposition o f the language is then
related to the persons using it (i.e. speakers or observers). In the case
o f prime propositions it is determined from the values o f probability of
success ascribed by users in respective experiments, whereas in the case
o f compound propositions it is determined from the rules o f obligation
incurred by whoever asserts them and formulated in the nomenclature o f
 Operaiionalisiic conception o f subjective probability 69

dialogue logic (see Lorenz (1961)).

The starting assumptions of the construction can be stated in the form
o f the two following principles:

(1) Everyone who asserts a prime proposition A commits themselves to

paying a certain sum o f money, say $1, when the experiment associ­
ated with A results in ςηο\ Secondly, to assert a proposition is not
the same thing as to assert the same proposition twice (in the same
debate).
(2) Every speaker (i.e. language user) is able to ascribe to any prime A
a real number p(A) £ [0,1], the so-called subjective probability, such
that, given any ε < 0, they are willing: (*) to assert A in return for a
payment o f $(1 —p{A) + ε); (*) to assert - Ά in return for a payment
o f $(p(A) + ε).

Both principles together yield that all prime propositions are probability
definite for all the speakers taken into consideration. Observe that here
the risk value o f A for a given person is < A > = 1 — p(A ), denoting the
expected loss if they assert it.
The meaning o f compound propositions is appointed by the rules of
debate o f two participants: a given person and their partner who can be a
fate as well. The rules tend to reduce all assertions o f such propositions to
the assertions o f sequences (sets) o f prime propositions. Giles adopts the
following rules (for any propositions p and 5):

ASSERTION OBLIGATION (COM M ITM ENT)

pVq undertaking to assert either p or q at one’s own choice

pAq undertaking to assert either p or q at the opponent’s choice
p —*q agreement to assert q if the opponent will assert p
-1p agreement to pay $1 to opponent if they will assert p.

The rule concerning negation is less handy than the others but it can
be ‘ standarized’ under the assumption that language contains a constant
F (a falsum) signifying any false proposition; next define -<p as p —»F.
Every dialogue runs as follows: one of the participants asserts a (com ­
pound) proposition, the other joins the game or not. Subsequently, the
uttered propositions that undergo no decomposition in a given step are
repeated, which especially concerns prime propositions. Every position o f
a dialogue game is o f the form

W : Z
W u W2 , . . . , W k : Zu Z 2 , . . . , Z m

W , Z being the names o f the participants, while W \ 7 W 2 , . . . , Wk and Z\, Zi,

70 9. Probability and many-valuedness

. . . , Zm are sequences o f propositions asserted by W and Z , or tenets o f W

and Z , respectively. A position comprising prime propositions exclusively
is final. Once a final position is reached, the game counts as finished,
appropriate experiments are carried out and the accounts are settled. The
following two examples illustrate the course o f a dialogue:

w Z w Z
0 A —* { B —* A ) 0 A -> BM A
A B —> A A b \j A
A ,B A

A :B 4:4

In the first case it is W who takes the risk o f paying $1 when the test proves
falsity o f B . In the second case, two strategies are possible: one safe for
both sides ( A : -A), the other ending with A : B ; according to the rules o f
dialogue the choice lies with Z and it depends on Z ’s evaluation o f the risk
values o f A and B .
The risk value o f a tenet ( A i , A 2 , . . . , A m) is the sum o f risks o f its
elements < Αχ, A*i, . . . , Am > = Σ < Aj > , and that o f final position for Z
is the difference:

< A.i, .4.2, ·. ·, A m : 5i, .02, · · · *Bn ^ Σ< B{ > Σ< Aj >.
Prom a fundamental result o f game theory stating that every game with
perfect information has a ‘ saddle point’ , it follows (see Giles (1974)) that
each valuation o f prime propositions (i.e. assigning to them risk values or,
equivalently, subjective probability values) has a unique extension onto the
whole language guaranteeing both participants no increase in the risk value
of the initial position - or to put it differently: that optimal strategy exists.
The extension in question is defined for any statements P and Q through
the schemes:

< P -> Q > = sup{0, < Q > - < P > }

< P V Q > = inf{< P >, < Q > }
< P A Q > = sup{< P > , < Q > }
< “ »P > = 1 - < P > ;

sup and in f are the symbols o f the operations o f lower upper bound and
greater lower bound in [0,1], respectively.
The formulae o f the considered language to which any valuation assigns
non-positive risk value are referred to as tautologies; there are propositions,
the utterance o f which may lead only to (theoretically) not losing final
positions. Making use o f the equality pr(P) = in f { 1,1 — < P > } one can
 Operationalistic conception o f subjective probability 71

describe the property o f being a tautology in terms of subjective probability.

A simple calculation shows that:

p r ( P - * Q ) = in f { 1, 1 - pr{P ) + pr(Q )}
pr(P V Q) = sup{pr(P), jn-{Q)}
pr(P Λ Q) = in f{p r (P ), pr(Q )}
pr(-*P) = 1 — pr(P).

Thus, pr is a valuation of Lukasiewicz logic L« 0 (compare 5.1). In conse­

quence, the set o f tautologies o f dialogue logic just constructed is equal to
the content o f the matrix M «0: only those formulae are tautologies whose
probability amounts to 1 independently o f the values assigned to prime
propositions as its components.
10. Classical characterization of
many-valued logics

Even the most radical authors of many-valued constructions devoted much

attention to the classical logic. The matrix method inspired by truth-tables
embodies a distinct shadow o f two-valuedness in the division o f the matrix
universe into two subsets of designated and undesignated elements. Be­
sides, it turned out that, by making use of some special, so-called standard,
connectives, it is possible to map this division into the language as well as
to describe consequence relations (operations) o f some many-valued logics
through appropriate sets o f formulae and the consequence o f the classical
logic (see Chapter 8).
In the 1970s the investigations o f logical formalizations bore several
descriptions of many-valued constructions in terms o f zero-one valuations.
Effectively, the interpretations associated with these descriptions shed new
light on the problem o f logical many-valuedness.

10.1 Suszko’s thesis

Suszko (1977) draws attention to the referential character o f homomor-
phisms associating propositions with their (possible) semantic correlates.
Subsequently, he sets them against the logical valuations being zero-one­
valued functions defined on For. Given a propositional language C and
matrix M = (A, D) for £ , the set of valuations TVm is defined as

TVm = {th · h G Hom (C,A)}i

where

1 if h(a) G D
th(a) =
0 if h(a) g D.

Notice that card(TVM) < card(Hom(C, A )) (in general, hi φ h2 does not

imply that thl φ ·ί&2!). Notice, moreover, that

X \=m if ^n(i only if for every t G TVm

t(a) = 1 whenever t ( X) C {1 }.
 Suszko’s thesis 73

The definition o f logical valuations can be simply repeated with respect

to any structural consequence operation C (or, equivalently, for any relation
\=c associated with C ) since for each such C there is a class o f matrices K
having the property that

C = fllC n jt f : M € i f } ;

see 4.2.2'. Thus, each (structural) propositional logic (£ , C) can be deter­

mined by a class o f logical valuations o f the language C or, in other words,
it is logically two-valued (Suszko (1977)).
The justification for Suszko’s thesis is that logical two-valuedness o f an
important family o f logics lacks the description o f valuations (i.e. elements
o f T V c) for an arbitrary relation \=c- Moreover, it seems that giving a
general method for the recursive description o f these valuations without
knowing precisely the structure of the class K of matrices adequate for C
is hardly possible. At the same time, however, even for simple relations
o f inference the conditions defining valuations are illegible. An example
o f a relatively easily definable set of logical valuations is LV3, the class
adequate for (—>, - 1)-version o f the three-valued Lukasiewicz logic (Suszko
(1975)). LV3 is the set o f all functions t : For —►{ 0 , 1 } such that for any
a ,/? ,7 G For the following conditions are satisfied:

(0) *(7 ) = 0 or *(“ ·7 ) = 0

( 1) t(oc — /?) = 1 whenever £(/?) = 1
(2) if J(«) = 1 and £(/?) = 0, then t(a —►/?) = 0
(3) if t(a) = t(P) and t(-«a) = t(—«/?), then t(a —>β ) = 1
(4) if t(a ) = *(/?) = 0 and t(->a) φ *(-■/?), then t(a —►β ) = t(->a)
(5) if ^(“ Όί) = 0, then £(—■—·α) = ^(<*)
(6) if t (α ) = 1 and t(/3) = 0, then t(->(a —>β )) = *(-·/?)
(7) if t(a) = t(->a) = t(/3) and *(-«/?) = 1, then t(-i(a —> β )) = 0.

Usually, the degree o f complexity of the many-valued logic description

increases with the quantity o f values. But in some cases, it can be simpli­
fied by the application o f extra connectives ‘identifying’ original (matrix)
values. Such a use o f the j-operators of Rosser and Turquette (see 8.1)
makes it possible to get a uniform description of logical valuations for fi­
nite Lukasiewicz logics (Malinowski (1977)).
The logical valuation procedure forms a part o f a broader semanti­
cal programme related to the conception o f so-called non-Fregean logics
(Suszko (1972)). According to Suszko there are situations which play
the role o f semantic correlates o f propositions. Logical valuations, for
their part, are nothing more than characteristic functions o f the sets of
formulae that are counterimages o f the sets o f positive situations, i.e. of
those which obtain, under homomorphisms settling the interpretation.
74 10. Classical characterization o f many-valued logics

Following Suszko, it can be said that the n-valued Lukasiewicz logic (n fi­
nite) is a two-valued logic o f n situations $i, s2, . . . , sn denoted by Lukasie­
wicz as 0 , 1/n — 1, . . . , 1, respectively. Obviously then, 0 and 1 must not
be identified with the logical values o f falsity and truth.

10.2 Scott’s method

Scott (1973, 1974) is aware of the deficiency of all the already known in­
terpretations of non-classical logical values. I11 contradistinction to other
scholars, he regards the division o f these values as designated and undes­
ignated as unnatural. And, by replacing more values by more valuations,
Scott endeavours to bestow a more intuitive character upon many-valued
constructions. The valuations are bivalent functions and generate a parti­
tion o f the set o f propositions o f a given language into types corresponding
to the original logical values. Scott considers only finite classes o f valua­
tions and he assumes that (many-valued) logics are determined by single
matrices. The above two papers merely comprise an outline o f a general
method and its exemplification within n-valued Lukasiewicz logics.
Let For be the set o f formulae of a given propositional language C and
V = {vo, v i , .. ·, vn- i } (n > 1) a finite set o f valuations: the elements o f V
are (for the moment) arbitrary functions t;,· : For —►{ t, f}, with t denoting
truth and f falsity. By a type o f propositions of C with respect to V we
mean an arbitrary set Ζβ of the form

Ζβ = {a E For : Vi(a) = Vi(fi) for any i E { 0 , 1 , . . . , η - 1}}.

It can easily be seen that by using an n-element set o f valuations one can
induce maximally 2n types: thus, for example, the two-element set {tuo, ttfi}
of valuations - see the table here - defines four
types, Z \, Z 2 , Z $, Z±. The confining of the valua­
Wq W1
tions diminishes the number o f types. The set of
Zx f f
valuations just considered will define three12 types
z% f t
Z i , Z 2,^4 when we require that wo(ct) < u>i(a)
Z3 t f
for every a E For, and two types Z 2, Z$ when
^4 t t
u>o(a) Φ tui(o:) for every a E For and ^ 1,^4 un­
der the condition that wq = w\.
The types are counterparts o f logical values: Scott (1973) refers to them
as ‘ indexes’ . The above example shows that a given valency < 2n can be
obtained in several ways. Which of these reductions should be taken into
account depends on the properties of propositional connectives that, for
their part, are type-valued operations, i.e. mappings o f sequences o f types
into types. An accurate choice o f the limiting conditions leads to a relatively
simple characterization of the connectives under consideration.

12At most.
 Scott’s method 75

Applying the above method, Scott gets a description o f the implicative

system o f n-valued Lukasiewicz logic through the (n — l)-element set of
valuations

VL*n =

such that for any *, j € {0, 1 , . . . , n — 2} and a € For*, For* being used to
denote the set o f formulae o f the language C* comprising negation -> and
implication —►connectives,

(mon) whenever v ,( «) = t and i < j, then iy (a ) = t

and, moreover, νο(«ι) Φ f and νη- 2(<*2) φ t for some a i, «2 G For*. The
table below shows that the set VL*n determines n types Zo, Z \,. . . , Zn-\
of propositions:

Vo Vl V2 Vn - 3 Vn- 2

So t t t t t
Si f t t t t
s2 f f t t t
. . . . . .

. . . . . .

. . . . . .

S n-2 f f f f t
S„_i f f f f f

The function f( Zi ) = n - i - l / n - l i s a one-to-one order-reversing

mapping o f the set o f types onto the universe o f the Lukasiewicz matrix
Mn : Z0 corresponds to 1 and Zn-\ corresponds to 0 in the matrix; compare
5.1. The negation and implication connectives are characterized in the
following way: Zi —> Zj = Zmax(0 j - i ) , -*Zi = £ n- * - 1. Accordingly, for
any i , j , k e { 0, l , . . . , n - 2},

(“ 0 = t if and only if vn-f c - 2( « ) = f

(—») Vk(a —>β) = t if and only if whenever i + fc < j and v, (a ) = t ,

then V j $ ) = t.

Simple calculation shows that the set o f all formulae o f C* true under
an arbitrary valuation Vi E VL*n is just the content o f M *, i.e. the (“ >,-►)-
reduct o f the Lukasiewicz matrix Mn (n finite!):

E( M*) = { a E For* : Vi(a) = t for i E { 0 , 1 , . . . , n —2}}.

Simultaneously, however, the consequence relation C 2Fot*x For*,

76 10. Classical characterization o f many-valued logics

X |=* a if and only if v ,(a ) = t whenever v,(-Y) C { t}

for any Vi G VL*n,

does not coincide with ^=n (Lukasiewicz’s consequence reduced to the lan­
guage £*). To verify this it suffices to check that {a —>β, a } /?, while
clearly { a —►/ ? , « } |=n β· |=*n is called a conditional assertion. Whether
and how it can be extended onto the whole language C should be evident.
Scott suggests that the equalities of the form = t’ , for i G
{ 0, . . . , n — 2}, should be read as ‘ (the statement) a is true to within
the degree i\ Consequently, he assumes that the numbers in the range
0 < i < n —2 stand for degrees of error in deviation from the truth. Degree
0 is the strongest and corresponds to ‘ perfect’ truth or no error: all the
tautologies of Lukasiewicz logic are schemas of the statements having 0 as
their degree o f error. Besides, Lukasiewicz implication may conveniently
be explained in these terms: assuming i + j < n — 2 we get that

Vi(a —►/?) = t and Vj(«) = t yield Vi+j(ft) = t.

Thus, by making use o f propositions a —>/?, one may express the amount
of shift o f error between the degree o f hypothesis and that o f the con­
clusion as being the measure o f error of the whole implication. An ex­
ample taken from Euclidean geometry justifies the construction. Where
a , &,c . . . are points o f a plane, let the metalinguistic statement ‘ |a — by
denote the distance between a and b. Let S be a propositional language
having the set For o f formulae consisting of all formulae made up from
atomic formulae ‘ α = V (possibly others) by the use of the connective —►
(and possibly others such as ‘ V’ and ‘ A’). Let us define the set o f valuations
Ve = {^Oi Vn-2} (n > 2) for S putting

Vi(‘a = 6’) = t if and only if \a — b\ < i

and assuming that the implication satisfies (—>). Let, finally, ^=e be the
consequence relation determined in S by Ve, i.e.

X ^=e a if and only if ν,·(α) = t whenever Vi(X) C { t}

for any Vi G Ve.

Depending on the choice of a unit distance, the relationship between a

and b is one o f ‘imperfect’ equality and as such it is not transitive. The
conditional assertion

(PI) { ‘ a = & ’ , ‘ & = c ’ } [ = e ‘a = c’

fails in general, while for any a, 6, c it is true that

(P2) ‘a = b'\=e ‘ b = c ^ a = c\
 Urquhart’s interpretation 77

Hence, the use of the Lukasiewicz implication permits us to formulate a

weakened version of the law of transitivity: (P2).

10.3 Urquhart’s interpretation

Urquhart (1973) deals with the relation h between the natural numbers of
the set Sn = { 0 , 1, . . . ,n — 2} and formulae h C Sn x For. He generally
assumes that

(Tr) If x h a and x < y E 5n, then yha

and adopts h to particular logics, thus specifying n, the language, and

providing recursive conditions which establish the meaning o f connectives.
Accordingly, in each of the cases considered we have to deal with some
Kripke-style semantics (compare 12.2):

Kn - ( S n , < , h).

The role o f valuations in K n is taken by mappings F : Var —> 2Sn such

that the relation hp, x hp iff x E F (p ), satisfies (Tr). The relations hjr
are extended onto the set o f all formulae respecting the conditions which
depend on the connectives. Then, a formula a is x-true in Ffn, x h a , if
x hjp a ίοτ any F as above. A formula a is K n-true iff it is true at the point
0, i.e. provided that 0 H a . K n is a semantics o f the system determined
by a given matrix M when that set o f all K n-ti\ie formulae is equal to the
content o f M , i.e. when

E ( M) = { a E For : 0 h a }.

Urquhart provides the semantics for the n- valued logics of Lukasiewicz

and Post, and for the three-valued Bochvar system. For Lukasiewicz calculi
h has to satisfy the conditions:

x h a —>β iff y h a yields x + y h β whenever x + y £ Sn

x 1— ·α iff (n — 2) — x \f a
x h a Vβ iff x h a or x h β
x\- a Α β iff x h a and x h β
x ha Ξ β iff x h a —>β and x h β —> a.

Notice that the ‘ translation’ o f Scott’s valuations in VL* to the instances

of h can be made according to the equivalence:

iho: if and only if v ;( «) = t.

The semantics for the (->, V)-variant o f Post propositional logic is es­
tablished through the conditions:
78 10, Classical characterization o f many-valued logics

x 1— ■« iff y h a fo r n o y E Sn or there is a y E Sn
su ch th a t y < x and y h a
x h a V /3 iff xh a or x h β.

Several meanings may be attached to ‘reference points’ x E Sn. As for

Lukasiewicz and Post logics, Urquhart suggests a temporal interpretation:
if 0 is the present moment, x ^ O a future moment, then ‘ x h a ’ means
‘ a is true at (the moment) x ’ . It is worth noting that the assumption (Tr)
guarantees that any proposition true at x is also true at every moment
y future to x. This obviously means that in the framework elaborated,
propositions are treated as temporally definitive units and, as such, they
must not contain any occasional, time-dependent expressions (such as, for
example, ‘ now’ , ‘ today’ , etc.). It may be appropriate, perhaps, to add that
even when the very originators of many-valued logics use occasional words
in the examples, they usually have in mind temporally-definite marks of
reference.
Under the above interpretation Lukasiewicz implication a —►β is true
at x if and only if the truth o f a at y means that β is true at x + y, i.e. at
the future moment y time units distant from x. In its turn, Lukasiewicz
negation ->a is true at x if and only if a is false at (n — 2) — x, i.e. at the
moment back to n — 2 (the last in Sn) o f x time units. Urquhart suggests
that such a way of understanding exhibits the sources o f difficulties in get­
ting a plainly intuitive interpretation of many-valued Lukasiewicz logics,
and he claims that the ‘ natural’ connectives o f implication and negation
should rather satisfy the conditions:

x h a —►β iff for any y E Sn

(y β whenever x < y and t / h a ) ,
x l— ·α iff y h a for no y E Sn-

Urquhart’s interpretation o f Post logics is, as easily seen, entirely com­

patible with the original interpretation envisaged by Post himself.
11. Quantifiers in many-valued
logic

Many-valued predicate calculi are usually built along the classical pattern
(compare 1.5). Then the enriched language also includes two quantifiers:
universal and existential. Nonetheless, in many cases the difficulties arise
either with getting an adequate semantic description o f quantifiers, or with
an axiomatization o f a defined system o f the first-order logic. On the
other hand, there is a visible tendency to enlarge many-valued predicate
logics with generalized quantifiers binding a number o f nominal variables
and predicates at a time (Rosser and Turquette (1952)) and also with
generalized quantifiers in Mostowski’s sense (see e.g. Rescher (1969)).
In this chapter, an account of ordinary and generalized forms o f quan­
tification in many-valued logic is given. Further, a very relevant topic,
namely the problem of foundation o f set theory on many-valued logic, is
also outlined.

11.1 Ordinary predicate calculi

Introducing two classical quantifiers to many-valued logic gave rise to the
problem o f their new semantical interpretation. In general, the starting
point consists in the so-called substitutional conception o f quantifiers ac­
cording to which V and 3 are treated as (infinite) generalizations o f con­
junction and disjunction, respectively. Accordingly, for any finite domain
U = {ai, a2, · · · i an }·

Vx.F(x) = u F (a i) Λ F (a2) Λ . . . Λ F (an)

3 x F ( x ) = u Flax) V F (a2) V . . . V F (an)

(= u means the equivalence o f the formulae at any interpretation in 17,

αχ,. . . , an being nominal constants ascribed to the objects of the domain).
On reflection this intuition seems justified since in all the acknowledged
systems o f many-valued logic both connectives are commutative and asso­
ciative.
In finite-valued logical calculi constructed upon linear matrices, quanti­
fiers are defined ‘directly’ by means o f algebraic functions determining the
80 11. Quantifiers in many-valued logic

above-mentioned connectives. Thus, for example, for finite Lukasiewicz

and Post logics it is supposed that for any interpretation / in a domain U :

/ ( V x F ( x ) ) = mi n{ f ( F( a) ) : a e U }
f ( 3 x F ( x ) ) = ma x { f ( F( a) ) : a € U } .

For other calculi the definitions o f quantifiers take on the form o f descrip­
tions - as an instance consider the clauses defining quantifiers in Bochvar
logic:

t w h en f ( F( a) ) = t fo r ev e ry a£U
/( V * F ( * ) ) = u w hen f( F( a) ) = u fo r s o m e aE U
f o th erw ise

f when f( F( a) ) = f for every a E U

/( 3 * J ( * ) ) = u when f ( F( a) ) = u for some a E U
t otherwise.

Axiomatic systems o f many-valued predicate logics are usually built as

extensions o f axiom systems o f the basic propositional calculi: to this end
axioms A16 and A17, the rule (GN) and condition (D) o f 1.5 are accepted.
Proofs o f completeness for finitely-valued calculi do not, in general, create
difficulties and can be obtained by adopting classical methods (see e.g.
Surma (1973)). Axiomatizability of several important calculi o f this kind is
assured by Rosser and Turquette’s result presented further in this chapter.
Introducing quantifiers to logics with infinitely many logical values in
the semantical plane can be problematic. It is possible to follow the method
employed in the case o f finite-valued calculi only when the (infinite) set o f
values o f a considered logic is closed under some infinite operations ‘natu­
rally’ related to the conjunction and disjunction connectives. As an illus­
tration o f a calculus where such a procedure is impossible see Lukasiewicz
logic L n0; the set of values of this logic is not closed under operations con­
sidered for finite calculi - in the case when U is infinite it may happen that
the set { f ( F ( a ) ) : a E U} does not contain the least or the greatest element
and ‘ natural’ extensions o f min and max operations (achieved through inf
and sup in [0,1], see below) are out of the question for similar reasons.
A great deal o f trouble is involved in the axiomatization o f the infinite
(-valued) predicate calculi, which is best demonstrated by Ni-valued Luka­
siewicz logic: in this logic quantifiers V and 3 are introduced provided that
for any interpretation in a non-empty domain U :

/ ( V x F ( x ) ) = i nf { f ( F( a) ) : a e U }
f ( 3 x F ( x ) ) = s up{ f ( F( a )) : a e U } ;

see Mostowski (1961). Scarpelini (1962) proved that Ni-valued predicate

calculus thus obtained is not axiomatizable. The investigations o f other
 Set theory and many-valued logic 81

scholars complete, to some extent, Scarpelini’s result characterizing sets of

valid formulae (i.e. true at any interpretation) in terms o f recursive count-
ability RC (see e.g.Kleene (1952)): Rutledge (1959) showed that the set of
valid formulae o f Ni-valued monadic predicate calculus is RC. Hay (1963)
proved that for any valid formula (of calculus considered) a there exists a
finite m > 0 such that ma is derivable from some ‘sound’ axiomatics. Fi­
nally, a very general result o f Mostowski (1961) implies that, among other
things, the set o f formulae valid in the Lukasiewicz matrix with the des­
ignated set (r, 1], 0 < r < 1, is axiomatizable - Mostowski’s proof is not
effective; the author provides no axiomatics. The adduced works allow, as
we think, an estimate o f the complexity and subtlety o f the problem. In
this connection it is also worth while mentioning that the greatest hopes
for proving completeness of Ni-valued predicate calculus lay with the com­
bination o f the algebraic method and MV algebras (see 5.4, Belluce and
Chang (1963)). The experience gained while attempting to constitute such
a proof finally yielded the form o f the interpretation theory o f the first-
order language with values in compact Hausdorff spaces, i.e. the so-called
continuous model theory (see Chang and Keisler (1966)).

11.2 Set theory and many-valued logic

Russell’s paradox13 leads to a more general question, namely whether
in set theory the comprehension axiom ( comprehension principle) can be
adopted:

(AC) The propositions of the form 3 xVy(y G x = Φ(τ/)), where Φ(τ/) is a

formula containing y, are true.

The presence o f (AC) signifies that every formula defines a certain set or,
more concisely, that for any property a set o f objects bearing that property
can be chosen.
Russell’s discovery excludes the acceptance o f (AC) in the set theory
based on the classical logic. Hence, the only method to allow the con­
struction o f set theory and preserving the comprehension principle is to
change its logic. The suggestion of Bochvar may obviously be conceived
as a step in this direction. However, it can hardly be accepted as satis­
factory; though it enables the classification o f some formulae (the formula
defining the Russell set included) as senseless, simultaneously it still results
in a very embarrassing distinction between two categories o f propositions
(compare 7.2). Next, as it turned out, the three-valued and, more generally,
all finite-valued Lukasiewicz logics cannot be seriously taken into consid­
eration either. Moh Shaw-Kwei (1954) provided the following method of
construction o f ‘ undesirable’ sets in Ln. For a given finite n > 2 we put

13See 2.2.
82 11. Quantifiers in many-valued logic

Zn = { x : x G X -* n p } 14.

The set Zn is antinomial since the following absorption rule

« -N.
(abs„)
« >n—1 β

is a rule o f Ln (equivalently (a — β) —> (a —»„_χ β ) € E( Mn)). The

definition of Zn implies that Zn £ Zn and Zn £ Zn —*n p are equivalent.
Therefore, using p -> p £ E( Mn) we get (1) Zn £ Zn -+ ( Zn £ Zn —►„ p)
and (2) ( Zn £ Zn — p) -> Zn £ Zn. Since applying (abs„) to (1) yields
(3) Zn £ Zn —*n p we also have (4) Zn £ Zn. Thus, after ( n —l)-ary
application of MP to (3) and (4) we finally get p .
As absorption rules (absn) are not the rules o f infinite-valued Lukasie­
wicz logics, much attention was given to the possibility o f founding set
theory with (AC) on these logics. Skolem (1957) put forward a hypothesis
that the proposition (AC) was consistent15 in Ηχ-valued Lukasiewicz logic
(or, more accurately, in the predicate calculus with the predicate € ). Up
till now Skolem’s hypothesis was only partly supported.
Using advanced proof theory techniques and applying Brouwer’s fixed
point theorem (for an n-dimensional cube) Skolem demonstrated that the
set o f formulae o f the form

(si) V®i. . . V®„3yVt(t £ y = U(t, y, * χ ,. . . , ®„)),

with U(t , 2/, ®χ,. . . , x n) being a formula containing no quantifiers wherein

free variables are at the outmost t , y and ®i, is consistent is Ηχ-
valued Lukasiewicz logic16. The result o f Skolem, after he had applied
his method o f proof, was expanded by Chang and Fenstad. Chang (1963)
showed that the assumption o f the absence of quantifiers in formulae U(t,y,
®x,. . . , x n) can be removed under the condition that bound variables in U
appear in atomic formulae u £ w only on the second position. Chang also
proved that in any formula

(c) 3 xVy(y £ x = Φ($/)),

14p denotes an arbitrary proposition (or formula) and thus also any formula incon­
sistent in L n ; compare p. 19. a —>n β is an ‘ascending’ implication a —►(a —►( . . .
—►(oc —►/3 ). . . ) ) with η — 1 antecedents a.
15 ok is consistent in the predicate calculus iff there exists an interpretation / (fj)) such
that f ( a ) = 1. In Lukasiewicz logic the concept o f interpretation is defined according to
the pattern o f 1.5 applied in ; see 5.1 and 11.1.
16The set X o f closed (i.e. having no free variables) formulae is consistent iff there
exists an interpretation / sending all formulae o f X into true propositions, i.e. such that
f ( X ) C { l } ; compare Note 15.
 Generalized quantifiers 83

(compare (AC)) where Φ(y) is a formula with one free variable y, is consis­
tent. Fenstad (1964) obtained a similar result in this direction: he showed
that the set o f Skolem’s formulae is consistent (in Νχ-valued logic) under
the assumption that the free variable t takes only the place o f w in atomic
formulae u € w.
All this shows that the question o f unlimited consistency o f the compre­
hension axiom in many-valued logics still remains open. And it obviously
leaves room for making several suppositions. For example, it can seem un­
natural that Skolem combined the problem directly with and not with
the logic o f countably many values. It should immediately be noticed that
the attempts to get a proof o f consistency of (AC) in L ^ 0 would have to be
connected with working out a new method: Brouwer’s fixed point theorem
for the set o f rational numbers in the interval [0,1] does not hold.

11.3 Generalized quantifiers

Traditional (ordinary) quantifiers can be treated as functions on the set o f
pairs (a^F1), where x is a nominal variable and F a formula, with values
in the set o f formulae. Starting from this intuition Rosser and Turquette
(1952) developed the principles of a very general theory o f quantification
for finitely many-valued logics. Operators considered in this theory are to
bind many variables and formulae simultaneously.
A generalized quantifier of Rosser and Turquette is any formula o f the
form:

( Q i) Qi(x lj -^1? -^2? · · · j F*. ) j

where . . . , x mi are nominal variables and F\, F2, . . . , Ft . formulae

built from predicates, nominal and propositional variables, and proposi­
tional connectives. In what follows we shall accept that the traditional
quantifiers are expressed by means of Q 1 and Q 2 as follows:

MxF{x) = Qx(z, F), 3 xF (x ) = Q 2 (x , F).

The intended meaning o f Q, functions in n-valued logic is determined

with the help of interpretations assigning to formulae values from the set
{ 1 , 2 , . . . , n } . At the same time, in order to get a full characterization of
any o f those functions it suffices to provide an interpretation o f the Qi in
which F i , . . . , Fti are atomic formulae or, speaking more precisely, (simple)
predicates. Operating on the links between formulae stated by means of
propositional connectives o f basic logic enables the construction o f non-
classical quantifiers. To illustrate the method we will provide an example
o f n-valued (n > 2) quantifiers Q$ = Q z(x,y, F i, F2) defined with the two
Lukasiewicz connectives, implication and disjunction, introduced here (see
8.1) as p —►q = m a x{l,q —p + 1} and p A q = m ax{p,q}.
84 11. Quantifiers in many-valued logic

Let P and S be (m + 2)-argument predicates. Suppose that P =

P (x ,y , ^i , · · · , Zm) and S = S ( z ,2/ , 2i , . . . , 2m). In order to simplify the
descriptions o f these and other generalized quantifiers we shall adopt the
following inward abbreviation: $ 3(2 , 2/, P, S) will also be used to denote
a (propositional) schema obtained from $3 through a ‘ partial’ interpre­
tation consisting o f determining the meaning o f predicates P and S in a
non-empty domain D. Assume that

( 1) $ 3(2 , 2/, P, S') takes the truth-value 1 for a given sequence o f objects
αχ,. . . , am £ D for zly. . . , zn if and only if there is an a £ D such
that P (a, 2/, αχ,. . . , am) —►£(a, 2/, a1?. . . , am) for all y £ D takes the
truth-value 2.
(2) $ 3(2 , 2/, P, S) takes the truth-value 2 for a given sequence o f ob­
jects £ D for if and only if for every x ,y £
D P ( 2 , y,a 1, . . . , am) A £ ( 2 , y,a χ,. . . , am) takes the truth-value 1.
(3) $ 3(2 , 2/, P, £ ) takes the truth-value 3 for a given sequence of objects
αχ,. . . , am £ D for . . . , zm if and only if for every x £ D there is
a b £ D such that P ( 2 , 6, α χ , , am) —> S (x , b, αχ,. . . , am) takes the
truth-value 3.
(4) $ 3(2;, 2/, P, 5 ) takes the truth-value 4 for a given sequence o f objects
αχ, · · ·, ctm € D for ζχ , . . . , zm in all cases except those stated in pre­
scriptions ( 1), (2) and (3).
(5) Qz{x, 2/, P, S') never takes the truth-values 5 , . . . , n.

We accept that the conditions ( l) - ( 5 ) determine the quantifier denoted as

Q 3 . Moreover, we shall likewise define two traditional quantifiers Qi and
Q 2 supposing that

( Q l ) Q i(x ,P ) takes the truth-value r (1 < r < n) for a given sequence of

objects αχ,. . . , am £ D for ζχ,. . . , zm if and only if there is an a £ D
such that Ρ (α, αχ,. . . , am) takes the truth-value r and, moreover, for
any x £ D the value of P (x , αχ,. . . , am) is not greater than r.
(Q2) Q 2 ( x , P) takes the truth-value r (1 < r < n) for a given sequence of
objects αχ,. . . , am £ D for ζχ,. . . , zm if and only if there is an a £ D
such that Ρ (α, αχ,. . . , am) takes the truth-value r and, moreover, for
any x £ D the value of P(®, a1?. . . , am) is not less than r.

The theory of representation of generalized quantifiers in Rosser and

Turquette logics (see 8.1) allows any generalized quantifier to be expressed
by means o f ‘ ordinary’ quantifiers. The means to achieve it consists of
so-called partial normal forms. A partial normal form o f a formula o f the
standard propositional calculus is a conjunctive-disjunctive formula giving
a ‘ two-valued’ specification of conditions under which the former takes on
a definite truth-value. To build this form we need any standard disjunction
and conjunction connectives and the j . The r-th partial normal form Nr(a)
 Generalized quantifiers 85

o f a formula a = α ( ρ χ ,. . . ,ps) is the disjunction o f all conjunctions o f the

form

* M p i ) A J2 r(P2 ) A . . . Λ JSr(p,) ,

where l r , 2r, . . . , sr is a sequence o f logical values such that v(a) = r for

any valuation v meeting the requirement v(pi) = ir. The above method is
extended to the predicate calculus with generalized quantifiers permitting
the use o f ordinary quantifiers (or, equivalently, the universal quantifier and
standard negation; compare 1.5). The procedure of forming the particular
partial normal forms is, speaking most generally, analogical. Quantifiers
either usually stand at the beginning or precede conjunctions; it is also
possible to obtain two part prenex normal forms: a ‘string’ o f quantifiers
plus a normal form formula with no bound variables. This cursory out­
line o f the manner o f representation will be illustrated by a (non-prenex)
representation o f Q3 in a four-valued logic (n = 4):

J M Q 3 ( z ,J/,-P,S)) = - V z - V y { ( J 2( 5 ) Λ Ji (P)) V ( J3(S)

A J2 ( P ) ) V (J4 (S) A J3 ( P) ) }
N 2 (Q 3 ( x , y , P, S) ) = VzVy { ^ ( P ) Λ M S ) }
N 3 {Q 3 ( x , y , P, S) ) = V z - V y - { ( J3(S ) Λ J i ( P ) ) V (J 4 (S)
AM P ))}
N4 (Q 3 ( x , y , P, S) ) = - { N 2 (Q 3 (z , y, P, 5 ) ) V N 2 (Q 3 (z , y, P, S))
VJV3 (Q3 ( z , y , P , S ) ) } .

The predicate logic formulated in a language containing the universal

quantifier and a finite number o f generalized quantifiers Q Qc built i , . . . ,

over n-valued propositional calculus with k designated values is standard if

its connectives satisfy the standard conditions (see 8.1). Furthermore, for
any formula of the form VxF

Vz(Fi(z) V . . . V Fk(z)) = (Nx V . . . V Nk),

where F\(z ) , . . . , Fk(z) and . . . , Nk are respectively the partial normal

forms o f F and MxF. (We are willing to explain to the reader surprised by
the employment o f a new variable z on the left side o f the equivalence that
this is intended and necessary in view o f the complexity o f the problem
of normal form representation here; compare Rosser and Turquette (1952
pp.57 and 61).)
Rosser and Turquette presented a method o f axiomatization o f the stan­
dard n-valued predicate calculi (with generalized quantifiers) enriching A l -
A7 o f 8.2 with the following schemes:

A 8. VxF(x) —►F(y) provided that the substitution y/x is

permissible (i.e. y will not be bound
in any place where x was free in F( x) )
86 11. Quantifiers in many-valued logic

A9. Vx(Fi —* F2 ) —> (Fi —> Vx F 2) under the assumption

that x is not free in F\
A10. Nr (Qi (®l, . · · , ®τη,·ί -Fli · · · j -Ffi)) * «Λ* {Qt (®lj · · · j ®m,·)
i i , . . . , Ft .)) where 1 < 7* < n, 1 < i < c

and adding the generalization rule

(GGN) F / VzF.

A significant role in the proof of adequacy o f these axiomatics is played by

the technique o f (partial) normal forms. The proof o f completeness was
achieved after having adopted the standard Henkin method.
The result o f Rosser and Turquette answers the problem o f axiomatiza-
bility o f many known n-valued predicate calculi, including Lukasiewicz and
Post logics (see Chapter 8). The question o f whether and to what extent
generalized quantifiers can be useful in practice still remains open.
12. Intuitionism and the modal
logics of Lewis

Intuitionistic logic and the modal logics o f Lewis are the most important
and best motivated non-classical logical constructions rendered axiomati-
cally. Their creators were not guided by the idea o f introducing supple­
mentary logical values. However, it occurred that these logics can be char­
acterized exclusively by means of infinite-valued matrices. On the other
hand, non-matrix semantics formed to describe intuitionistic and modal
logics became an important tool for interpreting and describing o f all the
non-classical logical constructions.

12.1 Intuitionistic logic

Intuitionism constitutes a constructivistic trend in the studies o f the foun­
dations o f mathematics. Its sources are found in some elements o f the
philosophy o f Kant who perceived the base o f mathematics in an a priori
intuition o f time and space and who emphasized the role o f construction
in justifying the existence of mathematical objects. The history o f the in­
tuitionistic conception is excessively rich and has links with such eminent
mathematicians as L. Kronecker, H. Poincare, E. Borel, H. Lebesgue, to
mention but a few o f them. The systematic and mature development of
the intuitionistic ideas initiated in 1907 was the life-work o f L.E.J. Brouwer
(see Heyting (1966)).
One o f the main assumptions of intuitionism is the postulate o f effec­
tiveness of existential mathematical theorems: a proposition concerning
the existence o f mathematical objects can be accepted only when we are
able to provide a method of construction o f those objects. Proposed by
Heyting (1930), the interpretation o f logical constants and quantifiers al­
lowed the formulation of an axiomatization o f intuitionistic logic generally
acknowledged as adequate. According to this interpretation the validity of
any proposition is identified with its provability, and proofs o f compound
propositions are composed of the proofs of simpler ones: the proof o f _____is
a construction_____

_____a Λ β ____ consisting o f a proof o f a and a proof o f β;

88 12. Intuitionism and the modal logics o f Lewis

ot \f β choosing one of the propositions α ,β and

laying down a proof o f it;
transferring any proof o f a onto the proof
o f β and verifying that the results o f it are
proofs o f β indeed;
—*<* equivalent to a proof of ol —►0 , where 0 is
an absurd sentence (falsum);
ΞχΦ(χ) choosing an object a and laying down
a proof o f Φ(α);
\/χΦ(χ) which to every object a of a given domain
assigns a proof of Φ(α) and subsequently
verifies it.

Heyting (1930) presents the intuitionistic propositional calculus as a

system INT based on the axioms:

(HI) p - > (pAp)

(H2) ( p A q ) - * ( qAp)
(H3) (P -►«) -► ((pA r) - » (q A r))
(H4) ((p —* q) A(q —* r)) -► (p -*· r)
(H5) q-*{p-*q)
(H6) (p A (p —►?)) — q
(H7) p-*(pVq)
(H8) (p V q) -* (q V p )
(H9) ((p -> r) A (« -*· r)) ((p V q) -► r)
(H10) ->p -* (p -* q)
(HU) ((p - * q ) A (p -* ->«)) -* -φ

and the rules MP and SUB. The ‘soundness’ of INT axiomatics can be
demonstrated through the above-mentioned interpretation o f logical con­
stants in terms o f proof.
It is readily checked that all the laws of the intuitionistic propositional
calculus are classical tautologies. Notwithstanding this, INT differs from
CPC, which can be shown by applying the following three-element Heyting
matrix

H3 = ( { 0 , 1 / 2 , 1 } , ^ V, Λ, {1}),

in which V and Λ are defined as in the Lukasiewicz matrix while -> and —►
are characterized by the tables:

a —>a 0 1 /2 1
0 1 0 1 1 1
1/2 0 1 /2 0 1 1
1 0 1 0 1 /2 1
 Intuitionistic logic 89

INT C E( H3) since axioms ( H l ) - ( H l l ) pertain to ϋ^-Ηβ) and E(Hs) is

closed under the rules MP and SUB. Subsequently, such laws o f the classical
logic as

P V ~>p, —>—>p -► p, (-.p - * p ) - * p

are not tautologies o f the matrix Hy. it suffices to consider any valuation
h such that h(p) = 1/2. It is also noteworthy that the strengthening of
the Heyting axiom system with the law o f the excluded middle leads to
the classical logic. The problems of relations between the intuitionistic and
classical propositional calculi were given much attention, and the so-called
intermediate logics (intermediate between the intuitionistic and classical
logics) were intensely studied (see e.g. the papers by Kabzinski and Kra-
jewski in Marciszewski (1987)) - the calculus determined by the Heyting
matrix can serve as an example o f such a logic. The weakening o f INT
resulting from the omission of (H10) leads to the system o f minimal logic
(consult Johansson (1936)), which is closer to intuitionism than Heyting
calculus.
Godel (1932) showed that INT cannot be described by a finite matrix
and, further, by a finite set o f finite matrices. Godel’s reasoning consists
of the construction of a sequence of matrices approximating INT, and then
in pointing out, on the base o f the matrices of that sequence, a suitable set
of formulae outside INT: the n-valued Godel matrix (n > 2, n finite) is of
the form

G „ = ({0, 1/n — 1 , . . . , n — 2 / n — 1, 1}, V, Λ, {1} ),

where

f 1 if x = 0 f 1 ifx < y x V y = max(x, y ),

= j x^ y= j
[ 0 if x φ 0 [ y iix>y x Λ y = m z n (® , y).

Notice that G 2 = Af2, G 3 = H 3 and that for every n € N:

INT 5 E( Gn) 5 E( Gn- i ) 5 . . . 5 E( G3) £ E( G2) = TAUT.

Let, further, = be the connective of equivalence defined customarily; for

any formulae a,/?, a = β —df (a —* β ) Λ (β —* a ). Then, a = a € INT
and the functions of matrices Gn corresponding to = are described by

1 when x — y

{ m in(x , y) when x φ y.

Consider the sequence {d,·} (z € N ) o f formulae:

(d*·) (pi = p2) V (pi = p3) V . . . V (pi = pi) V (p2 = P 3 ) V (p2 = p4) V . . .
90 12. Intuitionism and the modal logics o f Lewis

V(P2 = Pi) V . . . V (pi—2 = P i-l) v (p,-_2 = Pi) V (p ,_ i = p ,).

As everyone is quick to notice, d, ^ E (G n) if i < n and thus d2, d3, ..

d „ , . . . ^ INT. The proof of the non-existence of a finite matrix (weakly)
adequate to INT can be viewed as follows. If INT = E (M ) for some ra-
element matrix M = (.4 ,1 ), then v(a = a) £ I for any valuation v. Then,
by extensionality (see 3.1) for arbitrary formulae a,/3

(ο) ν (α ) = ν(β ) implies υ(α = β) £ Ι·

Now, consider a formula dfc where k > m. Since the number o f propositional
variables o f this formula exceeds the cardinality o f the matrix A f, then for
every v there are Pk1 ,Pk2 € \hr(dk) such that v(pk1) = v(pk2) and, in view
of (o), v(pk 1 = Ph2) € I. Applying the fact that (H7) £ E (M ), v(dk) € I
can be proved. Thus d* £ E (M ). A contradiction.
Keeping the original definitions of Godel’s connectives one may define an
infinite-valued logic. Thus, taking the set o f all rational numbers from the
real interval [0,1], we get the denumerable Godel matrix G n0. It follows that
INT ^ E (G * o). Dummet (1959) showed that the system o f propositional
logic thus obtained is axiomatizable: it may be obtained from INT by
adjoining the formula

(H12) (p —►?) V (g —►p)

to the axiom system (H l) - ( H ll) ; the rules of inference remain unchanged.

An infinite class o f finite matrices, adequate for INT, was introduced
by Jaskowski (1936). Jaskowski’s sequence of matrices begins with the
classical matrix and its succeeding terms are made out o f the preceding
ones by means o f a special operation G; an account and detailed proof of
the completeness theorem can be found in Surma (1971). Other researches
produced interesting and perspicious constructions o f relational semantics
for intuitionistic logic, such as the forcing by Cohen (see Fitting (1969)) and
the Kripke-Grzegorczyk semantics (Kripke (1965b), Grzegorczyk (1964)).
O f interest to us are the investigations by Beth (1956) concentrating on
the question of what is ‘intuitionistically true tautology’ and on a general
method o f finding ‘ intuitionistic proofs’ of the property. In connection
with this we should also mention the work by Lorenzen (1961) in which
a dialogue system for intuitionistic propositional calculus is constructed.
What appears to call for special attention is a topological interpretation
provided by Tarski (1938). Here, propositional variables are associated
with open sets val(p) o f a fixed topological space (X^I) (I is an interior
operation) and then this mapping extends onto all formulae thus setting:

val(a V β ) = val(a) U ναΐ(β)

val(a Λ β ) = val(a) Π ναΐ(β)
 Modal logics S4 and S5 91

val(a —>β) = I ((X —val(a)) U ναΐ(β))

val(-*a) = I ( X — val(a)).

If a € INT, then val(a) = X . Conversely, if val(et) = X for any space X

and for any valuation val, then a £ INT. Tarski’s interpretation inaugu­
rated the characterization of the intuitionistic logic in lattice-theory terms,
i.e. pseudo-Boolean, also called Hey ting (or Brouwer) algebras; see Rasiowa
and Sikorski (1963). Curiously enough, in practice almost all non-classical
logical constructions, many-valued included, are expressible by means o f
pseudo-Boolean algebras (compare 6.4).

12.2 M odal logics S4 and S5

Modal alethic functors (referring to truth) ‘it is possible that’ and ‘ it is
necessary that’ are the most intensely investigated non-truth-functional
propositional connectives. Their introduction to a group o f logical systems
(so-called modal logics) as central notions resulted from the criticism of
some classical laws referring to the properties of the implication connective
such as:

p -* (« ->■ p)
- ,p _> (p _> q )

(p q) V (q -* p).

Lewis (1918) formulated a programme of eliminating ‘ paradoxical’ prop­

erties o f implication through the formalization of strict implication defined
with the help o f possibility (O) and the classical negation and conjunc­
tion:

ct => β =df - Ό ( α Λ —>/?).

Taking O, -i and —►Lewis defines, apart from strict implication, further

propositional connectives, among them necessity (□ )

□ a = -i O -ι a,

and he produces axiomatically five modal systems SI, S2, S3, S4 and S5;
see Lewis and Langford (1932). Lewis’s axioms are formulae containing
connectives -■,=>, Λ and O, whereas rules consist o f the detachment rule
for strong implication, the adjunction rule (i.e. the rule o f introduction of
conjunction) and the rule of extensionality for ‘strict equivalence’ . Systems
S1-S5 determine different variants of strong implication and, consequently,
diverse classes o f alethic modalities definable through iterations o f O, □ and
the classical connectives; for example, infinitely many in SI and only six
in S5. This problem, however, is not our concern. Instead, we shall note
the fact that each o f these systems comprises CPC as one o f its subsystems
92 12. Intuitionism and the modal logics o f Lewis

and, moreover, that strict implication => can equivalently be defined with
the necessity connective:

<*=> β =df D(<* —►β )·

Godel (1933) put forward a formalization o f S4 and S5 which (see Lem­

mon (1957)) has become a pattern for a modern standard o f syntactic
characterization o f modal logics. The primitives o f Godel’s formalization
include, besides the classical connectives of negation and implication, the
connective o f necessity □. Further to that, O is defined as i.e. for
every a , O a = The role of axioms is taken on by the formulae

(Gl) p ->· (q -*■ p)

(G 2) ( p —> ( q —* r ) ) ((p — > q) —> (p — > r ) )
(G 3) (_,p _> -,5) (3 p)

(G 4) D p —► p

(G 5) □ (p - t g ) - * (Dp _► □ q )

(G 6) D p —¥ DDp (for S4 only)

(G7) O p —> D O p (for S5 only),

and the rules are as follows: MP, SUB and GodeVs rule

(R G ) h a/ h Da.

But it should be remarked that the above description o f S4 and S5 logics

was only incidentally obtained while investigating intuitionism; the sys­
tems worked out wer^ to depict the concept o f proof. Several years later
McKinsey and Tarski (1948) gave a modal interpretation of the intuitionis-
tic calculus determining precisely its relation to S4. The interpretation of
McKinsey and Tarski was fixed by the following ‘ translation’ o f formulae
of intuitionistic logic onto the modal formulae:

77l(a) = Da :: for propositional variables

ra (-i«) = □ -im (a )
m (a V β) = m(a) V τη(β)
m (a Λ β ) = m(a) Λ τη(β)
m (a —>β ) = □ (m (a) —>τη(β))

(the left-hand side connectives are intuitionistic, the right classical). A

formula a 6 INT if and only if m(a) 6 S4 (McKinsey and Tarski (1948)).
Dugundji (1940), making use o f Godel’s method, demonstrated that
none of the systems S1-S5 of Lewis can be described by a finite ma­
trix. But the ‘ universal’ reasoning of Dugundji proves more, namely that
for no reasonable non-trivial (system of) modal logic there exists an ade­
quate finite matrix such that the system coincides with the content o f the
 Modal logics S4 and S5 93

matrix. The first attempt to obtain a matrix characterization of S5 is due to

Wajsberg (1933) who invented a theorem-decision method for S5 referring
to the first-order logic; he showed that isomorphism holds between S5 and
the theorems o f one-argument predicate calculus. That isomorphism is es­
tablished by changing propositional variables into (unary) predicates, and
modal functors □ and O accordingly into universal and existential quanti­
fiers. An elegant matrix description of S5 can be procured with the help of
an infinite Boolean algebra (implication being defined as x —►y = —x U y)
with additional operations 7, C defined as follows17:

1 if x = 1 1 if x φ 0
I ( x) = C(x)
0 if x φ 1 0 if x = 0;

the idea o f such an algebra comes from Henle (compare Lewis and Langford
(1932) 1959 edn).
Inspired by Tarski (1938) a broad stream o f studies developed, devoted
to the algebraic semantics for modal logics (see Lemmon (1966)). The
principle o f algebraic interpretation o f the logics mentioned consists o f as­
signing to formulae the elements of a power set o f a given topological space
in such a way that the connectives o f disjunction, conjunction and nega­
tion are associated with the set-theoretical operations o f union, intersection
and complementation, and modal functors with the topological operations
of closure C (for O) and interior 7 (for □ ). Abstract modal algebras are
Boolean algebras with topological operators: classes o f such algebras cor­
responding to (or determining) some definite systems o f modal logic are
defined by imposing extra conditions on C or 7. The completeness theo­
rem for modal systems in respect to the algebraic semantics is o f the form:
a is a theorem in a system S if and only if a 6 E (A, { 1}) for any S-algebra
A (1 is used to denote the ‘ designated’ element o f A , maximal one). It is
evident that the algebraic semantics is a special case o f the matrix one.
The structure o f the form

A = ( A , U, Π, - , 1, 0, C, I)

is referred to as an S4-algebra provided that

(1) (A, U, Π, —) is a Boolean algebra

(2) C is a one-argument operation such that
C (x U y) = C(x) U C( y ),
x < C(x),
C( 0) = 0,
C( C( x) ) = C(x),

171 corresponds to necessity (□ ) and C to possibility (O). Obviously, I ( x ) = —C ( —x)

and C ( x ) = —/ ( —*).
94 12. Iniuitionism and the modal logics o f Lewis

(3) I ( x) = - C ( - x )

and an S5-algebra if additionally18

(4) x < IC(x).

A qualitatively different way of modelling modal logics was embodied in

Kripke (1963). The motivational layer has quite a long genealogy reaching
back to the philosophy o f Leibniz. The conception itself was anticipated
by Kanger (1957) and Hintikka (1963). The Kripke frame is a triple

K = (G, K , R)

where K is a universe, R a binary relation on K (R C /C2), and G € /C.

Kripke interprets K as a set o f possible worlds, G as a real world and R
as a relation between worlds: K 1 R K 2 means ‘ Hi 2 is possible relative to
K\\ It is supposed that R is at least reflexive and, thus, that each world is
possible relative to itself. Model structures for particular systems o f modal
logic are obtained through further specification o f the relation R. K is
an S4-frame if R is transitive. The class of S5-frames is singled out by
equivalence relations: R is assumed also to be symmetric and transitive.
The characterization of modal notions in Kripke frames is given by means
of the relation R according to the rule: a proposition a is necessary in (a
world) K if and only if a is true in all worlds possible relative to K and a
is possible in K if there exists a world possible relative to K wherein a is
true.
A function V : For x K —> { 0 ,1 } is a valuation o f the language o f
propositional modal logic in a given frame K whenever:

V(-<a, Κ) = 1 if and only if ^ (« ,* 0 = 0

ν(α^β,Κ) = 1 if and only if V(a,K) = 0 or ν ( β, Κ) = 1
V ( « V β, Κ) = 1 if and only if V(a,JT) = l or ν ( β, Κ) = 1
F ( « Λ β, Κ) = 1 if and only if V(a,K) = 1 and ν ( β, Κ) = 1
ν(αα ,ίΓ ) = 1 if and only if V(a,K') = 1 for every K f 6 K
such that K R K f
V (O a , Κ ) = 1 if and only if Π « .* ') = ι for some Kf6 K
such that K R K f.

Formula a is true in Kripke frame K if V (ct, G) = 1 for every valuation V

in K . The completeness theorem for Kripke semantics is o f the form: a is
a theorem o f a given modal system S if and only if a is true in any S-frame.
Kripke semantics in the version presented applies only to those modal
systems in which Godel’s rule, (RG), is accepted. Its range o f applica­

18 < is the lattice order: x < y if and only if x Π y = x (or, equivalently, if x U y = y) .

 Remarks on quantification 95

tion can be enlarged when we include non-normal worlds where every­

thing is possible in /C; see Kripke (1965a). There exists a case o f form­
ing Kripke’s semantics for many-valued non-modal logics (compare 10.3,
Kripke (1965b)). Grzegorczyk (1964) characterized the intuitionistic logic
in terms o f the process o f knowledge acquisition by means of frames based
on the set o f partially ordered time moments (R has to be reflexive, anti­
symmetric and transitive).

12.3 Remarks on quantification

In the intuitionistic and modal logics quantifiers have a special role. In the
former it is only on the level o f the predicate calculus that an exhaustive
justification can be provided for the intuitionistic repudiation o f several
apparently ‘ propositional’ laws of the classical logic. The extensions of
modal logics to first-order calculi provide a better insight into the gist of
possibility and necessity connectives and cause the essential revision o f the
set-theoretical semantics being the indirect reason for the construction o f
a large class o f philosophically well-motivated systems o f the so-called free
logic.
The chief assumption of intuitionism, the effectiveness postulate o f the
existential proofs for mathematical theorems, implies immediately the need
to eliminate some quantificational laws from the logic. Let us consider two
o f them:

(11) -.V x-.P (x) — 3 x P (x )

(12) -i3x->P(x) —>V xP (x).

Both formulae as well as their converse are obviously laws o f the classi­
cal logic (compare 1.5). Intuitionistic scholars baldly reject (11) and (12)
because, when interpreted in terms of proof (see 12.1), they afford the
possibility o f non-constructive inferences o f both existential and universal
sentences. (11) provides a method o f proving the existence o f (mathemat­
ical) objects having the property P by virtue o f the fact that it is not the
case that all objects (of a given universe) possess the property ->P, which
can be proved while showing that accepting Vx-»P(x) leads to contradic­
tion. (12) can be criticized similarly. Traditional counterexamples that
open the door to the withdrawal o f some CPC laws from the intuitionistic
logic are usually formulated in reference to known unsolved mathematical
problems. Brouwer employs in his argumentation, for instance, the prob­
lem o f regularity o f the decimal extension o f π. If P( x) is an arithmetical
predicate interpreted as a description of some regularity of a decimal ex­
tension o f 7Γ, for example ‘ the position x o f the extension equals 9 and is
preceded by eight 9s’ , then neither the proposition 3 x P (x ) nor ->3xP(x)
can be justified. Thus the formula
96 12. Intuitionism and the modal logics o f Lewis

(*) 3 x P (x ) V ->3xP{x)

is not intuitionistically valid. And this yields that the law o f the excluded
middle should be repudiated. It can be further proved that, for example,
the law o f double negation and notably the implication -«-ip —> p are not
laws o f intuitionistic logic either: to this end one has to verify, using the
proof interpretation (in 12.1), that -ι-·(3χΡ (χ) V -·3 χ Ρ (χ )) is intuitionisti­
cally true and then to consider the formula

(**) -i-i(3 x P (x ) v ->3xP(x)) —> 3 x P (x ) V -»3xP (x)

which is a substitution instance of the implication indicated earlier.

A syntactic formalization of the intuitionistic predicate calculus is ob­
tained through the extension o f the axiom system ( H l) - ( H ll) with the
addition o f the usual axioms (or rules) for quantifiers; compare 1.5. The
above-mentioned differences in respect o f the classical logic appear to be
consequences o f the intuitionistic characterization o f propositional connec­
tives (mainly, negation and implication). At the same time, the problem
of completeness is not easy (see e.g. van Dalen (1986 p.273)). A vast study
on the intuitionistic predicate logic is contained in Rasiowa and Sikorski
(1963). In the axiomatics considered there, quantifiers are characterized
by four rules o f inference. Rasiowa and Sikorski prove its completeness in
respect of: (1) the class of all pseudo-Boolean algebras and (2) the class of
topological models constructed over the pseudo-Boolean algebras.
Systematic investigations o f quantification in modal logic were started
by Barcan and Carnap in the 1940s. The subsequent response to Barcan’s
(1946) work makes it especially worth noting that she proposed to extend
S2, the modal propositional logic (a system) ‘ weaker’ than S4 and S5 by
adding new axioms to it, one o f them bounding quantification and modality
as follows:

(B) 0 (3 χ α (χ )) —> 3 x O a (x).

This axiom was consigned to posterity on account of its philosophical im­

plications. The following equivalent of (B)

(B*) V x (d a (x )) —> □ (V xa (x)),

sometimes together with its converse, is known as the Barcan formula.

Principles like (B) and (B*) are, speaking most generally, connections be­
tween two modality modes: de dido (about proposition) and de re (about
thing, an object). Therefore, their concern goes far beyond the logic o f func­
tors and propositional operators. The unwelcome consequence o f adopting
(B) derives from the possibility of inferring from a modal proposition in­
tuitively unjustified existential conclusions; commonly known examples of
 Remarks on quantification 97

such inference are o f the sort:

It is possible that the first woman

president of France will be blond

There exists someone who can be the first

woman president of France and blond.

The premise is readily acceptable while the conclusion is not: the person
spoken about might not have been born yet. The example allows us to
perceive instantly the source o f difficulty met in constructing the seman­
tics o f modal logics o f the first order. No general assumption can be made
that each name has a denotation in every interpretation domain. In many
practical realizations Kripke’s semantics for a definite modal system serves
as the foundation for the construction. And some assumptions concerning
the interpretations o f names in every single world as well as those concern­
ing links between denotations in worlds related with R are considered. At
this point it may be worth while mentioning that Kripke himself defined
the first-order extensions o f modal logics having the semantics established
by the condition that in all possible worlds names have the same inter­
pretations. In the end, we would like to emphasize that the difficulties
which arise when the idea o f quantification is introduced to the realm of
modal logic do not end with the problem o f interpretation o f names. Quite
similar questions appear for predicates, especially the predicate o f identity
which has been much considered (incidentally, not only in the context o f
modality) and which provides some authors such as Quine, for example,
with arguments for the unworkability o f any construction o f the first-order
modal logic, and others with the inspiration for studies on systems o f free
logic deleting any existential assumptions concerning individual constants;
see Lambert and van Fraassen (1972). Garson (1984) showed that free logic
may serve as a foundation for certain systems o f quantified modal logic.
13. Fuzzy sets and Zadeh logic

As is generally acknowledged, the possibility o f applying the apparatus

of traditional formal logic to the analysis o f everyday reasoning is very
limited. However, recent developments in logical studies have considerably
extended it. Our aim in the present chapter is to account for one o f the most
interesting but simultaneously most controversial conceptions inspired by
logical many-valuedness: the fuzzy set theory and logics based on it.

13.1 Fuzzy sets and logics o f imprecise concepts

Zadeh (1965) defines a fuzzy set A o f a given domain U as an abstract
object characterized by generalized characteristic function Ua with values
in the real set [0,1]:

UA : U —> [0,1].

The values of Ua are interpreted as degrees o f membership o f elements of

U to the fuzzy set A. The extreme values of this function, 0 and 1, denote,
respectively, not belonging to A and entire membership of A . Limiting the
scope o f Ua to {0 ,1 } results in an ‘ ordinary’ characteristic function and in
this each ‘ classical’ set is a special case o f a fuzzy set. Likewise, the affinity
between fuzzy sets and inexact classes (see 7.1) or ‘ n-valued’ sets (see 6.3)
can be found.
Fuzzy sets are an instrument for modelling inexact predicates appear­
ing in natural languages. Zadeh suggests, for example, that the prop­
erty o f ‘ being much greater than 1’ defined on the set o f real positives
R+ can be assigned to a fuzzy set W with an arbitrary non-decreasing
characteristic function R+ —> [0,1] which meets the following conditions:
R w ( 0) = 0, 1 M 1 ) = 0, Rw ( 5) = 0.01, jRnr(lOO) = 0.95, Rw {500) = 1,
etc. As one may easily guess, in the above example only the values Rw (0),
R w { 1) and the mono tonicity of Rw are unquestionable. The selection of
other values is somehow arbitrary; nonetheless it should be subjected to
the problem constituting the context of W predicate modelling.
In the family F(U) of fuzzy (sub)sets of a given domain 17, Zadeh (1965)
introduces the relation of inclusion putting
 Fuzzy sets and logics o f imprecise concepts 99

AC B if and only if Ua { x ) < UB(x) for any x E 17,

and the counterparts of the operations o f complement ( —), union (U) and
intersection (Π) through the following specification o f the characteristic
functions:

U - A(x) = 1 - UA(x)
Ua ub ( x ) = max{UA(x), UB(x)}
Va ub { x ) = min{UA(x), Ub ( x )}·

Bellman and Giertz (1973) showed that UauB and Uac\b are the unique
non-decreasing continuous functions warranting both the compatibility of
the construction with the standard algebra of sets and the fact that (F(U),
U, Π, —) is a de Morgan lattice if and only if U- a is defined as above.
In spite o f the naturalness of the proposal, several studies admit as (more)
helpful, fuzzy set algebras defined otherwise. An example is the algebra
considered by Zadeh (1976) having the operations U and Π defined by the
schemes

Ua u b ( x ) = UA(x) + UB(x) - Ua ( x ) · UB(x)

UAnB (x) = UA(x )-U B (x)

and being somewhat akin to the rules for calculating the measures o f unions
and intersections o f Borel sets on independent events in the probability
theory. These operations satisfy necessary conditions that reduce to the
classical ones in the binary case. However, no function gives the idempo-
tency, i.e. Uaua and Uaha coincide with Ua , and taken together they do
not give the distributivity (compare 1.6). In the following, by a fuzzy set
algebra we shall mean the first of Zadeh’s constructions.
The values o f generalized characteristic functions may be identified with
the logical values of propositions obtained from the predicates serving as
a basis for determining (particular) fuzzy sets. And, consequently, with
fuzzy set algebra there can be associated an uncountably many-valued logic.
The most transparent manner of presenting the mutual relation between
the base logic and fuzzy set algebra amounts to stating the inclusion and
algebraic operations by means of ‘generalized’ set-theoretical predicate E
and logical constants:

A C B =df Vx(x E A —►x E B)

—A = {x : —'(x E A )}
4 (J B = { x : x E i V x E B }
AC\B = {x : x e A A x e B }.

For Zadeh’s algebra the choice o f a base logic is to a great extent prejudiced:
it (this logic) must be based on the Ni-element matrix, in which negation
is expressed by the function 1 —p, disjunction and conjunction respectively
100 13. Fuzzy sets and Zadeh logic

by m ax{p,q} and m in {p ,g }, and the universal quantifier by the greatest

lower bound (inf). The function of implication is not uniquely determined.
However, evidently it should meet the requirement:

(.) if P —* Q = 1? then p < q.

Especially noteworthy implications for which (.) holds are those o f Lu­
kasiewicz and Godel (defined on [0,1]) (see 5.1 and 12.1). It may be of
interest that they are somehow maintained by the probability theory, the
former by subjective (compare 9.2) and the latter by conditional (see Gaines
(1976b)). Lukasiewicz’s implication and, consequently, his Ni-valued logic,
have been the most intensely applied. Nevertheless, the commonly shared
belief among scholars working on fuzziness, both theoreticians and practi­
cians, is that only a concrete application o f fuzzy set algebra can decide
the final form of the base logic (see Gaines (1976a)).
In the initial applications of fuzzy set theory much attention was focused
on the analysis o f reasoning using inexact predicates, standard proposi­
tional connectives and quantifiers. Hence, basic logics have mainly been
considered, though not without innovations. Remarkably susceptible to ex­
periment, there appeared to be paradoxes (antinomies) whose ‘successful’
analysis consolidated the motivational layer of the conception o f fuzziness.
This account yielded the (first) understanding o f the term ‘fuzzy logic’ as
a certain class o f many-valued logics with uncountably many values, with
Lukasiewicz logic in the foreground. A typical case of modelling an inexact
predicate in the latter is an attempt to analyse the classical paradox o f a
bald man by Goguen (1969). Intuitively, we would be ready to accept the
following two propositions:

(z l) A man with 20,000 hairs on his head is not bald

(z2) A man who has one hair less than somebody who
is not bald is not bald as well.

Applying the detachment rule 20,000 times, we get, by (z l) and (z2), the
conclusion that a man with no hair is not bald either. Naturally, the para­
dox stems from (z2) and more specifically from inexactness of the predicate
‘bald’ or, equivalently, ‘not-bald’. The paradox will vanish when the logical
value o f the proposition ‘ A man with n hairs is not bald’ is identified with
the degree o f membership of a man with n hairs to a fuzzyset‘ not-bald’ ,
since then (z2) would have a logical value less than 1, say1 — ε, where
ε > 0. If, for instance, in basic logic we use Lukasiewicz’s implication then
as a result o f 20,000 derivations we will obtain a proposition o f the logical
value amounting to 1 — 2 0,000ε, thus practically false. (We are leaving
scope for the reader to guess what value o f ε warrants this falsity!).
The development of fuzzy set theory has surpassed all expectations. For
its use almost all the important concepts of set theory, topology, algebra
 Fuzzy logic 101

and probability calculus were adopted. It is remarkable that these were the
applications in computer science and steering theory that gave an impetus
to the most extensive development o f the theory: they have confirmed
the usefulness o f the unconventional methodology worked out by means of
fuzzy set conception (see e.g. Gottwald (1981)).

13.2 Fuzzy logic

Zadeh’s (1975) fuzzy logic is a method of modelling (formalization) impre­
cise reasoning operating on imprecise concepts, more specifically inexact
predicates and truth-functional terms. Its construction conveyed the belief
that thinking in terms of fuzzy sets is a typical feature o f human percep­
tion. The examples of reasoning whose analysis might be possible due to
fuzzy logic are such ‘inferences’ as:

Reagan is very healthy Mario likes women that are tall and
Healthy men live a very long wicked
time Eva is not very tall but very wicked

Reagan will live a very long time Mario will probably like Eva.

Fuzzy logic seeks to formulate several rules o f approximate inference.

For this purpose it attempts to formalize colloquial linguistic usage o f cer­
tain ‘ hedges’ applied to imprecise concepts such as ‘ very’ , ‘ more or less’ ,
‘ too’ , etc. Zadeh logic is a two-level semantical construction allowing the
fuzzines o f predicates, their hedges and logical values. Its central elements
are:

(1) The denumerable set T V of linguistic logical values generated by

(means of) its element ‘ true’ with the help o f the hedge ‘ very’ and
logical (propositional) connectives.
(2) Hedges o f predicates and logical values, ‘ very’ being a special one
(compare ( 1)).
(3) The procedure of linguistic approximation compensating for the lack
o f closure o f the (object) language and the closure o f the set T V onto
logical connectives.

Fuzzy logic is formed on the basis of an ordinary Νχ-valued logic with the
connectives V, A, = and values in [0,1]. Its constituting principle
consists o f identifying predicates with fuzzy subsets o f a given universe
and logical values with fuzzy subsets o f the set o f values o f the basic logic,
i.e. [0,1]. The most frequently discussed is the ‘fundamental’ system FL
obtained on Lukasiewicz L^l (see Bellman and Zadeh (1977)). We shall
confine our further considerations to some aspects o f this construction.
102 13. Fuzzy sets and Zadeh logic

It is assumed that the set T V o f linguistic logical values o f FL is a set

o f the form

TV = { true, false, not true, not false, very true, more or less true,
true, not very true, not very false, . . . }

where ‘ true’ is a fixed fuzzy subset o f [0,1]? ‘ very’ a fixed hedge, and all other
elements are defined through ‘ true’ , ‘very’ and (operations determined by)
Lukasiewicz’s connectives. Obviously, the ‘names’ o f all logical values are
conventional labels and their proper meanings follow from characteristic
functions.
Hedges are one-argument operations sending fuzzy sets to fuzzy sets.
The most basic (a primitive) is a hedge reflecting the adverb ‘ very’ and
denoted as g. Zadeh (1972) supposes

(i) UgA(x) = (UA(x ) ) 2

and suggests that other hedges should be defined as superpositions o f g

and connectives o f the basic logic. The most important ‘ derived’ operator
d is described as follows:

(d) UdA(x) = ( UA(x))0S.

Zadeh urges that ‘more or less’ is a linguistic counterpart o f d. In the

relevant literature g and d are called standard hedges and expressibility by
their use has become a hedge definability criterion. Conversely, the direct
searching for other reasonable linguistic hedges is based on the analysis o f
compound statements containing g. Thus, if p is the statement:

The vase is beautiful (vase E beautiful object),

then dp = p A -*gp, i.e. ‘The vase is beautiful but not very beautiful’ may
be condensed to ‘The vase is slightly beautiful’ 19. Accepting that Λ and
are Lukasiewicz’s connectives we obtain:

u,lightly a (s ) = m in {U A(x), 1 - ( UA(x ))2}.

The procedure o f linguistic approximation originally stems from the
lack o f closure o f T V under logical connectives. More specifically, if Utrue
is a characteristic function o f the value labelled ‘ true’ then, o f course, all
other elements o f that set are thereby specified. Thus, for example, the
initial values o f T V are defined by the following functions:

19The argumentation supporting the given characterization o f ‘rather’ comes from

Zadeh. Obviously, the choice o f the very linguistic term for this modifier may be
disputable.
 Fuzzy logic 103

U fa ls e {x ) — £^tr«e(l ®)
Unot t r u e i^ ) = 1 £^tr«e(®)
Uvery tr u e (x>) = (U t r „ « ( x ) ) 2
U m ore or less tr u e ( x ) = ( t w * ) ) 0·5,

( x ranges over the set o f values o f the base logic, x E [0,1]). Consequently,
all linguistic logical values depend on the (subjective) introduction o f the
meaning o f ‘ true’ . Zadeh calls this feature a localization and the elements
o f T V , local values. Thus obtained, as a result o f localization, the sys­
tems o f FL are called local logics. For those logics there is the common
way o f defining logical connectives; it consists o f identifying them with the
operations o f the algebra o f fuzzy subsets o f [0,1] and thus, in a sense,
with connectives o f the base logic. Using the ‘ generalized’ E (see 13.1)
the meanings o f ~ , =Φ>, V’ and A’ o f FL may be approached by the equiv­
alences

x e [~ A]
x E [A ==> B]
x E [ i V ’ B]
x € [A A’ 5]
where - 1, — V, A, = are Lukasiewicz’s connectives and [Z] is a fuzzy set
representing, or to represent (on the left-hand side), a relevant linguistic
logical value. These postulates, however, warrant the closure o f the T V 0
set for practically none o f the local logics FL° - if, for example, a and β
are statements having values ‘rather true’ and ‘not very false and not very
true’ , respectively, then the value o f the statement a Α β generally is not a
linguistic value o f FL°. A linguistic approximation is a heuristic procedure
assigning linguistic logical values to any proposition. For obvious reasons it
is impossible to specify even a general principle o f that procedure; what can
be said is merely that in a concrete application o f a fuzzy logic it consists
o f searching for the closest value from T V 0 for a statement. Assuming that
LA is a predicate o f approximation one may characterize FL-connectives
as below:

[~ 4 = LA{ x : -*(x € [j4])}

[A ^B ] = L A{ x : (x 6 [A] —►x E [B])}
[ A y B] = LA{ x : ( x E [ i ] V x E [£ ])}
U a B] = LA{ x : ( x E [A] A x E [ 5 ] ) } .

In much the same way the concept o f approximation can be applied to

predicates and hedges.
As mentioned already, fuzzy logic aims at formulating the rules o f ap­
proximate reasoning (or inference). The basic as well standard rule o f that
kind is the compositional rule of inference
104 13. Fuzzy sets and Zadeh logic

ui is F Ui and U2 are F
Ui and U2 are G U2 and U3 are G

U2 is L A {F * G } Ui
Ui and
and u$
u$ are
are LLAA{F
{F **GG}}

where Ui and U2 are objects, F and G predicates (properties or relations)

and * an operation o f relational composition. In the former case * is defined
as follows:

U f * g ( u 2) — supx (U f {x ) A U g {x , u 2) ) ;

and L A { F * G} is a linguistic approximation to the (unary) fuzzy relation

F *G. The following is an exemplification o f the first scheme:

a is a small number
a and b arc approximately equal
b is more or less a small number

In more complicated rules, the premises are compound statements o f several

degrees o f complexity, e.g. if u\ is F then U2 is G, and may be quantified
by fuzzy quantifiers, i.e. expressions like most, many, several, few, etc. In
the formulation o f other rules o f fuzzy inference special operations on fuzzy
sets are used - here a good example is the rule of compositional modus
ponens

Ui is F
I f Ui is G, then U2 is H

u 2 is L A { F * ( G ' Q H ) }

where G ' φ ϋ Γ is the so-called bounded sum of cylindrification of G' (the

complement o f G) and F (see Bellman and Zadeh (1977)).
W hat constitutes the basis o f the practical application o f fuzzy inference
rules is assigning fuzzy predicates fuzzy subsets o f a given universe; in the
case o f (non-unary) relations, fuzzy subsets o f relevant Cartesian products
o f their domains. This procedure, often referred to as fuzzy restriction,
obviously exceeds logic. Notwithstanding this, within the scope o f a logic
it is possible and even advisable to formulate the general principles of
restriction resulting from the peculiarity o f formal devices.
Zadeh’s ‘fuzzy’ conception has found its place among accepted methods
o f artificial intelligence. It holds its ground due to reliable applications as,
for example, in medical diagnosis; see Turner (1984).
14. Applications and significance of
the topic

Projections o f the expected applications always constituted a distinguished

motivation for many-valued logical constructions. At first, these concep­
tions, as with Lukasiewicz, were a philosophical revolution, but there were
also much restricted ones, as with Kleene or Post. And, it is not at all
easy to answer the question o f whether and to what extent the expecta­
tions o f the authors o f several many-valued logical systems actually came
to fruition. Independently o f that, however, some concrete, worthy appli­
cations o f many-valued logics and algebras can be indicated; for example,
the use o f logical matrices to demonstrate the independence o f axioms, to
formalize (some) intensional functions or the application o f Moisil’s and
Post’s algebras in electrical network theory. A little look at those uses, at
the very least, gives a view o f the significance o f the topic. Just for this
purpose we will present some examples o f the applications o f many-valued
matrices and algebras to philosophical logic and to such practical areas as
switching theory and computer science.

14.1 Independence of axioms

The axiomatics o f a given logical system is independent when none o f its
axioms can be inferred from others by means o f the inference rules adopted.
The property o f independence not only prejudices the Econom y’ o f the
selection o f axioms but has a great methodological importance as well.
The method o f testing the independence o f axioms consisting o f the con­
struction o f algebras (or matrices) is credited to Bernays and Lukasiewicz.
To demonstrate that an axiom system is independent one singles out a
property which is common to all axioms besides the one chosen and is in­
herited by all those theorems o f the system that are derivable from the
axioms satisfying the property. The procedure is repeated as many times
as there is number o f axioms in the system. In the case o f applying logical
matrices, tautologousness, i.e. pertaining to the content o f a (many-valued)
matrix, is the most often used property o f formulae.
The application o f the method can be illustrated by the following ex­
ample. Consider a (-·, —►J-system o f the classical propositional calculus
106 14. Applications and significance o f the topic

originated by Lukasiewicz (see 1.3). Its axioms are

(Al) (->p - * p ) - + p
(A2) p —►(~>p —» q)
(A3) (p - > q ) ~ * ((q -> r) - » (p -► r)),

and the rules MP and SUB. Let now

Μ (Λ1) = ( { 0, 1}, - 1, { 1})

M(A2) = ( { 0, 1), - 2, { 1})

be matrices in which the implication connective —►is determined classically

(by the well-known truth-table), —»i(0) = —*i(l) = 0 and ->2(0) = “ •2( 1) = 1·
Moreover, let

M (A3) = ({o, 1/ 2, 1}, { 1})

be a matrix with the connectives defined by the tables:

a ->α -► 0 1/2 1
0 1 0 1 1 1
1/2 1/2 1/2 1 0 1
1 1 1 0 0 1

It is readily seen that E(M^A1)), E(M^A2 )) and E(M(A3)) are closed

under MP and SUB and that

(1) A l £ E ( M (A1)), A 2 € E ( MiA1)) and A3 E E ( M {A1))

(2) A l e E ( M (A2)), A 2 4 E ( M (A2)) and A3 E E ( M {A2))
(3) A l 6 E ( M (A3)), A 2 e E( M(a3)) and A3 4 E { M (A3)).

Therefore the axiomatics (A 1 )-(A 3 ) is independent.

The application o f the method described is not limited to the logical
calculi. The proofs of independence in set theory through the use o f ma­
trices built on the base o f Boolean algebras were presented by Scott and
Solovay (1969).

14.2 Formalization of intensional functions

Intensional propositional functions, in contrast to extensional functions, are
those functions o f propositional arguments whose logical value (its truth
or falsity) does not depend solely on the logical values o f the arguments.
Consequently, on the propositional calculus’ level intensionality is identi­
fied with non-truth-functionality. In that sense (almost) all connectives o f
many-valued logics are intensional. At the same time, as the example of
modal and intuitionistic logics shows, the most natural way to formalize
intensional propositional connectives is by their axiomatization.
 Formalization o f intensional functions 107

The application o f logical many-valuedness is not restricted to proposi­

tional functions. Los (1948) showed that, under some reasonable assump­
tions, the formalization o f functions o f the kind ‘John believes that p’ or,
more accurately, ‘John asserts, that p’ naturally leads to the many-valued
interpretation o f the belief-operators within the scope o f the classical logic
system.
All propositions ‘John asserts, that p’ are clearly substitutions (exem­
plifications) o f the schema ‘ x asserts that p’ , whose formal linguistic coun­
terpart is a function Lxp of two arguments x and p, assigning a logical
value to each couple (name, proposition). Hence the simplest language in
which syntactic characterization o f the (operator) L is possible, is the lan­
guage £ including two kinds o f variables, nominal and propositional, and,
additionally, quantifiers. Just such a language is considered by Los who,
as axioms o f his system for L, assumes the formulae

(L I) Lxp = -*Lx(->p)
(L2 i) Lx(Ai), where (Αί), i E {1 ,2 ,3 }, is the ith axiom
o f Lukasiewicz’s (-·, —^-system o f CPC
(L3) Lx(p —> q) -+ (Lxp —►Lxq)
(L4) ( Mx)Lxp —►p
(L5) LxLxp = Lxp

accepting the rules MP, the substitution rule extended onto the whole lan­
guage £ and the generalization rule. The intuitions captured by the author
are comprehensive and justified. By way o f example it is worth noting that
(L3) expresses the fact that everyone uses the detachment rule, i.e. assert­
ing a conditional statement and acknowledging the antecendent o f that
statement commiting one to assert the consequent, or that (L4) says that
a sentence acknowledged by everyone is a theorem o f the system.
The operators Lx, L y ,. . . are certainly not the only intensional functions
o f the system considered. What is more, the closed formulae o f the language
define intensional propositional functions, i.e.connectives. The case here is
the function S

Sol =df 3x3y(L x a Λ L y(-»a)),

which can be interpreted as ‘ it is questionable that a ’ . Intuitively, the

definition conveys the thought that saying ‘ it is debatable that a ’ means
‘ two people exist such that one asserts a , and the other asserts n ot-a ’ .
Any interpretation o f the system o f L-operators (and derived opera­
tors) starts with the selection of a definite range (domain) o f nominal
variables and a proposition set. The simplest is the case o f two persons
A and B ; let us suppose that they do not agree on all the issues, i.e.
that in a certain definite set o f propositions Z there are some accepted
by A but not accepted by B and vice versa. Then the assent property
108 14. Applications and significance o f the topic

divides Z into four classes, which shall be denoted symbolically as: 0,

1/3, 2/3, 1. The first class, 0, contains propositions which are ac­
knowledged by neither person, the second class, 1/3, propositions which
A acknowledges and B does not, and the third class, 2 /3 , propositions
which B acknowledges and A does not. Finally, the fourth class, 1, con­
tains all propositions acknowledged by A and H, i.e. all logical theorems
and perhaps other propositions acknowledged by both. Identifying the ac­
ceptance o f a proposition a by a person s with the formula Lsa we get a
truth-table assigning to classes 0, 1/3, 2/3, 1 the
pairs o f logical values o f truth and falsity (nota-
L A B
0 f f
1/3 t f
2 /3 f t
1 t t

a -ια S ol -► 0 1/3 2 /3 1
0 1 0 0 1 1 1 1
1/3 2 /3 1 1/3 2 /3 1 2 /3 1
2 /3 1/3 1 2/3 1/3 1/3 1 1
1 0 0 1 0 1/3 2 /3 1

Los suggests that the objects 0, 1/3, 2/3, 1 may be treated as logical
values. He further considers the matrix

14 = ({0 , 1/3, 2 /3 , 1}, {1 }),

demonstrating that E( 14) = TAUT{-*, —►} (the set o f classical (-·, —►)-
tautologies) and hence that the many-valuedness o f this system has just
formal character (compare 4.1).
The shifting o f Los’s interpretation onto those cases with more persons
is straightforward, and it results in formally many-valued versions o f CPC
(described by matrices with more than four elements).
Los’s construction shows that it is possible to get a many-valued
interpretation o f some intensional (epistemic) functions simultaneously ad­
hering to the intuition o f bivalence. The very method o f formalization,
however, has a more general aspect: the function L considered by Los ‘ uni­
fies’ a name and a proposition thus producing (one) proposition. Therefore,

200 ,1 /3 ,2 /3 ,1 are identified with pairs (x, y), where x , y £ {f,t }, defined by the table
for L . The connectives are defined just as in the product of matrices, i.e. -«(x,y ) =
and ( x i , y i ) (x2,y2) = (*i -> * 2 ,yi -> y2 )·
 Many-valued algebras and switching theory 109

many-valuedness appearing as a result o f interpretation is a certain qualita­

tive feature determining the relation o f two categorially different arguments
and as such is an untypical semantic correlate.

14.3 Many-valued algebras and switching theory

The successful applications o f the classical logic and Boolean algebras in
switching theory are widely known. Also, the use o f the theory o f Boolean
algebras in computer science became well established as a result o f the
actual construction o f computers. At the same time, the utility o f other
algebraic structures was enhanced, as, for example, the use o f (mathemati­
cal) groups for the description o f multi-positional switches. As early as the
1950s, interests centred on the possibility o f the use o f many-valued logic
algebras for similar purposes (see e.g. Epstein, Frieder and Rine (1974)).
These interests brought about the birth o f several techniques for the analy­
sis and synthesis o f electronic circuits and relays based (mainly) on Moisil’s
and Post’s algebras (see Rine (1977)). Since even the most superficial ac­
count o f this highly developed and specialized field goes far beyond the
intended limits o f this book, we shall confine ourselves to some remarks
justifying the purposefulness o f using many-valued algebras in switching
and relay circuit theory.
The most elementary composite o f the traditional electronic circuit is
a mechanical contact opening and closing some fragment o f an electrical
network. The switchover o f contacts is effected mechanically or electrome-
chanically (i.e. using relays). And, among the contacts o f a given network
one may find such pairs o f contacts which, according to the technical as­
sumptions, have to change their complementary positions simultaneously.
The simplest example o f such a situation is the arrangement o f two op­
positely oriented contacts x l and x2 positioned in parallel branches o f

xl xl x2

x2

Fig. 1 Fig. 2

a circuit (see Fig. 1): x l is normally closed while x2 is normally open.

When considering the ideal model o f the circuit one assumes that both
contacts react momentarily and thus move, as shown in Fig. 2, from the
state (1) to (0). Practically, however, it may happen that the contact x l
110 14- Applications and significance o f the topic

will open before x2 is closed and, consequently, contrary to the technical

presumptions, the circuit during this moment will be open. This is just the
reason for such a modelling in which the third
state (1 /2 ) (see Fig. 2) is considered; the table z xl x2
here characterizes the ‘ real’ switching function as 1 11 1 1
a function o f states and contacts (1 inside the ta- 1/2 0 1
ble denotes normal contact state and 0 its denial). 0 0 0
On the other hand, one also may read the table as
treating x l and x2 as (one-argument) functions o f states, i.e.of z, and their
values as states as well, putting x l = $χ(ζ), x2 = $2(2) · Let us notice then
that Si and $2 are Moisil’s operations on {0, 1/2, 1} or, equivalently, the
modal connectives L and M o f Lukasiewicz (see 5.4 and 2.3). Subsequently,
for getting a tool to describe the function state o f an arbitrary network built
o f the contacts x l , x2 and their complements x l, x2, one should define
two binary operations U and Π corresponding to two possible types o f
connectives (parallel and serial) and unary operation N such that xi =
N xi (2 = 1, 2) and that N N z = z for z € { 0, 1/2, 1 } . It appears that
the most accurate way o f introducing these operations leads to the three­
valued Moisil algebra on {0, 1/2, 1} : ({0, 1/2, 1}, U, Π, N, $i, $2)·
A generalization of the outlined construction for the case o f any number
of contacts similarly results in n-valued algebras. The algebraic treatment
of switching devices aims to provide several techniques for the analysis,
synthesis and minimalization o f multiplex networks. The most important
advantage o f the many-valued approach is the possibility o f eliminating
possible switching disturbances through the algebraic synthesis o f the net­
works (see e.g. Moisil (1966)).
The application o f many-valued algebras is not limited to binary con­
tacts. Investigations concerning multi-stable contacts and switches have
also been undertaken. However, as a result o f the difficulties which arise
with the technical realizations o f devices utilizing voltages, the possibilities
created by the highly developed theory o f many-valued algebras are utilized
only to a certain degree, particularly in computer technology. This course
o f events may be partially caused by the extraordinary progress o f tech­
nology for the production of binary components, including very large scale
integration. Nevertheless, many-valued logical constructions still attract
the attention o f engineers and computer scientists. The reader interested
in the current state o f investigation in this field is advised to consult Com­
puter, the IEEE Computer Society monthly, as well as the yearly editions o f
this influential American society’s Proceedings of the Multiple- Valued Logic
Symposia.
 Many-valuedness in computer science 111

14.4 Many-valuedness in computer science

One objective o f computer science is to investigate digital computing. The
chief domains o f this discipline concern two kinds o f problem: the com­
puter’s architecture (hardware) and the programming (software). It may
be said that logical many-valuedness has found successful use in both these
branches.
There were several technical difficulties (mentioned above) that led to a
reduction o f serious efforts aimed at the full hardware realization o f many­
valued computers. Full scale ternary computers were successfully com­
pleted at least twice: the arithmetical SETUN in the USSR in 1958 and
the logico-arithmetical TERNAC in the USA in 1973. However, the first
emulation o f TERN AC proved that both the speed and price are on the
same order o f magnitude as the speed and price o f binary computers. The
same experiments showed the necessity o f lengthening the instruction set
o f the machine (i.e.in the assembly language) and, therefore, the complica­
tion o f programming languages. Consequently, more attention was directed
towards the questions o f synthesis and construction o f several fragments or
electronic components o f digital devices (e.g. memories) using many-valued
algebras (see Epstein, Frieder and Rine (1974)). It is not excluded that fur­
ther developments o f new electronic technologies and software rationaliza­
tion will resurrect the now forgotten ideas o f building valuable many-valued
computers.
The Post algebras found an important application in the systemati­
zation o f theoretical research concerning programs and higher level pro­
gramming languages which contain instruction branching programs, such
as CASE, SELECT, etc. The application o f these instructions consider­
ably simplifies programs, thus making them more readable. In turn, the
structural reconstruction o f a branched program naturally leads to a Post
algebra o f order n. The typical CASE (or SELECT) situation is that in
which one the (sub)programs Ρ Ι , . , . , Ρ η should be performed according
to whether conditions W l or . . . or W n are satisfied. Then, the constant
functions o f the Post algebra, e o , . . . , e n_i, may be interpreted as devices
which keep track o f which o f W 1 , . . . , W n are true.
The most powerful tool o f the contemporary methodology o f program­
ming languages, algorithmic logic (see Salwicki (1970)), is formulated in
the language containing operators which represent composition, branch­
ing and iteration operations on programs. The systems o f this logic con­
tain expressions representing programs and formulae describing the prop­
erties o f these programs. Rasiowa’s ω +-valued extension o f algorithmic
logic not only is fully adapted to arbitrary ‘ wide’ branching programs
but it also constitutes a starting point for other, more advanced logical
constructions (see Rasiowa (1977)). Its semantics is based on Post alge­
bras o f order ω+ defined similarly as Post algebras o f order n in Rousseau
112 14. Applications and significance o f the topic

version. The simplest algebra o f that kind is o f the following form: Ρω =

U, Π, {ei}o<i<u>) (0, l , . . . , w are ordinal
numbers), where —βω — e0 < e\ < . . . < ew, e,· U e* = emax (,·,*), e, Π
ejk = emt-n ( ^ ) , e, ek = when i < fc, e, ek = ek when
i > fc, —e,· = e0 for z ^ 0, —eo = ew, 27; (e*) = if i < k and
Di(ek) = eo otherwise (compare 6.4).
Algorithmic logic is among the main predecessors o f the recent outcome
o f computing theory called dynamic logic (see e.g.Harel (1984)). The term
dynamic logic is the generic name given to logical systems appropriate for
reasoning about changes from one state to another, which ultimately may
also represent programs. Though it is based on features o f programming
and program verification, dynamic logic has borrowed heavily from modal
logic and has an autonomous philosophical importance.
Bibliography
Ackermann, R. (*1967). Introduction to many-valued logics. Routledge
and Kegan Paul, London.

Barcan, R. C. (Mrs. Marcus) (1946). A functional calculus o f first order

based on strict implication. The Journal of Symbolic Logic, 11, 1-16.

Bellman, R. E. and Giertz, M. (1973). On the analytic formalism o f the

theory o f fuzzy sets. Information Sciences, 5, 149-156.

Bellman, R. E. and Zadeh, L. A. (1977). Local and fuzzy logics [in:] Dunn,
J. M. and Epstein, G. (eds) Modem uses of multiple-valued logic. D. Reidel,
Dordrecht, 105-165.

Belluce, L. P. and Chang, C. C. (1963). A weak completeness theorem for

infinite-valued first-order logic. The Journal of Symbolic Logic, 28, 43-50.

Belnap, N. D. (1970). Conditional assertion and restricted quantification.

Nous, 4, 1-13.

Beth, E. W . (1956). Semantic construction o f intuitionistic logic. Med-

edelingen der koninklijke Nederlandse Akademie van Wetenhappen, new
series, 19, no. 11, 357-388.

Blarney, S. (1986). Partial logic [in:] Gabbay, D. and Guenthner, F. (eds)

Handbook of philosophical logic, vol. ΙΠ . D. Reidel, Dordrecht, 1-70.

Bloom, S. L. and Brown, D. J. (1973). Classical abstract logics. Disserta-

tiones Mathematicae, 102, 43-52.

Bochvar, D. A. (1938). Ob odnom trehznacnom iscislenii i ego primenenii

k analizu paradosov klassiceskogo rassirennogo funkcjonal’nogo iscislenia
(On a three-valued calculus and its application to analysis o f paradoxes
o f classical extended functional calculus). Matematiceskij Sbomik, 4, 287-
308.

Carnap, R. (1946). Modalities and quantification. The Journal of Symbolic

Logic, 11, 33-64.

Carnap, R. (1950). Logical foudations of probability. University o f Chicago

114 Bibliography

Press, Chicago.

Chang, C. C. (1958a). Proof of an axiom o f Lukasiewicz. Transactions of

the American Mathematical Society, 87, 55-56.

Chang, C. C. (1958b). Algebraic analysis o f many-valued logics. Transac­

tions of the American Mathematical Society, 88, 467-490.

Chang, C. C. (1959). A new proof of the completeness o f the Lukasiewicz

axioms. Transactions of the American Mathematical Society, 93, 74-80.

Chang, C. C. (1963). The axiom o f comprehension in infinite-valued logic.

Mathematica Scandinavica, 13, 9-30.

Chang, C. C. and Keisler, H. J. (1966). Continuous model theory. Prince­

ton University Press, Princeton, New Jersey.

Chang, C. C. and Keisler, H. J. (1973). Model theory. North-Holland,

Amsterdam.

Church, A. (1956). Introduction to mathematical logic, vol. I. Princeton.

Cleave, J. P. (1974). The notion o f logical consequence in the logic of

inexact predicates. Zeitschrift fur Mathematische Logik und Grundlagen
der Mathematik, 20, 307-324.

Czelakowski, J. (1980). Model-theoretic methods in methodology of propo­

sitional calculi. Polish Academy o f Sciences, Institute o f Philosophy and
Sociology, Warsaw.

Dalen, D. van. (1986). Intuitionistic logic [in:] Gabbay, D. and Guenthner,

F. (eds) Handbook of philosophical logic, vol. III. D. Reidel, Dordrecht,
225-339.

Dugundji, J. (1940). Note on a property o f matrices for Lewis and Lang­

ford’s calculi o f propositions. The Journal of Symbolic Logic, 5, 150-151.

Dumniet, M. (1959). A propositional matrix with denumerable matrix.

The Journal of Symbolic Logic, 24, 97-106.

Dwinger, Ph. (1977). A survey of the theory o f Post algebras and their
generalizations [in:] Dunn, J. M. and Epstein, G. (eds) Modem uses of
multiple-valued logic. D. Reidel, Dordrecht, 53-75.

Epstein, G. (1960). The lattice theory o f Post algebras. Transactions of

the American Mathematical Society, 95, 300-317.

Epstein, G., Frieder, G. and Rine, D. C. (1974). The development o f

 Bibliography 115

multiple-valued logic as related to Computer Science. Computer, 7, no. 9,

20-32.

Fenstad, J. E. (1964). On the consistency o f the axiom o f comprehension

in the Lukasiewicz infinite-valued logic. Maihematica Scandinavica, 14,
64-74.

Fitting, M. C. (1969). Intuitionistic Logic, Model Theory and Forcing.

North-Holland, Amsterdam.

Fraassen, B. C. van (1966). Singular terms, truth-value gaps and free logic.
Journal of Philosophy, 63, 481-495.

Frege, G. (1879). Begriffsschrift, eine der arithmetischen nachgebildete

Formelsprachc des reinen Dendens. Halle.

Frege, G. (1892). Uber Sinn und Bedeutung. Zeitschrift fur Philosophie

und philosophischc Kritik C , 25-50.

Gaines, B. R. (1976a). Foundations o f fuzzy reasoning. International Jour­

nal of Man-Machine Studies, 8, 623-668.

Gaines, B. R. (1976b). General fuzzy logics [in:] Proceedings of the 3rd

European Meeting on Cybernetics and Systems Research. Vienna.

Gardies, J.-L. and Malinowski, G. (1981). D ’un certain usage de la triva­

lence en logique deontique. Logique et Analyse, 94, 179-199.

Garson, J. W . (1984). Quantification in modal logic [in:] Gabbay, D. and

Guenthner, F. (eds) Handbook of philosophical logic, vol. II. D. Reidel,
Dordrecht, 249-307.

Giles, R. (1974). A non-classical logic for physics. Studia Logica, 33, 397-
416.

Godel, K. (1930). Die Vollstandigkeit der Axiome des logischen Funktio-

nenkalkuls. Monatschefte fur Mathematik und Physik, 37, 349-360.

G5del, K. (1932). Zum intuitionistischen Aussagenkalkul. Akademie der

Wissenschaften in Wien, Mathematisch-naturwissenschaftliche Klasse.
Anzeiger, LXIX, 65-66.

Godel, K. (1933). Eine Interpretation des intuitionistischen Aussagenkal-

kuls. Ergebnisse eines mathematischen Kolloquiums, IV , 34-40.

Goguen, J. A. (1969). The logic o f inexact concepts. Syntliese, 19, 325-

373.
116 Bibliography

Gonseth, F. (ed.) (1941). Les entretiens de Zurich sur les fondements et

la methode dcs sciences mathematiques 6-9 decembre 1938. Zurich.

Gottwald, S. (1981). Fuzzy-Mengen und ihre Anwendungen. Ein Uberblick.

Elektronische Informationsverarbeitung und Kybemetik , 17, 207-233.

Gratzer, G. (1968). Universal algebra. D. van Nostrand Co., Princeton,

New Jersey.

Grigolia, R. (1977). Algebraic analysis o f Lukasiewicz-Tarski’s n-valued

logical systems [in:] Wojcicki, R. and Malinowski, G. (eds) Selected papers
on Lukasiewicz sentential calculi. Ossolineum, Wroclaw, 81-92.

Grzegorczyk, A. (1964). A philosophically plausible formal interpretation

o f intuitionistic logic. Indagationes Mathematicae, X X V I , 596-601.

Harel, D. (1984). Dynamic logic [in:] Gabbay, D. and Guenthner, F. (eds)

Handbook of philosophical logic, vol. II. D. Reidel, Dordrecht, 497-604.

Hay, L. S. (1963). Axiomatization o f the infinite-valued predicate calculus.

The Journal of Symbolic Logic, 28, 77-86.

Hey ting, A. (1930). Die formalen Regeln der intuitionistischen Logik.

Sitzungsberichte der Preussiche Akademie der Wissenschaften, Phys.-Math.
Klasse, 42-56.

Hey ting, A. (1966). Intuitionism. An introduction. North-Holland, Ams­

terdam.

Hilbert, D. and Bernays, P. (1934). Grundlagen der Mathematik, vol. I.

Berlin.

Hilbert, D. and Bernays, P. (1939). Grundlagen der Mathematik, vol. II.

Berlin.

Hintikka, J. (1963). The modes o f modality. Acta Philosophica Fennica,

X V I , 65-81.

Jaskowski, S. (1936). Recherches sur le systeme de la logique intuitioniste.

Actes du Congres International de Philosophic Scientifique VI. Philosophic
de mathematiques. Actualites scientifiques et industrielles 393, Paris, 58-
61.

Johansson, I. (1936). Der Minimalkalkul, ein reduzierter intuitionistischer

Formalismus. Compositio Mathematicae, 4, 119-136.

Kanger, S. (1957). Provability in logic. Stockholm Studies in Philosophy,

1, Stockholm, 1-47.
 Bibliography 117

Kleene, S. C. (1938). On a notation for ordinal numbers. The Journal of

Symbolic Logic, 3, 150-155.

Kleene, S. C. (1952). Introduction to metamathematics. North-Holland,

Amsterdam.

Korner, S. (1966). Experience and theory. Routledge and Kegan Paul,

London.

Kotarbmski, T. (1913). Zagadnienie istnienia przyszlosci (The problem o f

existence o f the future). Przeglqd Filozoficzny, V I . 1.

Kotarbmski, T. (1964). Leqons sur Vhistoire de la logique. Presses Univer-

sitaires de France, Paris.

Kripke, S. (1963). Semantical analysis o f modal logic, I: Normal modal

propositional calculi. Zeitschrift fur Mathematische Logik und Grundlagen
dcr Mathematik, IX , 67-98.

Kripke, S. (1965a). Semantical analysis o f modal logic, II: Non-normal

propositional calculi [in:] Addison, J. W ., Henkin, L. and Tarski, A. (eds)
The Theory of Models. North-Holland, Amsterdam, 206-220.

Kripke, S. (1965b). Semantical analysis o f intuitionistic logic [in:] Cross-

ley, J. N. and Dummet, M. A. E. (eds) Formal Systems and Recursive
Functions. North-Holland, Amsterdam, 92-130.

Lambert, K. and Fraassen, B. C. van (1972). Derivation and counterexam­

ple: an introduction to philosophical logic. Dickenson, Encino, California.

Langholm, T. (1988). Partiality, truth and persistence. CSLI Lecture

Notes, Stanford University.

Lemmon, E. J. (1957). New foundations for Lewis’ modal systems. The

Journal of Symbolic Logic, X X I I , 176-186.

Lemmon, E. J. (1966). Algebraic semantics for modal logics. The Journal

of Symbolic Logic, 31, 46-65 (I), 191-218 (II).

Lewis, C. I. (1918). A survey of symbolic logic. University o f California

Press, Berkeley (new edition 1960).

Lewis, C. I. and Langford, C. H. (1932). Symbolic Logic. Dover, New York

(new edition 1959).

Lorenz, K. (1961). Arithmetik und Logik als Spielc. Kiel.

Lorenzen, P. (1961). Ein dialogisches Konstruktivitatskriterium [in:]

118 Bibliography

Infinitistic Methods. Proceedings of the Symposium on Foundations of

Mathematics. 2-9 September 1959, Warsaw.

Los, J. (1948). Logiki wielowartosciowe a formalizacja funkcji intensjonal-

nych (Many-valued logics and the formalization o f intensional functions).
Kwartalnik Filozoficzny, 17, 59-78.

Lukasiewicz, J. (1906). Analiza i konstrukcja poj^cia przyczyny. Przeglgd

Filozoficzny, 105-179.

Lukasiewicz, J. (1910). 0 zasadzie sprzecznosci u Arystotelesa. Studium

krytyczne. Krakow; English tr. On the principle of contradiction in Aris­
totle. Review of Metaphysics, X X I V , 1971.

Lukasiewicz, J. (1913). Die logischen Grundlagen der Wahrscheinlichkeits-

rechnung. Krakow; English tr. Logical foundations o f probability theory
[in:] Borkowski, L. (ed.) Selected works. North-Holland, Amsterdam, 16-
63.

Lukasiewicz, J. (1920). O logice trojwartosciowej. Ruch Filozoficzny, 5,

170-171. English tr. On three-valued logic [in:] Borkowski, L. (ed.) Se­
lected works. North-Holland, Amsterdam, 87-88.

Lukasiewicz, J. (1929). Elementy logiki matematycznej. Skrypt. Warszawa

(II edn, PW N, Warszawa 1958); English tr. Elements of Mathematical
Logic translated by Wojtasiewicz, O. Pergamon Press, Oxford, 1963.

Lukasiewicz, J. (1930). Phiilosophische Bemerkungen zu mehrwertigen Sys-

temen des Aussagenkalkiils. Comptes rendus des seances de la Societe des
Sciences et des Lettres de Varsovie Cl. Ill, 23, 51-77; English tr. Philo­
sophical remarks on many-valued systems o f propositional logic [in:] Mc­
Call, S. (ed.) Polish Logic 1920-1939. Clarendon Press, Oxford, 1967,
40-65.

Lukasiewicz, J. (1953). A system o f modal logic. Journal of Computing

Systems, 1, 111-149.

Lukasiewicz, J. (1961). Z zagadnien logiki i filozofii. Pisma wybrane.

PW N, Warszawa; English tr. Selected works (ed. Borkowski, L.). North-
Holland, Amsterdam, 1970.

Lukasiewicz, J. and Tarski, A. (1930). Untersuchungen liber den Aus-

sagenkalkhl. Comptes rendus des seances de la Societe des Sciences et des
Lettres de Varsovie Cl. Ill, 23, 30-50.

MacColl, H. (1897). Symbolical reasoning. Mind, 6, 493-510.

 Bibliography 119

McKinsey, J. C. C. and Tarski, A. (1948). Some theorems about the sen­

tential calculi o f Lewis and Heyting. The Journal of Symbolic Logic, 13,
1-14.

McNaughton, R. (1951). A theorem about infinite-valued sentential logic.

The Journal of Symbolic Logic, 16, 1-13.

Malinowski, G. (1977). Classical characterization o f n-valued Lukasiewicz

calculi. Reports on Mathematical Logic, 9, 41-45.

Marciszewski, W . (ed.) (1987). Logika formalna. Zarys encyklopcdyczny z

zastosowaniem do informatyki i lingwistyki (Formal logic: an encyclopaedic
outline with informatics and linguistics applied). PW N, Warszawa.

Meredith, C. A. (1958). The dependence o f an axiom o f Lukasiewicz.

Transactions of the American Mathematical Society, 87, 54.

Moh Shaw-Kwei (1954). Logical paradoxes for many-valued systems. The

Journal of Symbolic Logic, 19, 37-40.

Moisil, G. (1966). Zastosowanie algebr Lukasiewicza do tcorii uktadow

przekaznikowo-stykowych (Application of Lukasiewicz algebras to the study
of relay-contact networks). Ossolineum, Wroclaw (vol. II, 1967 edn).

Moisil, G. (1972). Essais sur les logiques non-chrisipiennes. Editions de

l’Academie de la Republique Socialiste de Roumanie. Bucharest.

Mostowski, A. (1961). Axiomatizability o f some many-valued predicate

calculi. Fundamenta Mathematicae, 50, 165-190.

Nowak, M. (1988). O mozliwosci interpretowania trojwartosciowej logiki

Lukasiewicza metod$ Slupeckiego (On the possibility o f interpreting the
three-valued Lukasiewicz logic using Slupecki’s method). Acta Universi-
tatis Lodziensis, Folia Philosophica, 5, 3-13.

Peirce, C. S. (1885). On the algebra o f logic: a contribution to the philos­

ophy o f notation. American Journal of Mathematics, 7, 180-202.

Picard, S. (1935). Sur les fonctions definies dans les ensembles finis quel-
conques. Fundamenta Mathematicae, 24, 198-302.

Post, E. L. (1920). Introduction to a general theory o f elementary propo­

sitions. Bulletin of the American Mathematical Society, 26, 437.

Post, E. L. (1921). Introduction to a general theory o f elementary propo­

sitions. American Journal of Mathematics, 43, 163-185.

Rasiowa, H. (*1974). An algebraic approach to non-classical logics. North-

120 Bibliography

Holland, Amsterdam; PWN, Warsaw.

Rasiowa, H. (1977). Many-valued algorithmic logic as a tool to investigate

programs [in:] Dunn, J. M. and Epstein, G. (eds) Modem uses of multiple­
valued logic. D. Reidel, Dordrecht, 79-102.

Rasiowa, H. and Sikorski, R. (1963). The Mathematics of Metamathemat­

ics. PW N, Warsaw.

Rautenberg, W . (ed.) (1987). Non-classical logics [in:] Muller, G. H. in

collaboration with Lenski, W . (eds) Ω-bibliography of mathematical logic,
vol. II. Springer-Verlag, Berlin.

Reichenbach, H. (1935). Wahrscheinlichkeitslehrc. Leiden; English tr. The

theory of probability. University o f California Press, Berkeley, 1949.

Rescher, N. (*1969). Many-valued logic. McGraw-Hill, New York.

Rine, D. C. (ed.) (1977). Computer Science and Multiple-valued Logic.

Theory and Aplications. Amsterdam, North-Holland.

Rose, A. and Rosser, J. B. (1958). Fragments o f many-valued statement

calculi. Transactions of the American Mathematical Society , 87, 1-53.

Rosenbloom, P. C. (1942). Post algebra. I. Postulates and general theory.

American Journal of Mathematics, 64, 167-188.

Rosser, J. B. and Turquette, A. R. (*1952). Many-valued logics. North-

Holland, Amsterdam. 1

Rousseau, G. (1969). Logical systems with finitely many truth-values. Bul­

letin de VAcademie Polonaise des Sciences, Serie des sciences mathema-
tiques, astronomiques et physiques, 17, 189-194.

Rutledge, J. D. (1959). A preliminary investigation of the infinitely many­

valued predicate calculus. Ph.D. thesis, Cornell University.

Salwicki, A. (1970). Formalized algorithmic languages. Bulletin de VAcad­

emie Polonaise des Sciences, Serie des sciences mathematiques, astrono­
miques et physiques, 18, 227-232.

Scarpelini, B. (1962). Die Nichtaxiomatisierbarkeit des unendlichwertigen

Pradikatenkalkuls von Lukasiewicz. The Journal of Symbolic Logic, 17,
159-170.

Scott, D. (1973). Background to formalisation [in:] Leblanc, H. (ed.)

Truth, Syntax and Modality. North-Holland, Amsterdam, 244-273.
 Bibliography 121

Scott, D. (1974). Completeness and axiomatizability in many-valued logic

[in:] Henkin, L. et al. (eds) Proceedings of Tarski Symposium. Proceedings
of Symposia in Pure Mathematics, vol. 25, 411-436.

Scott, D. and Solovay, R. (1969). Boolean valued models for set theory, Pro­
ceedings of the American Mathematical Society Summer Inst. Axiomatic
Set Theory 1967. University o f California, Los Angeles. Proceedings of
Symposia in Pure Mathematics, 13.

Skolem, T. (1957). Bemerkungen zum Kornprehensionsaxiom. Zeitschrift

fur Mathematische Logik und Grundlagen der Mathematik, 3, 1-17.

Slupecki, J. (1936). Der voile dreiwertige Aussagenkalkiil. Comptes rendus

des seances de la Societe des Sciences et des Lettres de Varsovie Cl. I l l , 29,
9-11; English tr. The full three-valued propositional calculus [in:] McCall,
S. (ed.) Polish Logic 1920-1939. Clarendon Press, Oxford, 1967, 335-337.

Slupecki, J. (1939a). Kryterium pelnosci wielowartosciowych systemow

logiki zdaii (A criterion o f completeness o f many-valued systems o f propo­
sitional logic). Comptes rendus des seances de la Societe des Sciences et
des Lettres de Varsovie Cl. I l l , 32, 102-109.

Slupecki, J. (1939b). Dowod aksjomatyzowalnosci pelnych systemow wielo­

wartosciowych rachunku zdafi (P roof of the axiomatizability o f full many­
valued systems o f propositional calculus). Comptes rendus des seances de
la Societe des Sciences et des Lettres de Varsovie Cl. I l l , 32, 110-128.

Slupecki, J. (1964). Proba intuicyjnej interpretacji logiki trojwartosciowej

Lukasiewicza (An attempt o f intuitionistic interpretation o f three-valued
Lukasiewicz logic) [in:] Rozprawy Logiczne. PWN, Warszawa.

Suchoh, W . (1974). Definition des founcteurs modaux de Moisil dans le cal-

cul n-valent des propositions de Lukasiewicz avec implication et negation.
Reports on Mathematical Logic, 2, 43-47.

Surma, S. J. (1971). Jaskowski’s matrix criterion for the intuitionistic

propositional calculus. Prace z logiki, V I, 21-54.

Surma, S. J. (1973). A historical survey o f the significant methods o f prov­

ing Post’s theorem about the completeness o f the classical propositional
calculus [in:] Surma, S. J. (ed.) Studies in the History of Mathematical
Logic. Ossolineum, Wroclaw, 19-32.

Suszko, R. (1957). Formalna teoria wartosci logicznych (A formal theory

o f logical values). Studia Logica, V I, 145-320.

Suszko, R. (1972). Abolition o f the Fregean Axiom [in:] Parikh, R. (ed.)

122 Bibliography

Logic Colloquium, Symposium on Logic held at Boston, 1972-73. Lecture

Notes in Mathematics, vol. 453, 169-239.

Snszko, R. (1975). Remarks on Lukasiewicz’s three-valued logic. Bulletin

of the Section of Logic, 4, no. 3, 87-90.

Suszko, R. (1977). The Fregean Axiom and Polish Mathematical Logic in

the 1920’s. Studia Logica, 36, no. 4, 377-380.

Tarski, A. (1930). Uber einige fundament ale Begriffe der Metamathematik.

Comptes Rendus des seances de la Societe des Sciences et des Lettres de
Varsovie CLIII, 23, 22-29; English tr. [in:] Tarski, A. Logic, Semantics,
Metamathematics: Papers from 1923 to 1938, translated by Woodger, J.
H. Clarendon Press, Oxford, 1956, 30-37.

Tarski, A. (1936). O poj^ciu wynikania logicznego (On the concept of

logical consequence). Przeglqd Filozoficzny, 39, 58-68; English tr. [in:]
Tarski, A. Logic, Semantics, Metamathematics: Papers from 1923 to 1938 ,
translated by Woodger, J. H. Clarendon Press, Oxford, 1956, 409-420.

Tarski, A. (1938). Der Aussagenkalkul und die Topologie. Fundamenta

Mathematicae, 31, 103-134; English tr. [in:] Tarski, A. Logic, Semantics,
Metamathematics: Papers from 1923 to 1938, translated by Woodger, J.
H. Clarendon Press, Oxford, 1956, 421-454.

Tokarz, M. (1974). A method o f axiomatization of Lukasiewicz logics.

Bulletin of the Section of Logic, 3, no. 2, 21-24.
I

Traczyk, T. (1964). An equational definition o f a class of Post algebras.

Bulletin de VAcademie Polonaise des Sciences Cl. III , 12, 147-149.

Traczyk, T. (1970). Wstqp do teorii algebr Boole’a (Introduction to the

theory of Boolean algebras). PWN, Warszawa.

Turner, R. (*1984). Logics for Artificial Intelligence. Ellis Horwood,

Chichester.

Urquhart, A. (1973). An interpretation of many-valued logic. Zeitschrift

fur Mathematische Logik und Grundlagen der Mathematik, 19, 111-114.

Urquhart, A. (*1986). Many-valued logic [in:] Gabbay, D. and Guenthner,

F. (eds) Handbook of philosophical logic, vol. III. D. Reidel, Dordrecht,
71-116.

Wade, C.I. (1945). Post algebras and rings. Duke Mathematical Journal,
12, 389-395.
 Bibliography 123

Wajsberg, M. (1931). Aksjomatyzacja trojwartosciowego rachunku zdah.

Comptes Rendus de la Societe des Sciences et des Lettres de Varsovie Cl.
III, 24, 126-148; English tr. Axiomatization of the three-valued proposi­
tional calculus [in:] McCall, S. (ed.) Polish Logic 1920-1939. Clarendon
Press, Oxford, 1967, 264-284.

Wajsberg, M. (1933). Eine erweiterter Klassenkalkiil. Monatshefte fur

Mathematik und Physik, 40, 113-126.

Webb, D. L. (1935). Generation o f any n-valued logic by one binary oper­

ation. Proceedings of the National Academy of Sciences, 21, 252-254.

Whitehead, A. N. and Russell, B. (1910). Principia Mathematica, vol. I.

Cambridge University Press.

Wolehski, J. (1985). Filozoficzna Szkola Lwowsko-Warszawska. PW N,

Warszawa; English tr. Logic and philosophy in the Lvov-Warsaw School.
Synthese Library, 198. D. Reidel, Dordrecht, 1989.

Wolf, R. G. (*1977). A survey of many-valued logic (1966-1974), Appendix

II [in:] Dunn, J. M. and Epstein, G. (eds) Modem uses of multiple-valued
logic. D. Reidel, Dordrecht, 167-324.

Woodruff, P. (1970). Logic and truth value gaps [in:] Lambert, K. (ed.)
Philosophical problems in logic. D. Reidel, Dordrecht, 121-142.

Wojcicki, R. (1970). Some remarks on the consequence operation in sen­

tential logics. Fundamenta Mathematicae, 68, 269-279.

Wojcicki, R. (1977). Strongly finite sentential calculi [in:] W ojcicki, R.

and Malinowski, G. (eds) Selected papers on Lukasiewicz sentential calculi.
Ossolineum, Wroclaw, 53-77.

Wojcicki, R. (1988). Theory of logical calculi. Basic theory of consequence

operations. Synthese Library, 199. Kluwer Academic Publishers, Dor­
drecht.

Zadeh, L. A. (1965). Fuzzy sets. Information and Control, 8, 338-353.

Zadeh, L. A. (1972). A fuzzy-set-theoretic interpretation o f linguistic

hedges. Journal of Cybernetics, 2, 4-34.

Zadeh, L. A. (1975). Fuzzy logic and approximate reasoning. Synthese,

30, 407-428.

Zadeh, L. A. (1976). A fuzzy-algorithmic approach to the definition of

complex or imprecise concepts. International Journal of Man-Machine
124 Bibliography

Studies, 8, 249-291.

Zawirski, Z. (1934a). Znaczenie logiki wielowartosciowej i zwi§zek jej z

rachunkiem prawdopodobienstwa (Significance of rnany-valned logic for
cognition and its connection with the calculus o f probability). Przeglgd
Filozoficzny, 37, 393-398.

Zawirski, Z. (1934b). Stosunek logiki wielowartosciowej do rachunku praw-

dopodobiehstwa (Relation of many-valued logic to the calculus of probabil­
ity). Prace Komisji Filozoficznej Polskiego Towarzystwa Przyjaciol Nauk,
4, 155-240.

Zinov’ev, A. A. (*1963). Philosophical problems of many-valued logic,

edited and translated by Kiing, G. and Comey, D. D. D. Reidel, Dordrecht.
Author index
Ackermann R. 68, 113 Gaines B.R. 100, 115
Aristotle 1 Gardies J.-L . 115
Barcan R.C. 96, 113 Garson J.W . 97, 115
Bellman R.E. 99, 101, 104, 113 Giertz M. 99, 113
Belluce L.P. 81, 113 Giles R. 3, 66, 68, 69, 70, 115
Belnap N.D. 57, 113 Godel K. 12, 46, 89, 92, 115
Bernays P. 10, 116 Goguen J.A. 100, 115
Beth E.W . 90, 113 Gonseth F. 21, 116
Blarney S. 56, 57, 58, 59, 113 Gottwald S. 101, 116
Bloom S.L. 64, 113 Gratzer G. 13, 15, 53, 116
Bochvar D.A. 3, 51, 54, 55, 56, Grigolia R. 40, 42, 43, 116
81, 113 Grzegorczyk A. 90, 95, 116
Boole G. 12 Harel D. 112, 116
Borel E. 87 Hay L.S. 81, 116
Brentano F. 16 Henkin L. 86
Brouwer L.E.J. 2, 87 Henle P. 93
Brown D.J. 64, 113 Hey ting A. 87, 88, 116
Carnap R. 67, 96, 113 Hilbert D. 10, 68, 116
Chang C.C. 12, 40, 41, 43, 81, 82, Hintikka J. 94, 116
114 Jaskowski S. 90, 116
Church A. 12, 114 Johansson I. 89, 116
Cleave J.P. 53, 114 Kabzmski J. 89
Cohen P. 90 Kanger S. 94, 116
Czelakowski J. 29, 114 Kant I. 87
Dalen D. van 96, 114 Keisler H.J. 12, 81, 114
Dugundji J. 92, 114 Kleene S.C. 3, 51, 52, 53, 54, 55,
Dummet M. 90, 114 56, 81, 117
Dunn J.M. 114 Korner S. 52, 117
Duns Scott 2 Kotarbmski T. 12, 17, 117
Dwinger Ph. 48, 114 Krajewski S. 89
Epstein G. 48, 109, 111, 114 Kripke S. 90, 94, 95, 97, 117
Fenstad J.E. 83, 115 Kronecker L. 87
Fitting M.C. 90, 115 Lambert K. 97, 117
Fraassen B.C. van 56, 97, 115, Langford C.H. 91, 93, 117
117 Langholm T. 59, 117
Frege G. 2, 25, 115 Lebesgue H. 87
Frieder G. 109, 111, 114 Leibniz 25, 94
126 Author index

Lemmon E.J. 92, 93, 117 Rousseau G. 49, 120

Lewis C.I. 2, 91, 93, 117 Russell B. 10, 16, 19, 44, 123
Lindenbaum A. 32 Rutledge J.D. 81, 120
Lorenz K. 69, 117 Salwicki A. I l l , 120
Lorenzen P. 90, 117 Scarpelini B. 80, 120
Los J. 107, 108, 118 Scott D. 74, 75, 106, 120, 121
Lukasiewicz J. 2, 3, 10, 16, 17, Sikorski R. 15, 91, 96, 120
18, 19, 20, 21, 22, 23, 36, 37, 38, Skolem T. 82, 121
39, 40, 41, 42, 43, 46, 52, 66, 118 Slupecki J. 21, 22, 23, 27, 38, 39,
MacColl H. 2, 66, 118 48, 49, 121
McKinsey J.C.C. 92, 119 Solovay R.M. 106, 121
McNaughton R. 38, 119 Suchoii W . 41, 121
Malinowski G. 73, 115, 119 Surma S.J. 10, 90, 121
Marciszewski W . 89, 119 Suszko R. 25, 72, 73, 121, 122
Meinong A. 16 Tarski A. 32, 37, 39, 64, 90, 92,
Meredith C.A. 40, 119 93, 118, 119, 122
Moh Shaw-Kwei 81, 119 Tokarz M. 40, 122
Moisil G. 21, 41, 110, 119 Traczyk T. 48, 122
Mostowski A. 80, 81, 119 Turner R. 104, 122
Nowak M. 23, 119 Turquette A .T . 3, 39, 40, 41, 60,
Peirce C.S. 2, 119 62, 64, 79, 83, 85, 86, 120
Peter de Rivo 2 Twardowski K. 16
Picard S. 27, 119 Urquhart A. 77, 122
Poincare H. 87 Vasil’ev N. 2
Post E.L. 2, 3, 8, 10, 26, 27, 44, Wade C.I. 48, 122
46, 47, 48, 119 ' Wajsberg M. 39, 40, 93, 123
Quine W . van O. 97 Webb D.L. 27, 123
Rasiowa H. 15, 49, 50, 91, 111, Whitehead A.N. 10, 44, 123
119, 120 William of Ockham 2
Rautenberg W . 4, 120 Woleiiski J. 16, 123
Reichenbach H. 66, 120 W olf R.G. 4, 123
Reseller N. 4, 68, 79, 120 Woodruff P. 57, 123
Rine D.C. 109, 111, 114, 120 Wojcicki R. 30, 33, 34, 123
Rose A. 40, 120 Zadeli L.A. 3, 98, 99, 101, 102,
Rosenbloom P.C. 48, 120 103, 104, 113, 123
Rosser J.B. 3, 39, 40, 41, 60, 62, Zawirski Z. 17, 66, 123
64, 79, 83, 85, 86, 120 Zinov’ev A.A. 124
Subject index
algebra 8, 24, 28, 30, 40, 101 base of consequence 34
absolutely free 25 bound 13
abstract 24, 93 greatest lower (inf) 13, 70, 80,
Boolean 13, 15, 22, 40, 42, 47, 100
93, 109 least upper (sup) 13, 70, 80
fuzzy sets 99, 103
Lindenbaum 15, 53 casual event 66
Lukasiewicz 27, 41 class
many-valued 26, 47, 109 inexact 52, 98
Moisil 21, 41, 105, 109, 110 of matrices (adequate) 73
o f formulae 24 closure 93
Post 26, 27, 47, 105, 109, 111 complement 13, 14, 21, 47, 53, 93,
n-valued 26, 44 99
o f order n 48,111 completeness 10, 12, 40, 43, 49,
of order ω+ 111 63, 93, 94, 96
pseudo-Boolean (Heyting) 49, computer science 4, 101, 105, 109,
91, 96 111
quotient 29 concept 98
Rousseau 49 imprecise 98
S-algebra 93 confirmation function 67
assertion 55 congruence 14, 29
conditional 56, 57, 76 compatible with a consequence
external 55 matrix 33
axiom 2, 9, 12, 34, 40, 50, 80, 88, conjunction 5, 6, 9, 18, 36, 53, 79
90, 92, 96, 106, 107 external 54
of choice 39 connectives 5, 18, 24, 26, 40, 52,
of comprehension 16, 81, 83 57, 61, 64, 99, 103, 108
axiomatics 9, 12, 39, 48, 60, 62, external 54
79, 85, 105 internal 54
adequate 10, 86 modal 20, 41, 91
complete 10, 12, 40, 62, 80, 86 necessity 91
independent 105 possibility 91
sound 10, 62, 88 strong 53, 58
128 Subject index

weak 53 primitive 45
consequence 28, 30, 32, 33, 34, rotation 26, 44
59, 63, 76 truth function 8
absolutely separable 33 functional completeness 2, 8, 24,
absolutely uniform 33 26, 38, 45, 47
classical 35, 64 for n variables 26
finite 34, 63 functor 1, 60
strongly 33, 35 modal 20, 91, 93
structural 32, 35, 73 of necessity 20, 91
uniform 33 of possibility 20, 91
content of matrix 28, 30, 71, 92 future contingents 1, 17
CPC 1, 5, 9, 28, 60, 88, 92
game theory 70
deduction theorem 34, 64
degree o f error 76 hedge 101, 102
denotation 25, 97 Hom (C,A) 25, 28, 72
derivation 9
designated element 28, 72, 93 implication 5, 6, 18, 23, 34, 36,
dialogue 69, 90 46, 49, 78, 91
disjunction 5, 6, 9, 18, 26, 36, 44, external 54
53, 79 Godel 46, 100
exclusive 8 Lukasiewicz 36, 46, 77, 100
external 54 strict 91
induction 17, 67
endomorphism 26, 33, 41 INT 88
epimorpliism 61 1 intensionality 2, 68, 105, 106
equivalence 5, 6, 9, 18, 36 interior 93
external 54 interjunction 58
strict 91 interpretation 11, 20, 26, 46, 73,
extensionality 8, 25 81, 91, 97, 107
McKinsey-Tarski 92
falsity 1, 6, 17, 25, 45, 56, 60 topological 90
field o f sets 14, 48 intuitionism 3, 87
For 5, 6, 11, 18, 24, 28, 56 involution 41
formula 5, 24, 62, 74, 83 isomorphism 14, 37, 43, 46, 93
Barcan 96
contingent 56 l f n-truth 77
contradictory 19 Kripke frame 94
function 6, 13, 18, 25, 36, 61, 72, S4-, S5- 94
107
characteristic 13, 40, 47, 52, language 7, 21, 28, 33, 98, 107
73, 98, 99, 102 propositional 7, 24, 25, 28, 72,
constant 13 74
maximum 26 lattice 13, 14
 Subject index 129

distributive 13, 41, 47 logical values 1, 13, 17, 22, 60, 74,
de Morgan 23, 53, 99 99, 101, 108
law 7, 45, 89, 95 designated 18, 30, 60, 74
o f double negation 7, 96 falsity 1, 17, 108
of the excluded middle 7, 19, linguistic 101, 102
20, 45, 59, 89, 96 local 103
o f transitivity 77 truth 1, 17, 108
Lindenbaum bundle 33 undefinite 17, 51
linguistic approximation 101, 103
localization 103 many-valuedness 2, 3, 17, 30, 32,
logic 1 66, 72, 108
algorithmic 111 in computer science 111
Bochvar 51, 54, 77, 80 matrix 2, 24, 28, 30, 61, 72, 87,
external 55 92, 105
internal 55 Bochvar 54, 55
classical 1, 2, 5, 35, 44, 89 congruence 29
dialogue 69, 90 consequence 28, 30, 32, 40, 53,
dynamic 112 60, 63
free 95, 97 Godel 46, 89, 90
fuzzy 3, 98, 101, 103 Heyting 88
intermediate 89 Jaskowski 90
intuitionistic 2, 87, 92, 96 Kleene 52
Kleene 51, 57 Lindenbaum 32
local 103 Lukasiewicz 3, 36, 38, 41, 61,
Lukasiewicz 3, 17, 36 75
No-valued 40, 43, 68, 80, of classical logic (M 2) 28, 37,
100 60, 89
n-valued 27, 40, 74 Post 44, 61
many-valued 2, 3, 19, 30, 36, n-valued 45
60, 64, 72 implicative 46
minimal 89 Q-standard 61
modal 2, 3, 21, 87, 91, 96, 112 quotient 29
S4, S5 91, 93 Slupecki 49
S 1 - S 5 91 standard 61, 64
partial 3, 51, 56 T-standard 61, 63
Post 44, 48, 77, 86 meet (intersection) 1 3 ,2 1 ,2 9 ,5 3 ,
propositional 1, 5, 27, 34, 73, 93, 99
88 modality mode 96
Rosser-Turquette 60, 62 de dido 96
three-valued 2, 16, 17, 21, 31, de re 96
36, 39, 51, 54, 73 modus ponens 7
two-valued 1, 5, 30, 74 modus tollens 7
Zadeh 98, 101 monotonicity 57, 58, 98
logical valuation 6, 58, 72 MV algebra 41, 42, 81
130 Subject index

M Vn algebra 42 reductio ad absurdum 7

relation
negation 5, 6, 18, 26, 78 partial ordering 13
cyclic 26, 44 representation theorem 43
external 54 risk value 69
normal form 10, 85 rule 7, 33
partial 84 absorption 82
prenex 85 axiomatic 34
compositional rule o f inference
operation 103
closure 93 detachment = modus ponens
interior 90, 93 (M P) 7, 9, 34, 39, 64, 88, 107
extensionality 91
predicate 11, 51, 79, 83 generalization 12, 86, 107
inexact 53, 98, 100 Godel’s (RG) 92, 94
predicate calculus 1, 11, 15, 79 of adjunction 91
standard 85 of debate 69
principle 69 of inference 7, 9, 12, 33, 56,
extensionality 25, 60, 67 96, 101
of bivalence 1, 7, 17 of obligation 68
of causality 17 substitution (SUB) 9, 34, 39,
of contradiction 7, 16, 19, 20 88, 107
probability 3, 17, 66, 100 structural 34
logical 36, 66 Russell’s
subjective 3, 66, 68, 100 paradox (antinomy) 16, 19, 81
probability measure 67 set 19
product of matrices 29, 56
proof 35, 37, 38, 87, 90, 92, 96 S-algebra 93
proposition 1, 16, 21, 28, 46, 96, S4-, S5-algebra 93, 94
108 scope of quantifier 11
meaningful 54 semantic correlate 25, 28, 72, 109
meaningless 54 semantics 77, 87, 90, 93
prime 68, 69 forcing 90
undefinite 17 Kripke-Grzegorczyk 90, 94,
proposition types 74 95, 97
propositional calculus 1, 24, 29 sequent 33
classical 1, 5, 24, 64 set 3, 12, 81
intuitionistic 87, 92 empty 12
fuzzy 98, 100, 103
quantifier 1, 3, 11, 20, 79, 84, 93, n-valued 47, 98
95 of formulae R-closed 34
generalized 79, 83 power 13
fuzzy 104 Russell 19, 81
 Subject index 131

universal 12 truth-value gap 56, 57

situation 73, 74 two-valuedness 72
standard conditions 60, 61 logical 72
strategy 70
optimal 70 union 13, 21, 53, 93, 99
subalgebra 14, 29
submatrix 29, 37
world 94, 97
substitution 9, 25, 32
nomnormal 95
supervaluation 3, 51, 56, 59
possible 94, 97
induced by K 56
real 94
induced by X 56
Suszko thesis 72
switching theory 4, 105, 109 validity 1, 7, 87
valuation 6, 7, 11, 18, 31, 37, 56,
TAUT 7, 9, 28, 61 59, 74
tautology 7, 9, 14, 18, 37, 45, 52, in Kripke frame 94
70, 88 logical 6, 11, 72
theory of models 12, 29 partial 58, 68
continuous 81 Var 6 , 2 4 ,5 6
partial 59 variable 5, 62
transplication 58 nominal 11, 79, 83, 107
truth 1, 6, 17, 25, 45, 56, 60 bound 11, 82
truth-tables 1, 5, 18, 26, 44, 52, free 11, 17, 82
55, 89, 106, 108 propositional 5, 24, 107