Logic, Epistemology, and the Unity of Science 40

WalterCarnielli
MarceloEstebanConiglio

Paraconsistent
Logic:
Consistency,
Contradiction
and Negation
Logic, Epistemology, and the Unity of Science

Volume 40

Series editors
Shahid Rahman, University of Lille III, France
John Symons, University of Texas at El Paso, USA

Editorial Board
Jean Paul van Bendegem, Free University of Brussels, Belgium
Johan van Benthem, University of Amsterdam, The Netherlands
Jacques Dubucs, CNRS/Paris IV, France
Anne Fagot-Largeault, Collge de France, France
Gran Sundholm, Universiteit Leiden, The Netherlands
Bas van Fraassen, Princeton University, USA
Dov Gabbay, Kings College London, UK
Jaakko Hintikka, Boston University, USA
Karel Lambert, University of California, Irvine, USA
Graham Priest, University of Melbourne, Australia
Gabriel Sandu, University of Helsinki, Finland
Heinrich Wansing, Ruhr-University Bochum, Germany
Timothy Williamson, Oxford University, UK
Logic, Epistemology, and the Unity of Science aims to reconsider the question
of the unity of science in light of recent developments in logic. At present, no single
logical, semantical or methodological framework dominates the philosophy of
science. However, the editors of this series believe that formal techniques like, for
example, independence friendly logic, dialogical logics, multimodal logics, game
theoretic semantics and linear logics, have the potential to cast new light on basic
issues in the discussion of the unity of science.
This series provides a venue where philosophers and logicians can apply specic
technical insights to fundamental philosophical problems. While the series is open
to a wide variety of perspectives, including the study and analysis of argumentation
and the critical discussion of the relationship between logic and the philosophy of
science, the aim is to provide an integrated picture of the scientic enterprise in all
its diversity.

More information about this series at http://www.springer.com/series/6936

Walter Carnielli Marcelo Esteban Coniglio

Paraconsistent Logic:
Consistency, Contradiction
and Negation

123
Walter Carnielli Marcelo Esteban Coniglio
Department of Philosophy and Centre Department of Philosophy and Centre
for Logic, Epistemology and the History for Logic, Epistemology and the History
of Science (CLE) of Science (CLE)
University of Campinas (UNICAMP) University of Campinas (UNICAMP)
Campinas, So Paulo Campinas, So Paulo
Brazil Brazil

ISSN 2214-9775 ISSN 2214-9783 (electronic)

Logic, Epistemology, and the Unity of Science
ISBN 978-3-319-33203-1 ISBN 978-3-319-33205-5 (eBook)
DOI 10.1007/978-3-319-33205-5

Library of Congress Control Number: 2016936981

Springer International Publishing Switzerland 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microlms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specic statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG Switzerland
To absent friends, and to Elias Alves, in
memoriam. To our children: B, Juju,
Matheus, Paolo, Gabriela, Vittorio and to
the kids we would have had, and to Juli and
Tati. Sine qua non.

Campinas, February 29, 2016

Preface

I protest against the use of innite magnitude as something completed, which in mathe-
matics is never permissible. Innity is merely a faon de parler, the real meaning being a
limit which certain ratios approach indenitely near, while others are permitted to increase
without restriction.

C.F. Gauss, Brief an Schumacher (1831); Werke 8, 216 (1831)

In a letter to the astronomer H.C. Schumacher in 1831, Gauss was rebuking

mathematicians for their use of the innite as a number, and even for their use of the
symbol for the innite. It would be difcult to sustain that kind of nitism,
regardless of any epistemological considerations: a good part of mathematics
simply cannot survive with only the potential innite.
The reaction against the innite, as well as against complex or imaginary
numbers, and against negative numbers before, are interesting examples of the
difculties faced, even by great minds, in accepting certain abstractions. Aristotle in
Chaps. 48 of Book III of Physics argued against the actual innite, advocating for
the potential innite. His idea was that natural numbers could never be conceived as
a whole.
Euclid in a certain sense never proved that there exist innitely many prime
numbers. What was actually stated in Proposition 20 of Book IX, carefully avoiding
the term innite, was that prime numbers are more than any previously thought
(total) number of primes, which agrees with his tradition.
It was only in the nineteenth century that G. Cantor dispelled all those accepted
views by showing that an innite set can be treated as a totality, as a full-fledged
mathematical object with honorable properties, no less than the natural numbers.
Imaginary numbers were introduced to mathematics in the sixteenth century
(through Girolamo Cardano, though others had already used them in different
guises). These numbers caused an embarrassment among mathematicians for cen-
turies, since they faced astonishing difculties in accepting an extension of the
concept of number, especially in light of the problem of computing the square root
of 1. Only after the fundamental works of L. Euler and Gauss did the complex

vii
viii Preface

numbers rid themselves of the label imaginary given them by Descartes in 1637,
and even then not without difculty.
The mathematics of the innite and of complex numbers, and all they represent
in contemporary science, are triumphant cases of amplied concepts, but are not the
only ones. A notable case of expansion of concepts, with deep implications for the
development of contemporary logic, can be traced back to Frege and his famous
article of 1891, Funktion und Begriff (see [1]).
In this seminal paper, Frege recalls how the meaning of the term function has
changed in the history of mathematics, and how the mathematical operations used
to dene functions have been extended by, as he says, the progress of science:
basically, through passages (or transitions) to the limit, as in the process of dening
a new function y0 f 0 x from a function y f x (provided that the limits
involved in the calculus exist), and through accepting complex numbers in domains
and images of functions.
Starting from this point, Frege goes further into adding expressions that now we
call predicates, such as =, < and >. Leaving aside his philosophical motivations
for seeing arithmetic as a further development of logic, what Frege started was a
real revolution, that made possible the development of quantiers and an
unprecedented unication of propositional and predicate logic into a far more
powerful system than any that preceded it.
Not only could the truth-values, True and False, be taken as outputs of a
function, but any object whatsoever could be similarly treated. To rephrase an
example from Frege himself, if we suppose the capital of x expresses a function,
of which the German Empire is the argument, Berlin is returned as the value of the
function. In this way, Freges system could represent non-mathematical thoughts
and predications, and founded the basis of the modern predicate calculus.
Freges idea of dening an independent notion of concept as a function which
maps every argument to one of the truth-values True or False was instrumental in
the development of a strict understanding of the notions of proof, derivation, and
semantics as parts of the same logic mechanism. Regarding concept as a wide
and independent notion based on an amplication of the idea of function was an
essential step for Freges fundamental break between the older Aristotelian tradition
and the contemporary approach to logic.
Paraconsistency is the study of logical systems in which the presence of a
contradiction does not imply triviality, that is, logical systems with a non-explosive
negation : such that a pair of propositions A and :A does not (always) trivialize the
system. However, it is not only the syntactic and semantic properties of these
systems that are worth studying. Some questions arise that are perennial philo-
sophical problems. The question of the nature of the contradictions allowed by
paraconsistent logics has been a focus of debate on the philosophical signicance
of paraconsistency. Although this book is primarily focused on the
logico-mathematical development of paraconsistency, the technical results
emphasized here aim to help, and hopefully to guide, the study of some of those
philosophical problems.
Preface ix

Paraconsistent logics are able to deal with contradictory scenarios, avoiding

triviality by means of the rejection of the Principle of Explosion, in the sense that
these theories do not trivialize in the presence of (at least some) contradictory
sentences. Different from traditional logic, in paraconsistent logics triviality is not
necessarily connected to contradictoriness; in intuitive terms (a more formal
account in given in Sect. 1.2) the situation could be described by the pictorial
equation:

contradictoriness consistency triviality

The Logics of Formal Inconsistency, from now on LFIs, introduced in [2] and
additionally developed in [3], are a family of paraconsistent logics that encom-
passes a great number of paraconsistent systems, including the majority of systems
developed within the Brazilian tradition. An important characteristic of LFIs is that
they are endowed with linguistic resources that permit to express the notion of
consistency of sentences inside the object language by using a sentential unary
connective referred to as circle: A meaning A is consistent. Explosion in the
presence of contradictions does not hold in LFIs, as much as in any other para-
consistent logic. But LFIs are so designed that some contradictions will cause
deductive explosion: consistent contradictions lead to trivialityintuitively, one can
understand the notion of a consistent contradiction as a contradiction involving
well-established facts, or involving propositions that have conclusive favorable
evidence. In this sense, LFIs are logics that permit one to separate the sentences for
which explosion hold, from those for which explosion does not hold. It is not
difcult to see that, in this way, reasoning under LFIs extend and expand the
reasoning under classical logic: although LFIs are technically subsystems of
classical logic, classical logic can be identied with that portion of LFIs that deals
with consistent contradictions. Therefore LFIs subsume classical reasoning. This
point will be explained in more detail in Sect. 1.2.
We may say that a rst step in paraconsistency is the distinction between triv-
iality and contradictoriness. But there is a second step, namely, the distinction
between consistency and non-contradictoriness. In LFIs the consistency connective
 is not only primitive, but it is also not necessarily equivalent to non-contradiction.
This is the most distinguishing feature of the logics of formal inconsistency.
Once we break up the equivalence between A and :A ^ :A, some quite inter-
esting developments become available. Indeed, A may express notions of con-
sistency independent from freedom from contradiction.
The most important conceptual distinction between LFIs and traditional logic is
that LFIs start from the principle that assertions about the world can be divided into
two categories: consistent sentences and non-consistent sentences. Consistent
propositions are subjected to classical logic, and consequently a theory T that
contains a pair of contradictory sentences A; :A explodes only if A is taken to be a
consistent sentence, linguistically marked as A (or :A). This is the only dis-
tinction between LFIs and classical logic, albeit with far-reaching consequences:
x Preface

classical logic in this form is augmented, in such a way that in most cases an LFI
encodes classical logic.
The concept of LFIs generalizes and extends the famous hierarchy of C-systems
introduced in [4] and popularized by hundreds of papers. At the same time, LFIs
expand the classical logical stance, and consequently the majority of the traditional
concepts and methods of classical logic, propositional or quantied (and even
higher-order), can be adapted, with careful attention to detail.
Since, as much as intuitionistic logic, LFIs are more of an epistemic nature,
rather than of an ontological, there is no point in advocating the replacement of
classical logic with paraconsistent logic. Because LFIs extend the classical stance,
the analogy with transnite ordinal numbers and with complex numbers is com-
pelling: in such cases, there is no rejection of what has come before, but a
renement of it.
It is not infrequent that an argument as of the skeptics, such as that given by
Sextus Empiricus1 against the sophists, is trumpeted against the need of paracon-
sistent logic, in science or reasoning in general:
[If an argument] leads to what is inadmissible, it is not we that ought to yield hasty assent to
the absurdity because of its plausibility, but it is they that ought to abstain from the
argument which constrains them to assent to absurdities, if they really choose to seek truth,
as they profess, rather than drivel like children. Thus, suppose there were a road leading up
to a chasm, we do not push ourselves into the chasm just because there is a road leading to
it but we avoid the road because of the chasm; so, in the same way, if there should be an
argument which leads us to a confessedly absurd conclusion, we shall not assent to the
absurdity just because of the argument but avoid the argument because of the absurdity. So
whenever such an argument is propounded to us we shall suspend judgement regarding
each premiss, and when nally the whole argument is propounded we shall draw what
conclusions we approve.

This argument, however, if it is not against the use of any logic, is indeed
favorable to the kind of paraconsistency represented by LFIs. The notion of con-
sistencysymbolized as  when applied to propositionsactually increases our
wisdom: it does not stop one to jump into the chasm, but rather marks out the
dangerous roads and, precisely, helps avoid such roads because of the chasm!
The idea that consistency can be taken as a primitive, independent notion, and be
axiomatized for the good prot of logic is a new idea, which permits one to separate
not only the notion of contradiction from the notion of deductive triviality, which is
true of all paraconsistent logics, but also the notion of inconsistency from the notion
of contradictionas well as consistency from non-contradiction. This rened idea
of consistency has great potential, as we shall see in detail in this book, as unan-
ticipated as the possibilities that imaginary numbers, completed innite, and Freges
idealization of a concept as a function mapping arguments to one of the
truth-values represented in mathematics, logic and philosophy. The rest of the book
will speak for itself.

1
Sextus Empiricus, Outlines of Pyrrhonism, LCL 273: 318319. http://www.loebclassics.com/
view/sextus_empiricus-outlines_pyrrhonism/1933/pb_LCL273.3.xml
Preface xi

Books Content: A Road Map

Chapter 1
Chapter 1 purports to clarify the whole project behind LFIs, making sense of its
idealization and basic tenets. The paraconsistent viewpointmaterialized by means
of LFIsobjects to classical logic, but only on the grounds that contradiction and
triviality cease to coincide, and that contradiction ceases to coincide with incon-
sistency. But this requires no opposition to the classical stance, just the awareness
that classical logic involves some hidden assumptions, as discussed above in this
chapter. In the light of this, Sect. 1.2 makes explicit some of the philosophical
underpinning implicit in LFIs.
Section 1.3 will briefly retrace the motivations for the forerunners of LFIs and
paraconsistency in general. No discussion of paraconsistency can avoid touching
on, if only summarily, questions of the nature of logic, and Sect. 1.4 does this. Next
challenges to be faced are questions about the nature of contradictions. Section 1.5
takes up this thorny philosophical topic from the times of the ancient Greece,
cursorily discussing some remarks from Aristotle concerning three alleged versions
of the Principle of Non-Contradiction that correspond to the three traditional
aspects of logic, namely, ontological, epistemological, and linguistic.
This stance helps to give a justication for the rational acceptance of contra-
dictory sentences, and to better appreciate the distinctions among contradiction,
consistency, and negation, as characterized in Sect. 1.6. It will also help to make
palatable the rationale behind the semantics of LFIs to be developed in all math-
ematical details in Chaps. 2 and 3, as well as to give support to alternative semantics
for LFIs developed in Chap. 6.
There is a wide variety of reasons for repudiating (or at least to be cautioned
against) classical logic, and many of them nd an expression among paraconsistent
logics. This chapter makes clear that LFIs are not coincidental with this spectrum of
philosophical views, neither are they antagonistic, but can be combined with, and
can complement, some of them. A summary of the main varieties of paraconsis-
tency is given in Sect. 1.7, which attempts to clarify the position of LFIs with
respect to other paraconsistent logics in the hope that this will justify some claims
made in next chapters.
Chapter 2
Chapter 2 offers a careful survey of the basic logic of formal inconsistency, mbC:
it is basic in the sense that, starting with positive classical logic CPL+ and adding a
negation and a consistency operator, it is endowed with minimal properties in order
to satisfy the denition of LFIs. The chapter also lays out the main notation,
ongoing denitions and main ideas that will be used throughout the book. Positive
classical logic is assumed as a natural starting point from which the LFIs will be
dened, although in Chap. 5 some LFIs will be studied starting from other logics
than CPL+. A non-truth-functional valuation semantics for mbC is dened in Sect.
2.2, and its meaning and consequences explored in Sect. 2.3.
xii Preface

A remarkable feature of LFIs in general, and of mbC in particular, as mentioned

above, is that classical logic (CPL) can be codied, or recovered, inside such
logics, as shown, for instance, in Sect. 2.4.
One of the criteria proposed by da Costa in [5], p. 498, is that a paraconsistent
calculus must contain as many of the schemata and rules of classical logic as can be
endorsed without validating of the laws of explosion and non-contradiction. This
vague criterion can be formalized in the sense that some LFIs can be proved to be
maximal with respect to CPL, as in the case, for instance, of some three-valued
LFIs treated in Chap. 4.
Moreover, in addition to being a subsystem of CPL, mbC is also an extension of
CPL, obtained by adding to the latter a consistency operator  and a paraconsistent
negation : (see Sect. 2.5). In this sense, mbC can be viewed, both, as a subsystem
and as a conservative extension of CPL. A similar phenomenon holds for several
other LFIs.
That section also sheds light on how CPL can be codied in mbC, showing that
this can be achieved by way of a conservative translation, or by establishing a
Derivability Adjustment Theorem (or DAT) between CPL and mbC. Section 2.5
also discusses an alternative formulation for mbC called mbC? , showing that by
means of linguistic adaptations mbC can be directly introduced as an extension of
CPL.
Chapter 3
Chapter 3 deals with extensions of mbC, which by its turn is a minimal extension
of CPL+ with a consistency operator  and a paraconsistent negation : charac-
terizable as an LFI. This chapter denes several extensions of mbC, strengthening
or expanding different characteristics of this basic system.
In mbC, however, negation and consistency are totally separated concepts. The
rst extension of mbC, called mbCciw, is dened as the minimal extension
guaranteeing that the truth-values of and : completely determine the truth-value
of .
Besides being a subsystem of classical logic, mbC is strong enough to contain
the germ of classical negation, possessing a kind of hidden classical negation, as
explained in Sect. 2.4 of Chap. 2. Section 3.2 of this chapter shows that here is
another hidden operator in mbC: an alternative consistency operator  , one for
each formula . This operator establishes an important distinction, from a con-
ceptual point of view, between mbC and mbC ? as claried in Sect. 3.4.
When he introduced his famous hierarchy Cn (1  n\ of paraconsistent
systems, da Costa dened, for each system Cn , a kind of well-behavedness
operator (later identied with consistency) in terms of the paraconsistent negation
and conjunction (see Sect. 3.7). A special type of LFIs called dC-systems, char-
acterized by the fact that the consistency operator can be dened in terms of the
others, has been dened in [2]. The systems Cn of da Costa turn out to be examples
of dC-systems. Section 3.3 of this chapter analyzes the formal notion of
dC-systems, and investigates how to expand mbC in order to dene the consistency
Preface xiii

and/or the inconsistency operator in terms of the other connectives of the given
signature.
In general terms, LFIs are concerned with the notion of consistency, expressed
by the operator . The notion of inconsistency of is usually dened via the new
operator :, expressing the (formal) inconsistency of . Section 3.5 studies the
balance (or better, unbalance) between the formal concepts of consistency and
inconsistency, dening a new LFI (mbC, which, in fact, is a dC-system) where
inconsistency is a primitive notion and consistency is a dened one.
A natural requirement when characterizing consistency, as much as negation, is
how consistency can be propagated through the remaining connectives. Sections
3.6 and 3.8 analyze extensions of mbC enjoying propagation of consistency in
different forms, in the spirit of the historical systems of da Costa.
Chapter 4
Chapter 4 deals with matrices and algebraizability, and their consequences. In
particular, the question of characterizability by nite matrices, as well as the
algebraizability of (extensions of) mbC is tackled. Some negative results, in the
style of the famous Dugundjis theorem for modal logics, are shown for several
extensions of mbC. This results in new, compact proofs of previously established
results, to the effect that a wide variety of LFIs extending mbC cannot be
semantically characterized by nite matrices. Despite these general results, some
three-valued extensions of LFIs can be characterized by nite matrices, and most
of them are algebraizable in the well-known sense of Blok and Pigozzi. This is
surprising, considering that several extensions of mbC, including the systems Cn of
da Costa, cannot be algebraizable in Blok and Pigozzis sense (and consequently,
not in Lindenbaum and Tarskis sense).
On the topic of LFIs that can be dened matricially, the chapter also covers
Halldns. logic of nonsense as well as Segerbergs variation, da Costa and
DOttavianos, logic J3, also known in its variants LFI1 and MPT, Settes logic
P1, Priests logic LP, the system Ciore, and several other related systems.
Chapter 5
Chapter 5 is devoted to giving an account of LFIs based on other logics, distinct
from what was done in previous chapters, in which LFIs based exclusively on
positive classical logic CPL+ were studied. Although several extensions of the
basic system mbC have been proposed, including several three-valued logics (some
of them even algebraizable in the sense of Blok and Pigozzi, which is not possible
in the case of mbC) the underlying basis was always CPL+. This chapter, instead,
analyzes LFIs dened over other logical basis, to wit: positive intuitionistic logic,
the four-valued Belnap and Dunns logic BD, some families of fuzzy logics, and
some positive modal logics.
Section 5.1 starts by dening LFIs based on positive intuitionistic logic, instead
of CPL+, beginning with paraconsistent logics based on IPL+ (taking as a basis
Johanssons minimal logic and Nelsons logic). A weaker version of mbC called
xiv Preface

imbC obtained from the former by changing the positive basis CPL+ to IPL+ is
also investigated,
Section 5.2 is dedicated to the task of combining two paradigms of uncertainty:
fuzziness and paraconsistency, with exciting possibilities. Taking as a basis the
monoidal t-norm based logic MTL introduced in [6] as a generalization of the
famous basic fuzzy logic BL due to P. Hjek (which, in turn, simultaneously
generalizes three chief fuzzy logics, namely ukasiewicz, Gdel-Dummet and
Product logics) several new LFIs had been recently developed (see [7]).
Justied by the fact that MTL is the most general residuated fuzzy logic whose
semantics is based on t-norms, the LFIs dened over MTL give a nely controlled
combination of fuzzy and consistency (as well inconsistency) operators, giving rise
to mathematical models for the novel notion of fuzzy (in)consistency operators,
which formalizes the nice and natural idea of degrees of consistency and
inconsistency.
Section 5.3 investigates a four-valued modal LFI based on N. Belnap and J.M.
Dunns logic BD, a logic (based on their famous bilattice logic F OUR) suitable for
representing lack of information (a sentence is neither true nor false) or excess of
information (a sentence is both true and false). The logic BD was dened from the
notion of proposition surrogates introduced by J.M. Dunn about ve decades ago
as a set-theoretic tool for representing De Morgan Lattices. The logic M4m , a matrix
logic expanding Belnap and Dunns logic BD by adding a modal operator, is then
dened and proved to be an LFI. Moreover, it is a dC-system based on the logic
preserving degrees of truth of the variety of bounded distributive lattices. The logic
M4m is based on the previous work by A. Monteiro on tetravalent modal algebras.
The chapter closes, in Sect. 5.4, with an overview of the notion of modal LFIs
and their unfoldings.
Chapter 6
Chapter 6 studies alternative semantics for the LFIs presented in Chaps. 2 and 3,
concentrating on the novel notion of swap structures. As much as modal logics,
LFIs are in general non-truth-functional, and (as much as modal logics) have access
to different kinds of semantics (like algebraic semantics, Kripke or relational
semantics, topological semantics, and neighborhood semantics, among others) to
better clarify their meaning, LFIs also naturally require a plurality of semantics. But
unlike modal logics, LFIs in general do not have non-trivial logical congruences,
and the question of dening other semantics for LFIs becomes more sensible.
Standard tools, like categorial or algebraic semantics, will not work so easily for
LFIs and the development of alternative semantical techniques for certain LFIs is
an ongoing and relevant task.
The chapter claries the heritage of swap structures from M. Fidels notion of
twist structures (studied in Chap. 5), and also discusses the close relationship
between the concept of Fidel structures, swap structures and non-deterministic
matrices (or Nmatrices).
Section 6.8 surveys the possible-translations semantics (PTSs), a broad
semantical concept introduced in 1990 that gives new philosophical interpretations
Preface xv

for some non-classical logics, and especially for paraconsistent logics. It happens
that PTSs is a very general semantical notion, to the point that virtually any logic
may have a PTS interpretation, under certain conditions. It also happens that
several other semantical notions can be seen as particular cases of PTSs; those
points are carefully explained in that section.
Chapter 7
Chapter 7 gives a full account of LFIs for rst-order languages. The quantied
versions of LFIs are essential for certain mathematical applications, such as set
theory, and also for concrete applications in computer science, such as databases
and logic programming. The combination of the consistency operator  with
quantiers 8 and 9 demands a careful treatment: now, the propagation of consis-
tency through quantiers has to be duly balanced, generalizing from the propaga-
tion of consistency for conjunction and disjunction. The intuitive idea, of course, is
to regard the existential quantiers as arbitrary conjunctions and disjunctions, but
this has to be done taking a certain technical care.
The chapter is structured around a complete treatment of the system QmbC, a
quantied exension of the system mbC, the basic LFI studied in Chap. 2. Other
extensions of QmbC, such as QCi and QmbC  (the latter including an equality
predicate), are also treated, keeping QmbC at the horizon. From the point of view
of semantics, Tarskian rst-order structures are now endowed with a paraconsistent
bivaluation, and what results is a wide generalization of familiar model theory. An
alternative approach to three-valued rst-order LFIs is developed in detail in Sect.
7.9, based on the theory of quasi-truth. This treatment, of course, can be extended to
other many-valued paraconsistent logics.
The paradigm of quasi-truth, which provides a way of accommodating the
conceptual incompleteness inherent in scientic theories as studied in [8], views
scientic theories from the perspective of paraconsistent logic. This paradigm offers
a rational account for the dynamics of theory change, allowing for theories
involving contradictions without triviality, with deep implications for the founda-
tions of science and for the understanding of the scientic method. A generalization
of the logical aspects of the theory of quasi-truth has been undertaken in [9], by
means of a three-valued model theory for an LFI called LPT1, which in turn
coincides (setting aside some details of language) with the quantied version of the
three-valued paraconsistent logic LFI1 introduced in Denition 4.4.41. An addi-
tional discusion on quasi-truth can be found in Sect. 9.3 of Chap. 9.
One of the aims of this chapter is to endorse the claim that basically the same
results of classical model theory hold for QmbC, and for rst-order LFIs in gen-
eral, with certain provisos. Well-established results in traditional model theory such
as the Completeness, Compactness and LowenhimSkolem Theorems can be
proved for rst-order LFIs along the same lines as the classical case. In this way,
the chapter makes clear that rst-order LFIs expand traditional logic, and allows for
a revision of the uses of logic in mathematics and computer science from the
vantage point of richer logics.
xvi Preface

Chapter 8
The confusion between the concept of set on the one hand, and of class, or species,
on the other hand, has plagued the foundations of set theory since its birth. The
Principle of Comprehension (also referred to as the Principle of Nave
Comprehension, or Abstraction) was proposed in the nineteenth century, fruit of the
somewhat romantic ideas of Dedekind, Cantor, and Frege, and states that for every
property, expressed as a predicate, there exists a set consisting of exactly those
objects that satisfy the predicate. This principle lurks behind certain tough para-
doxes, such as Russells paradox, and the history of contemporary set theory has
much to do with efforts to rescue Cantors nave set theory from triviality, an
inevitable consequence, in traditional logic, of the contradictions entailed by those
paradoxes. Paraconsistent set theory has been an endeavor to save set theory from
certain (it not all) paradoxes for at least three decades. Chapter 8 aims to offer a new
approach to this question by means of employing LFIs and their powerful con-
sistency operator. By assuming that not only sentences, but sets themselves can be
classied as consistent or inconsistent objects, the basis for new paraconsistent
set-theories that can resist certain paradoxes without falling into trivialism is
established. One of the main motivations of this chapter, as stated in Sect. 8.1, is to
rescue, together with Cantors nave set theory, the proper Cantors intuition
towards inconsistent sets. Indeed, the chapter attempts to show that Cantors
treatment of inconsistent collections can be related to the one provided by means of
LFIs.
Section 8.2 denes ZFmbC, a basic system of paraconsistent set theory whose
underlying logic is QmbC  , and which contains two non-logical predicates (be-
sides the equality predicate ): the binary predicate 2 (for membership), and the
unary predicate C (for consistency of sets). Section 8.3 proposes some extensions of
ZFmbC by means of employing stronger LFIs as underlying logics and setting
appropriate axioms for the consistency operator C for sets. Section 8.4 discusses the
relationship between the notions of to be a consistent object in set theory (as
formalized in the chapter) and to be a set. It shows that consistent objects can be
(without risk of trivialism) regarded as sets, by means of an appropriate axiom. In
the same spirit, proper classes can be regarded as inconsistent objects. Such
afnities between consistent objects in set theory and sets, and between proper
classes and inconsistent objects, though it cannot be strengthened into equivalence,
testify to the richness of this approach.
Section 8.5, the last in the chapter, starts the discussion of models of paracon-
sistent set theory. If the construction of models for standard set theory is a fraught
task, the analogue for paraconsistent set theory is adventurous, to say the least. One
might consider standard models of paraconsistent set theory, where the relation of
that model corresponds exactly to the membership relation 2 of the universe of
ZFmbC and its extensions, and the same for the consistency operator , but it is
also reasonable to make room for non-standard models. Only in this way could one
venture into deeper waters, such as extending forcing machinery to paraconsistent
Preface xvii

set theory. Although this is not done in this book, and it may be an ambitious
project, it is not unrealistic.
Chapter 9
Chapter 9 attempts to clarify the close connections between paraconsistency and
philosophy of science: in a nutshell, there are so many cases of contradictions, even
if temporary, arising between scientic theories, as well as between facts and
theories, that a paraconsistent approach to the foundations of science seem to be
almost inevitable. Section 9.1 advocates an epistemological understanding of
paraconsistency based upon the notion of evidence, and questions its signicance
for science supported by some examples of real situations, examined in Sect. 9.2.
Consistency and contradiction in scientic theories can be understood by an epis-
temic approach to paraconsistency, we claim, inspired by some Kantian insights
about the limits of human reason. Some historical examples of cases where sci-
entists have held contradictory positions, and where science as a whole has gained
from holding them, are reviewed in this section. The controversy surrounding the
movement of the luminiferous aether of the nineteenth century, the controversies in
the early development of quantum theory, the case of Mercurys orbit and the
failure in hypothesizing Vulcan, a planet that only existed in the heads of certain
astronomers, and the contradictions arising from the imponderable phlogiston in
the beginnings of the chemistry of the eighteenth century are illuminating cases.
The provisional contradictions faced by Einstein just before he formulated the
special theory of relativity in 1905 is another typical example of what we call
epistemic contradictions, which arise between two non-contradictory theories that,
when put together, yield contradictory results. The phenomenon is not restricted to
natural sciences: the imaginary numbers, which baffled mathematicians and
philosophers until the beginning of the twentieth century, is another piece of
epistemic contradiction.
Section 9.3 reviewsfrom a more philosophical perspectivethe concept of
pragmatic truth, also referred to as quasi-truth, or partial truth, already analyzed
from the formal point of view in Chap. 7. Quasi-truth, developed as part of efforts to
expand the bounds of the traditional Tarskian account of formalized truth, proposes
a partial (or pragmatic) notion of truth, intending to capture the meaning of wider,
more flexible, theories of truth held by anti-realist thinkers in philosophy of science.
Section 9.4 emphasizes the evidence-based approach to paraconsistency, in the
sense of understanding a pair of contradictory sentences as representing, and
allowing us to reason about, conflicting evidence, defending this view as particu-
larly promising for philosophical interpretations of paraconsistent logics.
The last section, Sect. 9.5, succinctly wraps up one of the chief points behind
LFIs: they are concerned with truth, since classical logic can be fully recovered
inside most of the LFIs, but they are also concerned with the notion of evidence, a
notion weaker than truth that allows for an intuitive and plausible understanding
of the acceptance of contradictions in some reasoning contexts. In this regard, both
intuitionistic and paraconsistent logics may be conceived as normative theories of
logical consequence endowed with an epistemic character. This view not only
xviii Preface

stresses the brotherhood between the intuitionistic and the paraconsistent para-
digms, but explains the adequacy of LFIs for wider accounts in the philosophy of
science, and also their applicability in the elds of linguistics, theoretical computer
science, inferential probability, and conrmation theory.

References

1. Frege, Gottlob. 1891. ber Funktion und Begriff (in German). In Jenaischen Gesellschaft fr
Medizin und Naturwissenschaft. Jena: Verlag Hermann Pohle. Translated by Peter Geach as
Function and Concept. In Translations from the Philosophical Writings of Gottlob Frege, ed.
Peter Geach and Max Black, Blackwell, 1980.
2. Carnielli, Walter A., and Joo Marcos. 2002. A taxonomy of C-systems. In Paraconsistency: The
Logical Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency
(WCP 2000), volume 228 of Lecture Notes in Pure and Applied Mathematics, eds. Carnielli,
Walter A., Marcelo E. Coniglio, and Itala M.L. DOttaviano, 194, New York: Marcel Dekker.
3. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal
Inconsistency. In Handbook of Philosophical Logic (2nd edn), ed. Dov M. Gabbay and Franz
Guenthner, vol. 14, 193. Springer, doi:10.1007/978-1-4020-6324-4_1.
4. da Costa, Newton C.A. 1963. Sistemas formais inconsistentes (Inconsistent formal systems, in
Portuguese). Habilitation thesis, Universidade Federal do Paran, Curitiba, Brazil, Republished
by Editora UFPR, Curitiba, Brazil, 1993
5. da Costa, Newton C.A. 1974. On the theory of inconsistent formal systems (Lecture delivered
at the First Latin-American Colloquium on Mathematical Logic, held at Santiago, Chile, July
1970). Notre Dame Journal of Formal Logic 15(4): 497510.
6. Esteva, Francesc and Llus Godo. 2001. Monoidal t-norm based logic: Towards a logic for
left-continuous t-norms. Fuzzy Sets and Systems 124(3): 271288.
7. Coniglio, Marcelo E., Francesc Esteva, and Llus Godo. 2014. Logics of formal inconsistency
arising from systems of fuzzy logic. Logic Journal of the IGPL 22(6): 880904, doi:10.1093/
jigpal/jzu016.
8. Bueno, Otvio and Newton C. A. da Costa. 2007. Quasi-truth, paraconsistency, and the
foundations of science. Synthese 154(3): 383399.
9. Coniglio, Marcelo E., and Luiz H. Silvestrini. 2014. An alternative approach for quasi-truth.
Logic Journal of the IGPL 22(2): 387410, doi:10.1093/ljigpal/jzt026.
Acknowledgments

We would like to express our gratitude to the many organizations and people who
read, wrote, and offered criticisms and comments, allowed us to quote their papers,
assisted in editing and proofreading, and provided support of all kindsmonetary,
philosophical, personal, and emotional. We would like to acknowledge support
from FAPESP (Thematic Project LogCons 2010/51038-0, Brazil) and from indi-
vidual research grants from The National Council for Scientic and Technological
Development (CNPq), Brazil. The intellectual environment of the Centre for Logic,
Epistemology and the History of Science (CLE) of the State University of
CampinasUNICAMP deserves a special mention: we thank the colleagues and
the ofcers of CLE for having provided all necessary facilities, from library
facilities, to secretarial work, computers, and good coffee.
Personal thanks go to Ablio Rodrigues (Belo Horizonte), Henrique Antunes
Almeida (Campinas), Peter Verde (Campinas and Brussels), Raymundo Morado
(Campinas and Mexico City), Itala DOttaviano (Campinas), Giorgio Venturi
(Campinas), David Gilbert (Campinas and Urbana), Gabriele Pulcini (Campinas),
Rodolfo Ertola (Campinas), Francesc Esteva (Barcelona), Llus Godo (Barcelona),
Josep Maria Font (Barcelona), Ramn Jansana (Barcelona), Tommaso Flaminio
(Varese), Carles Noguera (Prague), Joo Marcos (Natal), Juliana Bueno-Soler
(Limeira), Newton Peron (Chapec), Rafael Testa (Campinas), Marcio Ribeiro
(Guarulhos), Erin OConnor (Sorocaba), and Gareth J. Young (Glasgow).
Thanks also to Christi Lue and to the Springer team (Dordrecht) for their
continuous support over the many years this book took to complete. We beg
forgiveness from all those we might have unintentionally failed to mention: we
have made every effort to leave that set consistently empty.

xix
Contents

1 Contradiction and (in)Consistency . . . . . . . . . . . . . . . . . . . . ..... 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 1
1.2 On the Philosophy of the Logics of Formal Inconsistency. ..... 4
1.3 A Historical Sketch: The Forerunners of the Logics
of Formal Inconsistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Paraconsistency and the Nature of Logic . . . . . . . . . . . . . . . . . 11
1.5 Paraconsistency and the Nature of Contradictions . . . . . . . . . . . 15
1.6 Contradiction, Consistency and Negation . . . . . . . . . . . . . . . . . 18
1.6.1 On Contradiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6.2 On Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6.3 On Negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7 Varieties of Paraconsistency Involvement . . . . . . . . . . . . . . . . . 24
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 A Basic Logic of Formal Inconsistency: mbC . . . . . . . . . . . . . . . . . 29
2.1 Introducing mbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 A Valuation Semantics for mbC . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Applications of mbC-Valuations . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Recovering Classical Logic Inside mbC . . . . . . . . . . . . . . . . . . 43
2.5 Reintroducing mbC as an Expansion of CPL . . . . . . . . . . . . . . 50
2.5.1 The New Presentation mbC? of mbC . . . . . . . . . . . . . 51
2.5.2 Valuation Semantics for mbC . . . . . . . . . . . . . . . . . . . 53
2.5.3 Equivalence Between mbC and mbC? . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3 Some Extensions of mbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.1 A Wider Form of Truth-Functionality for Consistency . . . . . . . . 63
3.2 A Hidden Consistency Operator . . . . . . . . . . . . . . . . . . . . . . . 68
3.3 Consistency and Inconsistency as Derived Connectives. . . . . . . . 71
3.4 Some Conceptual Differences Between mbC and mbC? . . . . . . 91
3.5 Inconsistency Operators and Double-Negations . . . . . . . . . . . . . 95

xxi
xxii Contents

3.6 Propagating Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

3.7 da Costas Hierarchy and Consistency Propagation . . . . . . . . . . 111
3.8 A Stronger Consistency Propagation . . . . . . . . . . . . . . . . . . . . 115
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4 Matrices and Algebraizability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1 Logical Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2 Uncharacterizability by Finite Matrices . . . . . . . . . . . . . . . . . . . 122
4.3 The Problem of Algebraizability of LFIs . . . . . . . . . . . . . . . . . 129
4.4 Some 3-Valued LFIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.4.1 Halldns Logic of Nonsense (1949) . . . . . . . . . . . . . . 136
4.4.2 Segerbergs Logic of Nonsense (1965) . . . . . . . . . . . . . 138
4.4.3 da Costa and DOttavianos Logic J3 (1970). . . . . . . . . 140
4.4.4 Settes Logic P1 (1973) . . . . . . . . . . . . . . . . . . . . . . . 144
4.4.5 Asenjo-Priests Logic LP (19661979) . . . . . . . . . . . . . 149
4.4.6 Ciore and Other Related Systems . . . . . . . . . . . . . . . . 151
4.4.7 LFI1, MPT and J3 . . . . . . . . . . . . . . . . . . . . . . . . . . 158
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5 LFIs Based on Other Logics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.1 LFIs Based on Positive Intuitionistic Logic . . . . . . . . . . . . . . . . 171
5.1.1 Basic Features of Positive Intuitionistic Logic . . . . . . . . 171
5.1.2 Johanssons Minimal Logic . . . . . . . . . . . . . . . . . . . . . 175
5.1.3 Nelsons Paraconsistent Logic N4 . . . . . . . . . . . . . . . . 179
5.1.4 An Intuitionistic Version of mbC . . . . . . . . . . . . . . . . 187
5.2 LFIs Based on Fuzzy Logics. . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.2.1 Preliminaries on MFL . . . . . . . . . . . . . . . . . . . . . . . . 191
5.2.2 Fuzzy Logics with a Consistency Operator . . . . . . . . . . 197
5.2.3 Propagation of Consistency and DAT. . . . . . . . . . . . . . 205
5.2.4 Fuzzy Logics with an Inconsistency Operator . . . . . . . . 208
5.3 A Modal LFI Based on Belnap and Dunns Logic BD . . . . . . . . 211
5.3.1 The Logic M4m of Tetravalent Modal Algebras . . . . . . . 213
5.3.2 M4m as an LFI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.3.3 M4m as a dC-System . . . . . . . . . . . . . . . . . . . . . . . . . 222
5.3.4 The Contrapositive Implication . . . . . . . . . . . . . . . . . . 224
c
5.3.5 A Hilbert-Style Axiomatization of M4m ......... . . . . 226
5.4 Paraconsistent Modalities, Consistency and Determinedness . . . . 229
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
6 Semantics of Non-deterministic Character for LFIs . . . . . . . . . . . . . 237
6.1 Fidel Structures for mbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
6.2 Fidel Structures for Some Extensions of mbC . . . . . . . . . . . . . . 242
6.3 Non-deterministic Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.4 Swap Structures for mbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
6.5 Swap Structures for Some Extensions of mbC . . . . . . . . . . . . . 260
6.6 Axiom (cl) and Uncharacterizability by Finite Nmatrices . . . . . . 272
Contents xxiii

6.7 Some Remarks on Fidel Structures and Swap Structures . . . . . . . 279

6.8 The Possible-Translations Semantics . . . . . . . . . . . . . . . . . . . . 280
6.8.1 Possible-Translations Semantics for Some LFIs. . . . . . . 280
6.8.2 A 3-Valued Possible-Translations Semantics for Cila . . . 282
6.8.3 Some Remarks on Possible-Translations Semantics . . . . 288
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
7 First-Order LFIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
7.1 The Logic QmbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
7.2 Basic Properties of QmbC . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
7.3 Tarskian Paraconsistent Structures . . . . . . . . . . . . . . . . . . . . . . 302
7.4 Soundness Theorem for QmbC . . . . . . . . . . . . . . . . . . . . . . . . 308
7.5 Completeness Theorem for QmbC. . . . . . . . . . . . . . . . . . . . . . 310
7.5.1 Henkin Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
7.5.2 Canonical Interpretations. . . . . . . . . . . . . . . . . . . . . . . 312
7.6 Compactness and Lowenhim-Skolem Theorems for QmbC . . . . 316
7.7 QmbC with Equality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
7.8 First-Order Characterization of Other Quantied LFIs . . . . . . . . 322
7.9 First-Order LFI1 and the Logic of Quasi-truth . . . . . . . . . . . . . 324
7.9.1 Semantics of Partial Structures. . . . . . . . . . . . . . . . . . . 324
7.9.2 The Logic QLFI1 . . . . . . . . . . . . . . . . . . . . . . . . . . 332
7.10 First-Order P1 and Partial Structures . . . . . . . . . . . . . . . . . . . . 335
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
8 Paraconsistent Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
8.1 Antinomic Sets and Paraconsistency. . . . . . . . . . . . . . . . . . . . . 346
8.2 LFIs Predicating on Consistency . . . . . . . . . . . . . . . . . . . . . . . 349
8.3 Some Extensions of ZFmbC. . . . . . . . . . . . . . . . . . . . . . . . . . 354
8.4 Inconsistent Sets and Proper Classes . . . . . . . . . . . . . . . . . . . . 360
8.5 On Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
9 Paraconsistency and Philosophy of Science: Foundations
and Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
9.1 An Epistemological Understanding of Paraconsistency,
and Its Signicance for Science . . . . . . . . . . . . . . . . . . . . . . . . 369
9.2 Consistency and Contradiction in Scientic Theories . . . . . . . . . 371
9.2.1 The Heritage of Kant . . . . . . . . . . . . . . . . . . . . . . . . . 371
9.2.2 Some Historical Examples . . . . . . . . . . . . . . . . . . . . . 373
9.2.3 The Beginning of Quantum Theory and
Paraconsistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
9.2.4 Mercurys Orbit and a Non-existent Planet . . . . . . . . . . 374
9.2.5 Contradictions in Phlogiston, the Imponderable . . . . . . . 376
9.2.6 The Special Theory of Relativity . . . . . . . . . . . . . . . . . 377
9.2.7 Mathematics, and the Meaning of Objects
that Mean Nothing . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
xxiv Contents

9.3 Quasi-truth and the Reconciliation of Science

and Rationality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
9.4 An Evidence-Based Approach to Paraconsistency . . . . . . . . . . . 384
9.5 Summing Up. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391

Index of Logic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397

Chapter 1
Contradiction and (in)Consistency

1.1 Introduction

The target audience of this book is mainly the philosopher, the logician interested in
the philosophical aspects of paraconsistency, and the computer scientist looking for
new logics for applications. But the intended audience also includes the mathemati-
cian intrigued by the possibility of working in a logic that allows contradictions (a
paraconsistent logic), the linguist worried about the acceptance of contradictions in
the ordinary speech, and the scientist interested in the significance of contradictions
in the history of science.1
The reader of this book is invited, first of all, to take into account that contradictions
are pervasive in scientific theories, in philosophical argumentation, in several areas
of computer science such as abduction, automated reasoning, logic programming,
belief revision and the semantic web. People negotiating a contract, as buyers and
sellers, many times encounter contradictions, and strive to overcome them in order
to strike a deal. Paradoxes in formal semantics, as the famous liar paradox, are
seen as dangerous to the standard theories of truth, and paradoxes in naive (albeit
intuitively acceptable) set theory are seen as threats to the foundations of science and
mathematics.
However, contradictory information is not only frequent, and more so as systems
increase in complexity, but can have a positive role in human thought, in some cases
being desirable. Finding contradictions in juridical testimonies, in statements from
suspects of a crime or in suspects of tax fraud can be an efficient strategy. Contra-
dictions can be very informative: we will never know if people being questioned
coherently lie or not, unless they contradict each other!

1 Thischapter corresponds in part with the tutorial on Logics of Formal Inconsistency presented in
the 5th World Congress on Paraconsistency (Kolkata, India, February 2014), see [1]. Parts of that
material have already appeared in [2].

Springer International Publishing Switzerland 2016 1

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_1
2 1 Contradiction and (in)Consistency

The current orthodoxy is that all contradictions are equally virulent, in view of
the principle of Ex Contradictione Sequitur Quodlibet (ECSQ), or The Principle of
Explosion (PE), the principle that holds that from a contradiction, anything logically
follows. But how can standard logic, which endorses ECSQ, impose a principle that is
not followed by common reasoning? Are all contradictions really equally hazardous?
The so called Bar-Hillel-Carnap paradox (see [3], p. 229) has already suggested,
half century ago, the clash between the notions of contradiction and semantic infor-
mation: the less probable a statement is, the more informative it is, and so contradic-
tions carry the maximum amount of information, and in the light of standard logic
are, as a famous quote by Bar-Hillel and Carnap has it, too informative to be true.
This is a difficult philosophical problem for standard logic, which is forced to equate
triviality and contradiction, and to regard all contradictions as equivalent, as the fol-
lowing example illustrates. If two auto technicians tell me that the battery of my
car is flat, and its electrical system out of order, and add all the (potentially infinite)
statements about car electrics, I have an excessive amount of information, including a
huge amount of irrelevant information. Classically, this trivial amount of information
is exactly the same as the information conveyed by the car technicians telling me a
contradiction, such as the battery of my car is flat and that it is not flat. However, if
one of the car technicians tells me (among his statements) that the battery is flat, and
the other that the battery is not flat, between them they are contradictory, but now I
know where the problem is! Skipping all technicalities in favor of a clear intuition
(details are given elsewhere), the Bar-Hillel-Carnap observation is not paradoxical
for LFIs since, as will be clear in the following, LFIs do not treat all contradictions
equivalently, and do not equate contradiction with triviality.
The idea that any contradiction inexorably leads to deductive explosion (by means
of ECSQ) seems to have entered logical orthodoxy towards the end of the 19th century
at the hands of G. Frege, B. Russell, D. Hilbert and W. Ackermann, pioneered by G.
Boole. As outlined in [4], the logic of antiquity did not endorse the validity of ECSQ,
and the principle only became a topic of debate in the Middle Ages or Medieval era.
It is a plausible, though debatable, conjecture that what is now known as ECSQ,
sometimes confused2 with Ex Falso Sequitur Quodlibet, or Ex Impossibile Sequitur
Quodlibet 3 might have been originated in the 14th century ideas of John of Cornwall
(quite possibly the Pseudo-Duns Scotus himself).
The incorporation of the principle ECSQ into contemporary logic had resounding
consequences: in standard logic, a theory  is by definition consistent if no pairs
of contradictory sentences , are deducible from . Consequently, in light of
ECSQ, the notion of consistency is inescapably dependent on negation, and  is
consistent if and only if it is not deducibly trivial.

2 Example 14 (p. 15) of [5] provides an example of a logic that respects the principle of Ex Falso
Sequitur Quodlibet, but not the ECSQ, showing that those principles do not need to be identified,
contrary to what is commonly held in the literature.
3 To the best of our knowledge, the exact expressions Ex Contradictione Sequitur Quodlibet and Ex

Contradictione Quodlibet have been independently coined by, respectively Priest and Bobenrieth-
Miserda, see [6].
1.1 Introduction 3

Glorious results of the 20th century such as K. Gdels proof of the consistency of
Cantors continuum hypothesis with the axioms of set theory, his celebrated theorem
on the impossibility of proving the consistency of arithmetic by finitary means, as well
as P. Cohens forcing technique for proving consistency and independence results in
set theory, among many others, all depend upon the acceptance of that definition of
consistency.
However, a logical system need not endorse ECSQ; paraconsistent logic avoids
this by adopting a more cautious strategy: a logical consequence relation   is
said to be paraconsistent if it is not explosive, and yet preserves enough properties
to be accounted as a logic. Relatedly, the notion of consistency need not be seen
from the aforementioned perspective: the spirit of the LFIs to be developed in full in
the next chapters shows that other mathematically well-founded and philosophically
palatable approaches exist, and have exciting and relevant consequences.
The LFIs are a family of propositional and quantified paraconsistent logics that
encode consistency (and inconsistency) as operators independent of negation in their
object language. Encoding consistency and inconsistency in this way has the effect
of permitting an explicit separation between, respectively, contradiction from incon-
sistency, inconsistency from triviality, consistency from non-contradiction, and non-
triviality from consistency. The rich language of LFIs make it possible to investigate
contradictory theories without assuming that they are necessarily trivial.
The LFIs are proper fragments of classical logic (hence non trivial) yet rejecting
ECSQ in the presence of a contradiction, unless the contradictory sentence is taken to
be consistent. The family of LFIs incorporate a great number of paraconsistent sys-
tems of various sorts, in a sense to be explained, including the well-known hierarchy
of logics introduced by Newton da Costa in Brazil in the sixties.
The idealization behind LFIs is that assertions about the world should be divided
into two categories: consistent sentences and non-consistent sentences. Consistent
propositions are subjected to classical logic, and consequently a theory  that contains
a pair of contradictory sentences , explodes only if (or ) is taken to be a
consistent sentence, linguistically marked as (or ). This is the only distinction
between LFIs and classical logic, albeit with far-reaching consequences: classical
logic in this way is expanded, in such a way that in most cases an LFI encodes
classical logic.
The rest of the book will make all this crystal clear, but some previous misunder-
standings should be noted and avoided. Firstly, paraconsistent logic in general, and
LFIs in particular, do not prove contradictions: these logic systems only support rea-
soning under hypothetical contradictions. Second, LFIs do not have any relationship
with dialetheism, a philosophical view in which there are true contradictions (see,
for instance, [7, 8]). If dialetheists turn out to be correct, and some contradictions
are actually true, the LFIs will be on their side. If not, the LFIs will continue to
be of value, independently of this debate. In this regard LFIs are a metaphysically-
free theory of deduction. Third, the LFIs are extensible to first-order logic (and
to higher-order logics as well), and are powerful enough to reproduce, under sub-
tle adaptations, all the main metamathematical results of traditional logic, such as
completeness, compacteness, most model theoretical properties of quantified logic,
decidability, and so on.
4 1 Contradiction and (in)Consistency

The history of expansion of concepts in mathematics has a long tradition, since the
negative numbers, zero and imaginary numbers have been called, variously, sophis-
tic, impossible or inexplicable. Science and engineering would be hardly pos-
sible without such fictions today, and the centuries following the introduction of
imaginary numbers (the name itself suggested a pejorative connotation) have wit-
nessed many other sophistic creatures, such as G. Cantors cardinal arithmetic,
A. Robinsons non-standard numbers, the hyperreal numbers, fractal dimensions,
and a number of other similar ideas. Perhaps the idea of consistency as a primi-
tive concept will be recognized as of a similar sortthe reader is invited to judge
her/himself.
The following section discusses some primary philosophical issues related to
paraconsistency in general, and especially to the LFIs. Our basic standpoint is that
there are two basic and philosophically legitimate approaches to paraconsistency that
depend on whether the contradictions are understood ontologically or epistemologi-
cally. LFIs are well suited to both options, but we shall emphasize the epistemological
interpretation of contradictions. The main argument depends on the duality between
paraconsistency and paracompleteness. Briefly, the idea is as follows: just as excluded
middle may be rejected by intuitionists for epistemological reasons, explosion may
also be rejected by paraconsistentists for epistemological reasons as well.

1.2 On the Philosophy of the Logics of Formal

Inconsistency

It is a fact that contradictions appear in a number of real-life contexts of reasoning.

Databases very often contain not only incomplete information but also conflicting
(i.e. contradictory) information.4 Since ancient Greece, paradoxes have intrigued
logicians and philosophers, and, more recently, mathematicians as well. Scientific
theories are another example of real situations in which contradictions seem to be
unavoidable. There are several scientific theories, however successful in the areas
with which they are primarily concerned, that yield contradictions, either by them-
selves or in combination with other successful theories. Contradictions are problem-
atic when the Principle of Explosion, or ECSQ, holds:
(PE) ( ).5

4 The term information is not used here in a strictly technical sense. We might say, in an attempt to

elucidate, rather than define, that information means any amount of data that can be expressed by
a sentence in natural language. Accordingly, there may be contradictory or conflicting information
(in a sense to be clarified below), vague information, or lack of information.
5 The symbol will always denote the classical negation, while usually denotes a paraconsistent

negation but sometimes a paracomplete (e.g. intuitionistic) negation. The context will make clear
in each case whether the negation is being used in a paracomplete or a paraconsistent sense.
1.2 On the Philosophy of the Logics of Formal Inconsistency 5

In this case, since anything follows from a contradiction, one may conclude anything
whatsoever. In order to deal rationally with contradictions, explosion cannot be valid
without restrictions, since triviality (that is, a circumstance such that everything
holds) is obviously unacceptable. Given that in classical logic explosion is a valid
principle of inference, the underlying logic of a contradictory context of reasoning
cannot be classical.
Simply put, paraconsistency is the study of logical systems in which the presence
of a contradiction does not imply triviality, that is, logical systems with a non-
explosive negation such that a pair of propositions and does not (always)
trivialize the system. However, it is not only the syntactic and semantic properties
of these systems that are worth studying. Some questions arise that are perennial
philosophical problems. The question of the nature of the contradictions allowed
in paraconsistent logic has been a particular focus of debates on the philosophical
significance of paraconsistency.
In philosophical terminology, we say that something is ontological when it has to
do with reality, the world in the widest sense, and that something is epistemological
when it has to do with knowledge and the process of its acquisition. A central question
for paraconsistency is the following: are the contradictions that paraconsistent logic
deals with ontological or epistemological? Do contradictions have to do with reality
proper? That is, is reality intrinsically contradictory, in the sense that we really need
some pairs of contradictory propositions in order to describe it correctly? Or do
contradictions have to do with knowledge and thought? Contradictions of the latter
kind would have their origin in our cognitive apparatus, in the failure of measuring
instruments, in the interactions of these instruments with phenomena, in operations
of thought, or even in simple mistakes that in principle could be corrected later on.
Note that in all of these cases the contradiction does not belong to reality properly
speaking.
The question of nature of contradictions, in its turn, is related to another central
issue in philosophy of logic, namely, the nature of logic itself. As a theory of logical
consequence, the task of logic is to formulate principles and methods for establishing
when a sentence follows from a set of premises . But a question remains: what are
the principles of logic about? Are they about language, thought, or reality? That logic
is normative is controversial, but if logic is anyhow normative for thought, its nor-
mative character may be combined both with an ontological and an epistemological
approach.
The epistemological side of logic is present in the widespread (but not unanimous)
characterization of logic as the study of laws of thought. This concept of logic, which
acknowledges an inherent relationship between logic and human rationality, has been
put aside since classical logic has acquired the status of the standard account of logical
consequencefor example, the work of Frege, Russell, Tarski, Quine and a many
other influential logicians.
Classical logic is a very good account of the notion of truth preservation, but it
does not give a sustained account of rationality. This point shall not be developed
in detail here, but it is well known that some classically valid inferences are not
really applied in real-life contexts of reasoning, for example: from , to conclude
6 1 Contradiction and (in)Consistency

that anything implies ; from , to conclude the disjunction of and anything; from
a contradiction, to conclude anything. The latter is the principle of explosion, and
of course it is not rational to conclude that 2 + 2 = 5 when we face some pair of
contradictory propositions. Nevertheless, from the point of view of preservation of
truth, given the classical meaning of sentential connectives, all the inferences above
are irreproachable.
We assume here a concept of logic according to which logic is not restricted to
the idea of truth preservation. Logical consequence is indeed the central notion of
logic, but the task of logic is to tell us which conclusions can be drawn from a given
set of premises, under certain conditions, in concrete situations of reasoning. We
shall see that sometimes it may be the case that it is not only truth that is at stake.6
Among the contexts of reasoning in which classical logic is not the most suitable
tool, two are especially important: contexts with excess of information and lack
of information. The logics suited to such contexts are, respectively, paraconsistent
and paracompletein the former, explosion fails, in the latter excluded middle fails.
There are two basic approaches to paraconsistency. If some contradictions are
true, since it is not the case that everything holds, we need an account of logical
consequence that does not collapse in the face of a contradiction. On the other hand,
if contradictions are epistemological, we shall argue that the rejection of explosion
goes hand in hand with the rejection of excluded middle by intuitionistic logic. In
the latter case, the formal system has an epistemological character and combines a
descriptive with a normative approach.
In the next sections, some basic concepts will be introduced in order to distin-
guish triviality from inconsistency. In addition, we shall make a first presentation
of LFIs, distinguishing paraconsistency and paracompleteness from the classical
approach. In Sect. 1.3 a brief historical digression on the origins of paraconsistency
and the forerunners of the Logics of Formal Inconsistency (henceforward referred to
as LFIs) will be delineated. Section 1.4 will examine the relationship between para-
consistency and the issue of the nature of logic. We will argue that, like the rejection
of excluded middle by intuitionistic logic, the rejection of explosion may be under-
stood epistemologically. Section 1.5 discusses paraconsistency from the point of view
of the issue of the nature of contradictions, and considers whether they should be
understood ontologically or epistemologically. We shall argue that both positions are
philosophically legitimate. Finally, we will be ready to show how the simultaneous
attribution of the value 0 (or false) to a pair of sentences and may be interpreted
as conflicting evidence, not as truth and falsity of .
We have seen that paraconsistent logics are able to deal with contradictory sce-
narios, avoiding triviality by means of the rejection of the principle of explosion.
Let us put these ideas more precisely. A theory is a set of sentences closed under
logical consequence. Given a set of sentences  in the language of a given logic L,
let T = { :  L } be the theory whose non-logical axioms are the sentences of

6 This idea has some consequences for Harmans arguments [9] against non-classical logics, a point

that we intend to develop elsewhere.

1.2 On the Philosophy of the Logics of Formal Inconsistency 7

 and the underlying logic is L. Suppose the language of T has a negation . We

say that T is:
Contradictory if and only if there is a proposition in the language of T such
that T proves and T proves .
Trivial if and only if for any proposition in the language of T , T proves ;
Explosive if and only if T trivializes when exposed to any pair of contradictory
formulasi.e.:
T {, }  , for all and .

In books of logic we find two different but classically equivalent notions of con-
sistency with respect to a deductive system S with a negation .
i. S is consistent if and only if there is a formula such that  S ;
ii. S is consistent if and only if there is no formula such that  S and  S .
What (i) says is that S is non-trivial; and (ii) says that S is non-contradictory. In
classical logic both are provably equivalent.
So, a theory whose underlying logic is classical is contradictory if and only if it
is trivial. But this is the case precisely because such a theory is explosive, since the
principle of explosion holds in classical logic. It is clear, then, that contradictoriness
is not necessarily coincident with explosiveness. The obvious move in order to deal
with contradictions is, thus, to reject the unrestricted validity of the principle of
explosion. This is a necessary condition if we want a contradictory but not-trivial
theory.
The first formalization of paraconsistent logic to appear in the literature is to
be found in [10]. In the beginning of the paper he presents three conditions that a
contradictory but non-trivial logic must attend:
1. It must be non-explosive;
2. It should be rich enough to enable practical inference;
3. It should have an intuitive justification.
Condition (1), as we have seen, is a necessary condition for any paraconsistent
system. We want to call attention to conditions (2) and (3). Indeed, the biggest
challenge for a paraconsistentist is to devise a logical system compatible with what
we intuitively think should follow (or not follow) from what. This is the idea expressed
by the criteria (2) and (3) presented by Jaskowski. An intuitive and applicable notion
of logical consequence should be appropriate for describing and reconstructing the
actual reasoning going on in real-life contexts. An intuitive account of the meaning
of the logical connectivesmore precisely, of paraconsistent negationshould be
an integral part of such an account of logical consequence. It follows that an intuitive
interpretation of a paraconsistent notion of logical consequence depends essentially
on an intuitive interpretation of negation.
A negation is classical or strong if the following conditions hold:
1. 
2. 
8 1 Contradiction and (in)Consistency

According to (1), there is no model M such that holds in M. (2) says that
for every model M, holds in M. Now, given the definition of classical
consequence, follows from anything, and anything follows from .7
We say that a negation is paracomplete if it disobeys (2), and that a negation is
paraconsistent if it disobeys (1). From the point of view of rules of inference, the
duality is not between non-contradiction and excluded middle, but rather between
explosion and excluded middle. Notice that the notion of logical consequence has
priority over the notion of logical truth: the latter must be defined in terms of the
former, not the contrary. The principle of non-contradiction is usually taken as the
claim that reality is not contradictory. But we may well understand the principle of
explosion as a stronger way of saying precisely the same thing: and cannot
hold together, otherwise we get triviality. From the above considerations it is clear
that in order to give a counterexample to the principle of explosion we need a weaker
negation and a semantics in which there is a model M such that neither nor
hold in M (here, is a paraconsistent negation) but for some , does not hold in
M. Dually, a paracomplete logic must have a model M such that both and does
not hold in M (here, is a paracomplete negation), no matter what happens with
any other formula .
A central feature of classical negation (but not of all negations, as we shall see)
is that it is a contradictory forming operator. This is due to its semantic clause,
M() = 1 iff M() = 0
that, in turn, holds because both (1) and (2) above hold. Applied to a proposition ,
classical negation produces a proposition such that and are contradictories
in the sense that they cannot simultaneously receive the value 0, nor simultaneously
the value 1. In classical logic the truth-values 0 and 1 are understood respectively as
false and true, but in non-classical logics this does not need to be the case. It is not
necessary that a paracomplete logic takes a pair of formulas and as both false,
nor that a paraconsistent logic takes them as both true.
Obviously, neither a paracomplete nor a paraconsistent negation is a contradictory
forming operator, and neither is a truth-functional operator, since the value of is
not unequivocally determined by the value of . Now a question arises: can we say
that such negations are really negations? Our answer is yes.
It should not be surprising that the meaning of a classical connective splits up into
some alternative meanings when its use in natural language and real-life arguments is
analyzed. Indeed, different meanings are sometimes attached to conditional, disjunc-
tion, and conjunction, and the connectives so obtained are still called the conditional,
disjunction, and conjunction, of course with some qualifications. What would be the
reason by which the same cannot occur with negation? In fact, both paracomplete
and paraconsistent negations do occur in real life. An obvious example of the for-
mer is intuitionistic negation: it may be the case that we do have a classical proof
of a proposition but have no constructive proof of . Given that we cannot have

7 For a more detailed account on the duality between paracompleteness and paraconsistency, see
e.g. [11].
1.2 On the Philosophy of the Logics of Formal Inconsistency 9

proofs of and , from the constructive point of view, we have neither nor .
On the other hand, sometimes it happens that we have to deal simultaneously with
conflicting information about . In these cases, we may have reasons to accept both
and , but we do not need to say that both are true. Finally, the above consid-
erations show that a paraconsistent negation is a negation to the same extent that a
paracomplete (including intuitionistic) negation is a negation. Nevertheless, what is
of major importance is that the question of whether or not a paraconsistent negation
may have an intuitive meaning has a positive answer.
In this section we shall present the basic ideas of LFIs without going into the
technical details, which will be left for the following chapters. As remarked, LFIs
have resources to express the notion of consistency inside the object language by
means of a sentential unary connective called circle: means that is consistent.
As in any other paraconsistent logic, explosion does not hold in LFIs. But it is handled
in a way that allows distinguishing between contradictions that can be accepted from
those that cannot. The point of this distinction is that no matter the nature of the
contradictions a paraconsistentist is willing to accept, there are contradictions that
cannot be accepted. In LFIs, negation is explosive only with respect to consistent
formulas (that is, formulas that are taken to be consistent):
,  L F I , while , ,  L F I .
An LFI is thus a logic that separates the sentences for which explosion holds from
those for which it does not hold. The former are marked with . For this reason, they
are called gently explosive.
The idea of expressing a kind of logical well-behavior in the object language
is also found in da Costas Cn hierarchy. In C1 , the normality of a sentence is
expressed by , in such a way that:
, C1 , while , , C1
(see Sect. 3.7 of Chap. 3). However, in C1 , is an abbreviation of ( ),
which makes the consistency of a proposition equivalent to saying that is non-
contradictory.8 The generalization from the original hierarchy Cn of da Costa and
the concept of LFIs is quite significative: da Costa in [12] named his logical systems
inconsistent formal systems, while he was really referring to contradictory and
non-trivial formal systems. The notion of the consistency of a sentence () is
only apparently similar to the well-behavedness of a sentence ( ): actually, the
idea of consistency as a primitive notion represents a radical departure from main-
stream paraconsistency, since it allows us to simultaneously capture a number of
paraconsistent systems (old and new) and to give a more comprehensible account of
paraconsistency generally.
We may say, thus, that a first step in paraconsistency is the distinction between
triviality and contradictoriness. But there is a second step, namely, the distinction

8 Actually,da Costa defined a hierarchy of systems, starting with the system C1 . A full hierarchy
of calculi Cn , for n natural, is defined and studied in [12]. Each Cn has its own definition of
well-behavedness.
10 1 Contradiction and (in)Consistency

between consistency and non-contradictoriness. In LFIs the consistency connective

is not only primitive, it is also not always logically equivalent to non-contradiction.
This is the most distinguishing feature of the logics of formal inconsistency. Once we
break up the equivalence between and ( ), some very interesting devel-
opments become available. Indeed, may express other notions of consistency, not
necessarily related to freedom from contradiction.

1.3 A Historical Sketch: The Forerunners

of the Logics of Formal Inconsistency

The advent of paraconsistency occurred more than a century ago. In 1910 the Russian
philosopher and psychologist N.A. Vasiliev proposed the idea of a non-Aristotelian
logic, free of the laws of excluded middle and non-contradiction. By analogy with the
imaginary geometry of Lobachevsky, Vasiliev called his logic imaginary, meaning
that it would hold in imaginary worlds. Despite publishing, between 19121913,
some conceptual papers on the subject, Vasiliev was not concerned with formalizing
his logic (see [13], pp. 307ff.).
Jaskowski in [10], trying to answer a question posed by ukasiewicz, presented the
first formal system for a paraconsistent logic, called discussive logic. This system
is connected to modalities, and later on came to be regarded as a particular member
of the family of the logics of formal inconsistency (see [5]).
Intending to study logical paradoxes from a formal perspective, S. Halldn
proposed in [14] a logic of nonsense by means of 3-valued logical matrices,
closely related to the nonsense logic introduced in 1938 by the Russian logician
D.A. Bochvar. Since its third truth-value is distinguished, Hlldens logic is para-
consistent, and it can also be reckoned as as one of the first paraconsistent formal
systems presented in the literature. In fact, like Jaskowskis logic, it is also an LFI.
Nelson in [15] proposed an extension of positive intuitionistic logic with a new
connective for constructible falsity or strong negation,9 intended to overcome
non-constructive features of intuitionistic negation. By eliminating the principle of
explosion from this system, [16] obtained a first-order paraconsistent logic, although
paraconsistency was not his primary concern. Indeed, the famous Nelsons paracon-
sistent logic was proposed some decades after, in [17].
Paraconsistency also has some early links to K. Poppers falsificationism. In 1954
(see [18]), K.J. Cohen, following the suggestion of his supervisor K. Popper, sub-
mitted to the University of Oxford a thesis entitled Alternative Systems of Logic
in which he intended to develop a logic dual to intuitionistic logic. In Cohens
logic, the law of explosion is no longer valid, while the law of excluded middle
holds as a theorem. Cohens thesis, according to Kapsner et al., escaped scholarly
attention, having been only briefly mentioned in Poppers famous Conjectures and

9 The reader is warned that, along this book, the expression strong negation is reserved for a
negation with a Boolean character.
1.3 A Historical Sketch: The Forerunners of the Logics of Formal Inconsistency 11

Refutations (see [19], footnote 8, p. 321). It did, however, in some sense antic-
ipate more recent work on dual-intuitionist logics (which, as shown in [11], are
paraconsistent).
In [20] we find a discussion of the status of contradiction in mathematics, intro-
ducing the Principle of Non-Trivialization, according to which non-triviality is more
important than non-contradiction. The idea is that any mathematical theory is worth
studying, provided it is not trivial. While we agree that non-trivial mathematical (and
logical) systems are worth studying, a little more is required for an account of logical
consequence to be accepted as an account of reasoning.
As already mentioned in the previous section, in 1963 N.C.A. da Costa (see [12])
presented his famous hierarchy of paraconsistent systems Cn (for n 1), constituting
the broadest formal study of paraconsistency proposed up to that time. It is worth
mentioning here what has been said by da Costa, in private conversation: As with
the discovery of America, many people are said to have discovered paraconsistent
logic before my work. I can only say that, as with Columbus, nobody has discovered
paraconsistency after me, just as nobody discovered America after Columbus.
The Argentinian philosopher F. Asenjo introduced, in 1966 (see [21]), a three-
valued logic as a formal framework for studying antinomies. His logic is essentially
defined by Kleenes three-valued truth-tables for negation and conjunction, where a
third truth-value is distinguished. Asenjos logic is structurally the same as the Logic
of Paradox LP studied by Priest in [22] more than a decade later.
From the 1970s on, after the Peruvian philosopher F. M. Quesada, at da Costas
request, coined the name paraconsistent logic to encompass all these creations,
several schools with different aims and methods have spread out around the world.10

1.4 Paraconsistency and the Nature of Logic

A central question in philosophy of logic concerns the nature of logical principles,

and specifically whether these principles are about reality, thought, or language. We
find this issue brought forth, either implicitly or explicitly, in a number of places. In
this section we shall discuss the relationship between paraconsistent logic and the
problem of the nature of logic.
Aristotle formulates three versions of the principle of non-contradiction, each one
corresponding to one of the aforementioned aspects of logic (more on this below).
Tugendhat et al. [23], Chap. 1, presents the problem mainly from a historical view-
point, relating the three approaches (ontological, epistemological and linguistic) to
periods in the history of philosophyrespectively ancient and medieval, modern, and
contemporary. Popper [19], pp. 206ff, presents the problem as follows. The central
question is whether the principles of logic are:
(I.a) Laws of thought in the sense that they describe how we actually think;
(I.b) Laws of thought in the sense that they are normative laws, i.e., laws that tell us
how we should think;

10 See Carta de F.M. Quesada a N.C.A. da Costa, 29.IX.1975 in [13], p. 609.

12 1 Contradiction and (in)Consistency

(II) The most general laws of nature, i.e., laws that apply to any kind of object;
(III) Laws of certain descriptive languages.
There are three basic options, which are not mutually exclusive: the laws of logic
have (I) epistemological, (II) ontological, or (III) linguistic character. With respect to
(I), they may be (I.a) descriptive or (I.b) normative. These aspects may be combined.
In many accounts, logic is taken as having a normative character, no matter whether
it is conceived primarily as having to do with language, thought or reality. The point
of asking this question is not really to find a definitive answer. It is a perennial philo-
sophical question, which, nonetheless, helps us to clarify and understand important
aspects of paraconsistent logic.
According to widespread opinion, a linguistic conception of logic has prevailed
during the 20th century. From this perspective, logic has to do above all with the
structure and functioning of certain languages. We do not agree with this view. For us,
logic is primarily a theory about reality and thought.11 The linguistic aspect appears
only inasmuch as language is used in order to represent what is going on in reality
and in thought. Although the linguistic aspects of logic are related to epistemology
(since language and thought cannot be completely separated) and to ontology (by
means of semantics), we do not think that a linguistic conception of logic is going
to help much in clarifying a problem that is central for us here, that of whether
contradictions have to do with reality or thought.
Aristotle, defending the principle of non-contradiction (PNC), makes it clear that
it is a principle about reality, language, and thought, but there is a consensus among
scholars that its main formulation is a claim about objects and properties: it cannot be
the case that the same property belongs and does not belong to the same object. Put
in this way, PNC is ontological in character. Like a general law of nature, space-time
phenomena cannot disobey PNC, nor can mathematical objects.
The epistemological aspects of logic became clear in the modern period. A very
illuminating passage can be found in the so-called Logic of Port-Royal ([25], p. 23),
where we read that logic has three purposes:
The first is to assure us that we are using reason well.
The second is to reveal and explain more easily the errors or defects that can occur in mental
operations.
The third purpose is to make us better acquainted with the nature of the mind by reflecting
on its actions.

Notice how the passage above combines the normative character of logic with an
analysis of mind. This view of logic does not fit very well with the account of logical
consequence given by classical logic, but it has a lot to do with intuitionistic logic.
Freges Begriffsschrift [26] had an important role in establishing classical logic
as the standard account of logical consequence. Although there is no semantics in
Freges work, it is well known that we find in the Begriffsschrift a complete and
correct system of first-order classical logic. At first sight, Freges approach is purely

11 A rejection of the linguistic conception of logic, and a defense of logic as a theory with ontological

and epistemological aspects, can be found in the Introduction to [24].

1.4 Paraconsistency and the Nature of Logic 13

proof-theoretical, but one should not draw the conclusion that his system has no
ontological commitments. We cannot loose sight of the fact that the idea of truth
preservation developed by Frege, although worked out syntactically, is constrained
by a realist notion of truth.
Frege had a realist concept of logic, according to which logic is independent of
language and mind. In fact, since he was a full-blooded platonist with respect to
mathematics, and his logicist project was to prove that arithmetic is a development
of logic, he had to be a logical realist. For Frege, the laws of logic are as objective as
mathematics, even though we may occasionally disobey them.12 Freges conception
of logic is very well suited to the idea of truth-preservation. He indeed famously
explains the task of logic as being to discern the laws of truth [28], or more precisely,
the laws of preservation of truth. Hence, it is not surprising that laws of logic cannot
be obtained from concrete reasoning practices. In other words, logic cannot have a
descriptive aspect, in the sense of (I.a) above.13 It is worth noting that Frege proves
the principle of explosion as a theorem of his system; it is Proposition 36 of the
Begriffsschrift.
It is important to emphasize the contrast between Freges and Brouwers concep-
tions of logic. This fact is especially relevant for our aims here because of the duality
between paracompleteness and paraconsistency pointed out in Sect. 1.1 above. From
the point of view of classical logic, the rejection of excluded middle by intuitionistic
logic is like a mirror image of the rejection of explosion.
It is well known that for Brouwer mathematics is not a part of logic, contrary to
what Frege wanted to prove. Quite the contrary, logic is abstracted from mathematical
reasoning. Mathematics is a product of the human mind, and mathematical proofs
are mental constructions that do not depend on language or logic. The role of logic
in mathematics is only to describe methodically the constructions carried out by
mathematicians.14 We may say that intuitionistic logic has been obtained through
an analysis of the functioning of mind in constructing mathematical proofs. To the
extent that intuitionistic logic intends to avoid improper uses of excluded middle,
it is normative, but it is descriptive precisely in the sense that, according to Frege,
logic cannot be descriptive. Intuitionistic logic thus combines a descriptive with a
normative character.

12 See [27], p. 13: they are boundary stones set in an eternal foundation, which our thought can
overflow, but never displace.
13 There is a sense in which for Frege laws of logic are descriptive: they describe reality, as well as

laws of physics and mathematics. But we say here that a logic is descriptive when it describes, in
some way, actual reasoning.
14 Brouwer [29]: Mathematics can deal with no other matter than that which it has itself constructed.

In the preceding pages it has been shown for the fundamental parts of mathematics how they can
be built up from units of perception. [] The words of your mathematical demonstration merely
accompany a mathematical construction that is effected without words [] While thus mathematics
is independent of logic, logic does depend upon mathematics. A more acessible presentation of
the motivations for intuitionistic logic is to be found in [30].
14 1 Contradiction and (in)Consistency

The view according to which intuitionistic logic has an epistemological character

that contrasts with the ontological vein of classical logic is not new.15 Note how the
intuitionistic approach fits in well with the passage quoted above from logic of Port
Royal. Furthermore, even if one wants to insist on an anti-realist notion of truth, the
thesis that intuitionistic logic is not about truth properly speaking, but about mental
constructions, is in line with the intuitionistic program as it was developed by Heyting
and Brouwer.16
Now we may ask: does intuitionistic logic give an account of truth preservation?
Our answer is negative, because intuitionistic logic is not only about truth; it is
about truth and something else. We may say that it is about constructive truth in the
following sense: it is constrained by truth but it is not truth simpliciter; rather, it is truth
achieved in a constructive way. Accordingly, not only the failure of excluded middle,
but the whole enterprise of intuitionistic logic, may be seen from an epistemological
perspective.
It is worth noting that Brouwers and Heytings attempts to identify truth with a
notion of proof have failed, as [32] shows, because the result is a concept of truth that
goes against some basic intuitions about truth. The notion of constructive provability
is stronger than truth in the sense that if we have a constructive proof of , we know
that is true, but the converse may not hold.
An analogous interpretation can be made with respect to contradictions in para-
consistent logics. While in intuitionistic logic (and paracomplete logics in general)
the failure of excluded middle may be seen as a kind of lack of information (no proof
of , no proof of ), the failure of explosion may be interpreted epistemologically
as excess of information (for example, conflicting evidence both for and for ,
but no concluding evidence for any ; we shall consider this point in more detail
later). The acceptance of contradictory propositions in some circumstances need not
mean that reality is contradictory: on the contrary, it may be accounted as a step in
the process of acquiring knowledge that, at least in principle, could be revised.
Imagine a context of reasoning such that there are some propositions well estab-
lished as true, or as false, and some others that have not been conclusively established
yet. Now, if among the latter there is a contradiction, one does not conclude that the
moon is made of green cheese, but, rather, one takes a more careful stance with respect
to the specific contradictory proposition. On the other hand, the inferences allowed
with respect to propositions already established as true are normally applied. In fact,
what does happen is that the principle of explosion is not unrestrictedly applied. The
contradictory propositions are still there, and it may happen that they are used in
some inferences, but they are not taken as true propositions.

15 See, for example, [31]: two [logics] stand out as having a solid philosophical-mathematical
justification. On the one hand, classical logic with its ontological basis and on the other hand
intuitionistic logic with its epistemic motivation.
16 In [30], p. 1, we read You ought to consider what Brouwers program was []. It consisted

in the investigation of mental mathematical construction as such, without reference to questions

regarding the nature of the constructed objects, such as whether these objects exist independently
of our knowledge of them.
1.4 Paraconsistency and the Nature of Logic 15

By means of a non-explosive negation and the consistency operator, , an LFI may

formally represent this scenario. The discussion about this point will be later resumed
in more detail. For now, we want to emphasize that the sketch of a paraconsistent logic
in which contradictions are epistemologically understood as conflicting evidence,
and not as a pair of contradictory true sentences, is inspired by an analysis of real
situations of reasoning in which contradictions occur. The notion of evidence is
weaker than truth in the sense that, if we know that is true, then there must be some
evidence for , but the fact that there is evidence for does not imply that is true.
A paraconsistent logic may thus be obtained analogously to the way intuitionistic
logic has been obtained.

1.5 Paraconsistency and the Nature of Contradictions

We now turn to a discussion of paraconsistency from the perspective of the problem

of the nature of contradictions. The latter is a very old philosophical topic that can
be traced back to the beginnings of philosophy in ancient Greece, and, as we have
just seen, is closely related to the issue of the nature of logic. There is an extensive
discussion and defense of the principle of non-contradiction in Aristotles Meta-
physics, book . According to Aristotle, PNC is the most certain of all principles
(Metaphysics 1005b19). It is a proposition that has no other propositions prior to it
(Posterior Analytics 72a5), and, as such, cannot be demonstrated from more basic
principles. Although Aristotle claims that PNC is, strictly speaking, indemonstrable,
he presents arguments in defense of it. This is not in fact a problem, since these argu-
ments may be thought of as elucidations, or informal explanations, of PNC, rather
than demonstrations in the strict sense. In Metaphysics  we find three versions of
PNC that correspond to the three aspects of logic mentioned above, ontological, epis-
temological and linguistic. We refer to them here respectively as PNC-O, PNC-E,
and PNC-L.
I. PNC-O (1005b 19-20)
Let us next state what this principle is. It is impossible for the same attribute at
once to belong and not to belong to the same thing and in the same relation.
II. PNC-E (1005b)
And if it is impossible for contrary attributes to belong at the same time to the
same subject (the usual qualifications must be added to this premiss also), and an
opinion which contradicts another is contrary to it, then clearly it is impossible
for the same man to suppose at the same time that the same thing is and is not
III. PNC-L (1011b10)
the most indisputable of all beliefs is that contradictory statements are not at the
same time true. If it is impossible at the same time to affirm and deny a thing truly,
it is also impossible for contraries to apply to a thing at the same time.
The point is that PCN-O is talking about objects and their properties, PCN-E about
beliefs, and PCN-L about propositions. ukasiewicz [33] calls these three versions
16 1 Contradiction and (in)Consistency

ontological, psychological and semantic.17 ukasiewicz strongly attacks Aristotles

defense of PNC, and claims that the psychological (i.e. epistemological) version is
simply false and that the ontological and the semantic (i.e., linguistic) versions have
not been proven at all. He ends the paper by saying that Aristotle might well have
himself felt the weaknesses of his argument, and so he announced his principle a
final axiom, an unassailable dogma [33], p. 509. We are not here going to analyze
Aristotles arguments, nor ukasiewiczs criticisms in detail. Rather, we are interested
in the following question: what should be the case in order to make true each one of
the formulations of PNC? We will see that the weaknesses of Aristotles arguments
have a lot to reveal about contradictions.
The basic idea of PNC-O corresponds to a theorem of first-order logic:
x(P x P x), i.e., the same property cannot both belong and not belong to
the same object. An object may have different properties at different moments of
time, or from two different perspectives, but obviously these cases do not qualify as
counterexamples for PNC (see Metaphysics, 1009b1 and 1010b10). PNC-O depends
on an ontological categorization of reality in terms of objects and properties. This
categorization has been central in philosophy and is present in logic since its begin-
nings.18 PNC-O has an ontological vein even if one is not sympathetic to the notion
of property. It is enough to change the object a has the property P to the object a
satisfies the predicate P. In any case, we are speaking in the broadest sense, which
includes objects in space-time as well as mathematical objects.
The linguistic formulation, here called PNC-L, although talking about language,
also has an ontological vein because of the link between reality and the notion of
truth. If there is a claim that is to a large extent uncontentious about truth, it is that
if a proposition (or any other truth-bearer) is true, it is reality that makes it true;
or, in other words, truth is grounded in reality. Understood in this way, PNC-O and
PNC-L collapse, the only difference being that the former depends on ontological
categorization in terms of objects and properties, while the latter depends on language
and an unqualified notion of truth. Note that Aristotle seems to conflate both, since in
the passage III quoted above PNC-O is the conclusion of an argument whose premise
is PNC-L.
A violation of PNC-O would be an object a and a property P such that a has and
does not have P. Hence, in order to show that PNC-O is true, one needs to show that
there can be no such object. This problem may be naturally divided into two parts,
one related to mathematics, the other related to empirical sciences. With respect to
the former, a proof of PNC-O would be tantamount to showing that mathematics is
consistent. But this cannot be proven, even with respect to arithmetic. With respect
to the latter, there is an extensive literature about the occurrence of contradictions
in empirical theories (see, for example, Chap. 5 of [35, 36]). However, to date, there
is no indication whether these contradictions are due to the nature of reality, or

17 This tripartite approach is also found in [34], where these three versions are called, respectively,
ontological, doxastic and semantic.
18 For example, the issue of particulars/universals, the Fregean distinction between object and func-

tion, and even Quines attacks to the notion of property.

1.5 Paraconsistency and the Nature of Contradictions 17

whether they belong to theories, which are nothing but attempts to give a model of
reality in order to predict its behavior. In other words, there is no clear indication,
far less a conclusive argument, that these contradictions are ontological and not only
epistemological.
The linguistic version of PNC is exactly the opposite of the dialetheist thesis as
it is presented by [7]:
A dialetheia is a sentence, , such that both it and its negation, , are true [] Dialetheism
is the view that there are dialetheias. [] dialetheism amounts to the claim that there are
true contradictions.

Thus, a proof of PNC-L would be tantamount to a disproof of dialetheism.

Although dialetheism has antecedents in the history of philosophy and is legitimate
from the philosophical point of view, it is a thesis that is far from being conclusively
established as true. Further, if we accept that every sentence says something about
something, a thesis that has not been rejected by logical analysis in terms of argu-
ments and functions, what makes PNC-O true would also make PNC-L true, and
vice-versa. Our conclusion is that neither PNC-O nor PNC-L has been conclusively
established as a true principle. And this is not because Aristotles arguments, or any
other philosophical arguments in defense of the two principles are not good. Rather,
the point is that this issue outstrips what can be done a priori by philosophy itself. It
seems to be useless for the philosopher to spend time trying to prove them.
Now we turn to PNC-E. As it stands, the principle says that the same person
cannot believe in two contradictory propositions. Here, the point is not how it could
be proved, because it really seems that there are sufficient reasons to suppose that
it has already been disproved. It is a fact that in various circumstances people have
contradictory beliefs. Even in the history of philosophy, as [33], p. 492 remarks, con-
tradictions have been asserted at the same time with full awareness. Indeed, since
there are philosophers, like Hegel and the contemporary dialetheists, that defend the
existence of contradictions in reality, this should be an adequate counterexample
to PNC-E. Furthermore, if we take a look at some contexts of reasoning, we will
find out that there are a number of situations in which one is justified in believing
both and . Sometimes we have simultaneous evidence for and , which
does not mean that we have to take both as true, but we may have to deal simul-
taneously with both propositions. Nevertheless, the problem we have at hand may
be put more precisely: PNC-E is somewhat naive and does not go to the core of
the problem. The relevant question is whether the contradictions we find in real sit-
uations of reasoningdatabases, paradoxes, scientific theoriesbelong to reality
properly speaking, or have their origin in thought and/or in the process of acquiring
knowledge.
Let us see which lessons may be taken from all of this. It is a fact that contradic-
tions appear in several contexts of reasoning. Any philosophical attempt to give a
conclusive answer to the question of whether there are contradictions that correctly
describe reality, is likely to be doomed to failure. However, the lack of such a conclu-
sive answer does not imply that it is not legitimate to devise a formal system in which
contradictions are interpreted as true. If there are some ontological contradictions
18 1 Contradiction and (in)Consistency

among the sentences that describe reality correctly, we are going to find some true
contradictions. But of course reality is not trivial, so we need a logic in which explo-
sion does not hold. Therefore, if contradictions are ontological, a justification for
paraconsistency is straightforward. Regarding epistemological contradictions, even
if some contradictions belong to reality, certainly at least some of the contradictions
we face are epistemological in the sense presented in Sect. 1.2. In general, con-
tradictions to be resolved at some later time, arising from conflicting information,
including the sometimes-contradictory results of scientific theories, may be taken
to be epistemological contradictions. It is perfectly legitimate, therefore, to devise
formal systems in which contradictions are understood either epistemologically or
ontologically. In the latter case, it may be the case that both and are true; in
the former, just take and to mean conflicting evidence about the truth-value
of . In both cases, explosion does not hold without restrictions; in both cases, the
development of paraconsistent logics is line with the very nature of logic.
A philosophical justification for paraconsistent logics, and in particular for the
LFIs, depends essentially on showing that they are more than mathematical struc-
tures with a language, a syntax, and a semantics, about which several technical
properties can be proved. Working on the technical properties of formal systems
helps us to understand various logical relations and properties of language and a
number of concepts that are philosophically relevant. However, in order to justify a
whole account of logical consequence it is necessary to show that such an account
is concerned with real situations of reasoning. From this perspective, given a formal
system, the key question is whether or not it provides an intuitive account of what
follows from what in real situations of reasoning. Depending on the answer given,
the logic at stake acquires a philosophical citizenship.
In an intuitive account, LFIs can be regarded as representing contexts of reasoning
in which contradictions occur because reality itself is contradictory, or as represen-
tations of contexts in which contradictions are provisional states that (at least in
principle) are going to be corrected later. What is important is that LFIs are able to
deal with contradictions, no matter whether they are understood epistemologically
or ontologically.

1.6 Contradiction, Consistency and Negation

A question that often arises with respect to paraconsistent logics regards the justifi-
cation for the rational acceptance of contradictory sentences. This question will have
influence on the rationale behind the semantics of LFIs to be presented in details in
Chaps. 2 and 3, and even in the alternative semantics for LFIs given in Chap. 6.
Contradictions, as we have argued, can be approached from an epistemic angle,
and it is not necessary to believe that there are real contradictions in order to devise
a paraconsistent logic. This section offers some remarks on three concepts central
to paraconsistency: contradiction, consistency and negation. The purpose is to show
1.6 Contradiction, Consistency and Negation 19

that these concepts have more than one meaning. This indicates several possibilities
for a philosophical appraisal of paraconsistency not yet explored.

1.6.1 On Contradiction

It is not surprising that Hegel is always recalled when it comes to contradictions.

Several passages can be found in his writings that seem to be in perfect opposition to
the principle of non-contradictionin his Science of Logic, for instance, we read that
Everything is inherently contradictory, and in the sense that this law in contrast to the
others expresses rather the truth and the essential nature of things. [37], p. 439. He
was perhaps the thinker who most emphatically addressed the issue, stating that both
thought and the world are contradictory. However, the meaning of contradiction in
Hegel is equivocal.
Explaining any concept of Hegels system in a few words, without appealing to
jargon, is not an easy task. It is perhaps appropriate to begin with a warning. We will
not become involved here in the details of the idiosyncratic Hegelian terminology,
but will simply try to make clear the essential points needed to grasp the ambiguity
in the meaning of contradiction in Hegel.
Kants analysis establishes an insurmountable gap between subjectivity and objec-
tivity. Scientific knowledge depends on our cognitive apparatus; we know nothing
about the objects of experience apart from the conditions by which they are given
to us. As an immediate consequence, unconditioned knowledge is not possible. In
opposition to Kant, Hegel believes that unconditioned knowledge not only should
but must be possible. A necessary condition for attaining it is to overcome the gap
between subject and object. In Hegelian terminology, nothing can be left outside
the absolute, otherwise the absolute would not really be absolute. The point is that
the truth must be the whole truth, the absolute truth, and hence must encompass the
whole of reality in its temporal development, together with subjectivity, in a unity.
But what could this unity be? We can safely say that it includes reality in all of its
constituent moments, opposing forces and conflicts, and collapses the distinction
between subject and object.19
The point is that in Hegel a contradiction also means a mere opposition that
is a result of conflicts between different things or moments of the same thing.
Hence, strictly speaking, this is not a contradiction in the sense of a violation of
the Aristotelian claim that the same attribute cannot at the same time belong and

19 The idea that there are opposing elements existing simultaneously in a whole is found in many

places in Hegels writings. The following passage from [38], Sect. 20, illustrates this interpretation
and the weaker sense of contradiction: The True is the whole. But the whole is nothing other
than the essence consummating itself through its development. Of the Absolute it must be said
that it is essentially a result, that only in the end is it what it truly is; and that precisely in this
consists its nature, viz. to be actual, subject, the spontaneous becoming of itself. Though it may
seem contradictory that the Absolute should be conceived essentially as a result, it needs little
pondering to set this show of contradiction in its true light .
20 1 Contradiction and (in)Consistency

not belong to the same subject in the same respect (Metaphysics 1005b19-21) and
expressed in first-order logic by the schema

x(P x P x)

for a given predicate (property) P.

We wish to call attention to the fact that the above interpretation, which identifies
a weaker notion of contradiction in Hegel, is closely related to the basic idea of
possible-translations semantics (and in particular to society semantics), introduced
in [39] and briefly treated in this Book in Sect. 6.8 of Chap. 6.
Indeed, the internal logic of a group of individuals is not necessarily the same as the
logic of individuals taken separately. There are different standards of rationality with
respect to individuals on the one hand and groups on the other. Non-contradictoriness,
for instance, when applied to groups, differs from that applied to individuals, precisely
because among the individuals (members of the group) there may be conflicting
beliefs, opinions, judgements, etc. Individuals very often act like contrary forces
within a society. The whole society, however, is subject to stricter standards of rigor.
As a very simple example, suppose a group of referees has to decide upon a selection
of papers to be published. Only some of them can be selected, and with respect to
each paper the whole group must take a position A or not A, but not both. Clearly, in
the case of two different referees that defend respectively positions and for a
certain paper, we may well say that there is a contradiction between them, but this
is clearly a weaker sense of contradiction. The example above shows that having
more than one notion of contradiction can be a useful tool for expressing a number
of features in different contexts of reasoning.

1.6.2 On Consistency

Turning now to the notion of consistency, it is not difficult to be convinced that

classical logic works with two different but equivalent notions of consistency, called
by [40], Sect. 24, simple and absolute consistency. Simple consistency is tantamount
to non-contradictoriness, while absolute consistency is tantamount to non-triviality.
It is clear that in classical logic, due to the validity of the principle of explosion, these
two notions are equivalent. Of course, in paraconsistent logics these two notions do
not coincide, since the point is precisely to allow some contradictions (i.e. simple
inconsistency) without triviality (maintaining absolute consistency). However, as has
been argued elsewhere (see e.g. [41]), in informal reasoning there are a number of
different notions of consistency.
First of all, notice that absolute consistency definitely does not match any informal
or intuitive notion of consistency. As pointed out earlier, there are situations where
we have to deal with contradictions, and this is perfectly rational and acceptable.
Triviality, on the contrary, is unacceptable in any situation. Furthermore, there are
intuitive notions of consistency, present in natural language, that do not correspond
1.6 Contradiction, Consistency and Negation 21

exactly to the notion of simple consistency. We sometimes use consistency in a sense

that is not necessarily related to or defined in terms of negation. In scientific research
we sometimes come across information that for some reason is suspicious. This does
not need to involve a contradiction. Simultaneous but non-conclusive evidence for
and indicates that both are suspicious. But there may be suspicious propositions
not involving any contradiction. The information conveyed by may be very unlikely,
or incoherent with respect to previous data represented by a set  of sentences, even
though there may be no formal contradiction derivable from and  together. It is
not uncommon to say, in such cases, that is inconsistent.20
Moral dilemmas are another typical situation where a person has to deal with
inconsistency, but not necessarily with a formal contradiction. A moral agent may
believe that (s)he has moral obligation to perform two actions, and , but cannot
do both, because they are mutually exclusive. We may well say that the belief of the
agent is inconsistent, in a very reasonable sense of inconsistency. Let us take as an
example the well-known dilemma, posed by [42], of the man in occupied France who
on the one hand wants to fight the Nazis but on the other must take care of his mother.
He believes that each alternative is a moral obligation, and that doing one implies
not doing the other. However, he may well believe that he has both obligations, while
not believing a formal contradiction in the strict sense, i.e. not believing that he
has and does not have the obligation to do something. In order to obtain a formal
contradiction, we need some other principles of deontic logic that the agent may be
not aware of.
These notions of consistency should be independent of model-theoretical and
proof-theoretical considerations. According to [43], it would be possible to treat a
notion of logical consistency as a primitive metalogical notion not reducible to seman-
tic consistency or syntactic consistency. However, it seems that his efforts were not
successful, as [44] points outFields attempts essentially amount to abbreviations
for statements about the consistency of some metatheoretical sentences. This shows
that it is not obvious how to treat consistency as independent from model-theoretic
or proof-theoretic means. It seems that a better approach is to treat the question more
geometrico, axiomatizing the notion of consistency as a primitive one, but making it
relate to negation and other logical connectives in an appropriate way, as it is done
in mbC and other LFIs (see [45]).

1.6.3 On Negation

A point that sometimes causes uneasiness at first sight is that logics of formal incon-
sistency have more than one negation. This is strange, one might say, because it
seems that a negation without the features of classical negation would not properly

20 In the logic mbC, to be studied in detail in Chap. 2 of this Book, the consistency of , represented

by , is not equivalent to non-contradictoriness of : implies ( ), but the converse does

not hold in general.
22 1 Contradiction and (in)Consistency

be a negation. Our reply is that this uneasiness is misplaced. It is quite possible that
this uneasy feeling may be caused by a realist conception of truth, connected to the
meaning of classical negation, even if one is not completely aware of it. However,
we can safely affirm that there are at least two senses of negation, one ontological
and one epistemic, corresponding to the two aspects of logic we tried to distinguish
in Sect. 1.2.
We do use, in informal reasoning, a negation that is weaker than the classical
one and that arguably has an epistemic character. It happens, for instance, when we
say not , but are not sure about how much we are denyingor, in other words,
when we are not completely sure that is consistent (or solid, well-established, etc.).
We write where means paraconsistent negation. On the other hand, suppose
that is not the case has already been confirmed, and we can safely say that it
has been well established. We now affirm not , but this time in a stronger sense,
employing a stronger negation. We are now making an assertion that intends to say
something in the framework of classical logic, and we write where means
classical negation. What occurs, in this case, is that in conclusively establishing
we have also established the consistency of . When we write we also mean that
we cannot have (a contradiction of the form is unacceptable, just as in
classical logic). On the other hand, in writing we still leave open the possibility
that the conclusion may be revised, i.e. that we are not completely convinced that
is not the case. Notice that it is not by chance that epistemic notions (such as
to be convinced, to establish conclusively, etc.) occur here. Thus it seems clear
that classical and paraconsistent negation have, respectively, an ontological and an
epistemic character. Of course, the distinction we have made in Sect. 1.4, namely,
that a contradiction may have an epistemic or an ontological character, also naturally
applies to the negation used to express the contradiction. Even if one admits that
negation is unique from an ontological viewpoint (a reasonable position, since there
should not be more than one ontology), from the epistemic point of view we have
at least one additional negation, precisely the negation that occurs in contradictions
that have an epistemic character. Furthermore, nothing prevents us from having still
more negations.
Now we may take consistency as a primitive operator, , which is justifiable once
we acknowledge that consistency does not coincide with non-triviality, that there is
more than one sense of consistency, and that consistency is not always intuitively
explained in terms of negation. We may also take as primitive a paraconsistent nega-
tion, which, from the epistemological point of view, antecedes classical negation.
With all of this in place, in mbC, a starting logic of formal inconsistency studied in
Chap. 2, a stronger negation is defined in the following way (details to be found in
Sect. 2.4):
:= ( ).
1.6 Contradiction, Consistency and Negation 23

Anticipating what is coming in the next chapter, it is interesting to note that it may
be proved in mbC that so defined has all the properties of classical negation.21
To sum up, suppose that we have some grounds, working with some data and
previous results, for thinking that is not the case. But we still have doubts with
respect to the previous results; in other words, we have not yet found that is
consistent and so write . Once we establish, by whatever means, that is indeed
not the case, we can now use the classical, stronger negation to affirm not , i.e.
. We have here a primitive notion of consistency plus a paraconsistent negation
that has an epistemic character, and, by combining them, we obtain through further
investigations, classical negation.
The fact that negation is an ambiguous notion that sometimes has a weaker sense
is not fully understood, and occasionally causes mistaken reactions against paracon-
sistency. Slater [46] argues against Priests paraconsistent system LP [22], but his
criticism supposedly holds for paraconsistent logics in general. He appeals to the
traditional notions of sentences being subcontraries but not contradictories, claiming
that paraconsistent negations are not real negations but a kind of subcontrariety
operator. We say that two sentences and are subcontraries when they cannot be
both false but can be both true. Of course, if paraconsistent negation were only a way
to express subcontrariness, it would hardly be of any philosophical significance.
In [47, 48], we find an effective logical and philosophical defense of paraconsistent
logics against Slaters arguments, but we think that there are still some remarks to
be made on this issue from the point of view of logics of formal inconsistency.
We have just seen above how we can go from a paraconsistent negation to a strong
one, a negation that has all the properties of classical negation. It seems to us that
it makes no sense to suppose that paraconsistent negation is a kind of subcontrary
operator, otherwise classical negation would not be definable from paraconsistent
negation. More precisely, it is hard to see how paraconsistent negation, as it is con-
ceived in mbC (to be detailed in Chap. 2) for example, could be a kind of subcontrary
sentential operator, in principle just allowing the attribution of the value 1 to and
, but at the same time being able to define classical negation based in the con-
junction of , and . We saw that in mbC consistency is primitive, not related
to negation, although we need the help of the consistency operator to define classical
negation. How could the property of subcontrariety between and be turned into
a classical contradictoriness between and if the role of were just to form
subcontraries?

21 With the help of this definition, and ( ) is also provable in mbC. Since

mbC is, by the very definition, an extension of classical propositional positive logic, classical logic
may be restored within mbC. Notice the difference between, on the one hand, restoring classical
consequence by means of a definition of a classical negation and, on the other, by means of a
DAT which, roughly speaking, states that each derivation within classical logic can be recovered
in mbC, by adding as additional hypothesis the consistency of certain formulas (see Sect. 2.4 and
Theorem 2.4.7). In the latter case, the point is the information that has to be available in order to
restore classical logic. The former shows that, in a certain sense, although the idea is to restrict
inferences valid classically, mbC is an extension of classical logic.
24 1 Contradiction and (in)Consistency

Even if we accept that is a subcontrary operator in a secondary sense, similar to

the sense one could say that intuitionistic negation is a kind of contrariety operator,
the only conclusion is that a paraconsistent negation that allows the attribution of 1
to and is not primarily a subcontrary operator, rather, it is primarily a negation.
It seems to us that Slater understands subcontrariety and contradictoriness as strictly
related to a notion of truth that complies with the properties of classical negation. But
the non-classical negation of mbC, as shown here, is better understood as expressing
a notion weaker than truth.

1.7 Varieties of Paraconsistency Involvement

If there is still a need to define the notion of paraconsistency at this point, we might
recall that paraconsistency is the study of contradictory but non-trivial theories, both
in its technical and philosophical aspects. A necessary condition for a logic to be
paraconsistent, as we have seen, is the invalidity of the principle of explosionbut
explosion may be rejected for several different reasons, related to different attitudes
with respect to the nature of contradictions. In fact, there are several types of para-
consistency in the market.
Beall and Restall, in [49], pp. 79ff, paraphrasing Quine, list four grades of
paraconsistency involvement. The first grade, that is the weakest, is simply dis-
satisfaction with explosion as a valid inference. This is the case of the relevantists,
whose motivation is to avoid the so-called paradoxes of material implication; one of
them is
( ),

that is nothing but a version of explosion. Indeed, relevant logics are a kind of para-
consistent logic. But there is no metaphysical discussion about the natures of con-
tradictions, since the point is that there may be no connection between the premises
and the conclusion. So, if one concludes that Descartes is French from Aristotle
is Greek and Aristotle is not Greek, since the premises have no connection with the
conclusion, i.e. are not relevant, the inference is thus rejected.
The second grade is the view according to which there are interesting, contradic-
tory but non trivial theories. They give as examples of such theories naive set theory
and naive truth theory. It is well known that these theories yield contradictions as
theorems because of Russells set and the Liar sentence. There are two remarks to
be made here. First, it seems that this position just leaves unnoticed the question
about the nature of contradictions. Second, they do not mention the important fact
that there are interesting, contradictory but non-trivial empirical theories.
1.7 Varieties of Paraconsistency Involvement 25

The third grade is a somewhat stronger position: some of the given inconsistent
but not trivial theories may be true, that is, it is possible that real contradictions exist.
This position has already been defended by N.C.A. da Costa.22
The fourth grade is dialetheism: the thesis that there exist inconsistent but non-
trivial theories that are true. These theories truly describe reality by means of some
contradictory sentences. This is the position defended by G. Priest and collaborators
in several places.
We do not agree that contradictions are possible, rather, we just admit that we
cannot prove that contradictions are impossible. The alternatives listed by Restall
and Beall are not exhaustive, and our position does not fit any of them. It seems to
us that a very important position is the one that accepts the existence of epistemic
contradictions, the view defended by us here. This view is not new, and has already
appeared in the history of philosophy. Maybe one of the best places to understand
the distinction between real and epistemic contradictions is in the response given by
Hegel to Kant, who took for granted that contradictions are a signal of error and, as
such, can only be produced by reason. Hegel disagrees:
[According to Kant] The stain of contradiction ought not to be in the essence of what is in
the world; it has to belong only to thinking reason, to the essence of the spirit [mind]. It
is not considered at all objectionable that the world as it appears shows contradictions to
the spirit [mind] that observes it; the way the world is for subjective spirit, for sensibility,
and for the understanding, is the world as it appears. [] [But the true significance of the
Kantian antinomies is that] Everything actual contains opposed determinations within it, and
in consequence the cognition and, more exactly, the comprehension of an object amounts
precisely to our becoming conscious of it as a concrete unity of opposed determinations
([51], pp. 91ff.).

It is not one hundred percent clear that what Hegel means by determination could
be represented by a unary predicate of a first-order language. Hence, it is also not
one hundred percent clear that opposed determinations would be a violation of the
principle of non-contradiction. Furthermore, it is plausible that the correct reading
is the one according to which contradictions in Hegel are due to the ongoing motion
of reality, rather than a violation of the principle of non-contradiction.23 What we
want to call attention to is how clearly Hegel explains the view, rejected by him, that
contradictions belong, or are yielded by, thinking reason.
Hegels description of Kants position is very close to what we understand by
epistemic contradictions: those produced by limitations in our cognitive apparatus,
flaws in the instruments used in experiments, the inability of the available theories
to deal with the data at handin a word, contradictions originated not in the world
itself, but in the way the world appears to the mind that observes it. Accordingly,

22 See [50], pp. 147 and 237: [A]t the macroscopic level, the experience seems to indicate that there

are no contradictions; however, at a microscopic level, there is nothing to prevent real contradictions.
[] [R]eal contradictions are not impossible, although there is nothing so far proving that they exist.
23 Aristotle had already been clear in saying that an object having different properties at different

moments of time, or from different perspectives, would not be a counterexample for the principle
of non-contradiction (see Metaphysics, 1009b1 and 1010b10). Actually, saying this one more time
is almost a platitude.
26 1 Contradiction and (in)Consistency

the rejection of the principle of explosion and the acceptance of some epistemic
contradictions by paraconsistent logics may be understood as a tool for dealing with
pairs of contradictory sentences that should not be taken as true but, on the other
hand, cannot be simply thrown away. This summarizes the position with respect to
paraconsistency that this book endorses, a position that cannot be accommodated
in any of the grades of paraconsistency mentioned by Restall and Beall, but that
nonetheless has a vast wealth of consequences still to be revealed.

References

1. Carnielli, Walter A., and Ablio Rodrigues. 2016. On the philosophy and mathematics of the
Logics of Formal Inconsistency. In New Directions in Paraconsistent Logic, ed. by Jean-Yves
Beziau, Mihir Chakraborty, and Soma Dutta, 5788. India: Springer. doi:10.1007/978-81-322-
2719-9_3.
2. Carnielli, Walter A., and Ablio Rodrigues. 2014. Towards a philosophical understanding of
the Logics of Formal Inconsistency. Submitted for publication.
3. Carnap, Rudolf, and Yehoshua Bar-Hillel. 1952. An outline of a theory of semantic information.
Research laboratory of electronics technical report 247, Massachusetts Institute of Technology.
Republished in [Bar-Hillel, Yehoshua, ed. 1964. Language and Information. London: Addison-
Wesley], pp. 221274.
4. Priest, Graham, Koji Tanaka, and Zach Weber. 2013. Paraconsistent logic. In The Stanford
Encyclopedia of Philosophy, ed. Edward N. Zalta, Fall 2013 edn.
5. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of formal inconsis-
tency. In Handbook of Philosophical Logic (2nd edn.), vol. 14, 193, Springer. doi:10.1007/
978-1-4020-6324-4_1.
6. Bobenrieth Miserda, Andrs. 2010. The origins of the use of the argument of trivialization in
the twentieth century. History and Philosophy of Logic 31(2): 111121.
7. Priest, Graham, and Francesco Berto. 2013. Dialetheism. In The Stanford Encyclopedia of
Philosophy, ed. Edward N. Zalta. Summer 2013 edn.
8. Priest, Graham. 2007. Paraconsistency and dialetheism. In Handbook of the History of Logic.
The Many Valued and Nonmonotonic Turn in Logic, vol. 8, ed. Dov M. Gabbay, and John H.
Woods, 129204, North Holland. doi:10.1016/S1874-5857(07)80006-9.
9. Harman, Gilbert. 1986. Change in view: Principles of reasoning. Massachussets: MIT Press.
10. Jaskowski, Stanisaw. 1948. Rachunek zdan dla systemw dedukcyjnych sprzecznych (in Pol-
ish). Studia Societatis Scientiarun TorunesisSectio A, I(5): 5777. English edition: Stanisaw,
Jaskowski. 1999. A propositional calculus for inconsistent deductive systems. Logic and Log-
ical Philosophy 7: 3556. Proceedings of the Stanisaw Jaskowskis Memorial Symposium,
held in Torun, Poland, July 1998.
11. Brunner, Andreas, and Walter A. Carnielli. 2005. Anti-intuitionism and paraconsistency. Jour-
nal of Applied Logic 3(1): 161184.
12. da Costa, Newton C.A. 1993. Sistemas formais inconsistentes (Inconsistent formal systems, in
Portuguese). Habilitation thesis, Universidade Federal do Paran, Curitiba, Brazil. Republished
by Editora UFPR, Curitiba, Brazil.
13. Gomes, Evandro L. 2013. Sobre a histria da paraconsistncia e a obra de da Costa: a
instaurao da lgica paraconsistente (On the history of paraconsistency and da Costas work:
The establishment of paraconsistent logic, in Portuguese). Ph.D. thesis, IFCH, State University
of Campinas. http://www.bibliotecadigital.unicamp.br/document/?code=000921849&opt=4&
lg=en_US.
14. Halldn, Sren. 1949. The Logic of Nonsense. Uppsala: Uppsala Universitets rsskrift.
15. Nelson, David. 1949. Constructible falsity. The Journal of Symbolic Logic 14(1): 1626.
References 27

16. Nelson, David. 1959. Negation and separation of concepts in constructive systems. In Construc-
tivity in Mathematics. Proceedings of the colloquium held at Amsterdam, 1957. Studies in Logic
and the Foundations of Mathematics, vol. 39, ed. Arend Heyting, 208225, North-Holland.
17. Almukdad, Ahmad, and David Nelson. 1984. Constructible falsity and inexact predicates. The
Journal of Symbolic Logic 49(1): 231233.
18. Kapsner, Andreas, David Miller, and Roy Dyckhoff. 2014. Report on Kalman Cohens 1954
Oxford Thesis on Alternative Systems of Logic. In Handbook of the 5th World Congress of
Paraconsistency.
19. Popper, Karl. 1963. Conjectures and Refutations. New York: Harper.
20. da Costa, Newton C.A. 1959. Observaes sobre o conceito de existncia em matemtica
(Remarks on the concept of existence in mathematics, in Portuguese). Anurio da Sociedade
Paranaense de Matemtica, 2: 1619.
21. Asenjo, Florencio G. 1966. A calculus for antinomies. Notre Dame Journal of Formal Logic
16(1): 103105.
22. Priest, Graham. 1979. The logic of paradox. Journal of Philosophical Logic 8(1): 219241.
23. Tugendhat, Ernst, and Ursula Wolf. 1989. Propedutica lgico-semntica. Editora Vozes,
Petrpolis, Brazil, 1996. Portuguese translation of Logischsemantische Propdeutik (in Ger-
man). Ditzingen: Reclam.
24. Chateaubriand, Oswaldo. 2001. Logical Forms, Part I: Truth and Description, vol. 34 of
Coleo CLE. Campinas: CLE-UNICAMP.
25. Arnauld, Antoine, and Pierre Nicole. 1996. Logic or the Art of Thinking. Cambridge: Cambridge
University Press.
26. Frege, Gottlob. 1879. Begriffsschrift: eine der arithmetischen nachgebildete Formelsprache des
reinen Denkens (in German). Halle. English edition: Gottlob, Frege. 1879. Begriffsschrift: a
formula language, modeled upon that of arithmetic, for pure thought. In [Heijenoort, Jean van,
ed. 1967. From Frege to Gdel; A Source Book in Mathematical Logic, 18791931. Harvard
University Press], pp. 182.
27. Frege, Gottlob. 1982. Grundgesetze der Arithmetik, Band I/II (in German). Verlag Hermann
Pohle, Jena, 1983. Partial translation of Band I by Montgomery Furth as The Basic Laws of
Arithmetic. Berkeley: University of California Press.
28. Frege, Gottlob. 1918. Der Gedanke. Eine logische Untersuchung (in German). Beitrge zur
Philosophie des deutschen Idealismus, I: 5877. The Frege Reader, pp. 325345. Trans. and
ed. Michael Beaney. Oxford: Blackwell Publishing, first edn.
29. Brouwer, Luitzen E.J. 1975. On the foundations of mathematics. Ph.D. thesis, 1907. In ed.
Arend Heyting, L.E.J. Brouwer Collected Works vol. I. North-Holland Publishing Company.
30. Heyting, Arend. 1956. Intuitionism: An introduction. London: North-Holland Publishing Com-
pany.
31. van Dalen, Dirk. 1986. Intuitionistic Logic. In Handbook of Philosophical Logic, ed. Dov M.
Gabbay, and Franz Guenthner, 225339., Volume III: Alternatives in Classical Logic, volume
166 of Synthese Library. Dordrecht: D. Reidel Publishing Company.
32. Raatikainen, Panu. 2004. Conceptions of truth in intuitionism. History and Philosophy of Logic
25: 131145.
33. ukasiewicz, Jan. 1910. ber den salz des widerspruchs bei Aristoteles (in German). Bullettin
International de lAcadmie des Sciences de Cracovie, 1538. Trans. and ed. Vernon Wedin.
1971. On the Principle of Contradiction in Aristotle. The Review of Metaphysics 24(3): 485
509.
34. Gottlieb, Paula. 2013. Aristotle on non-contradiction. In The Stanford Encyclopedia of Philos-
ophy, ed. Edward N. Zalta. Spring 2013 edn.
35. da Costa, Newton C.A., and Steven French. 2003. Science and Partial Truth: A Unitary
Approach to Models and Scientific Reasoning. London: Oxford University Press.
36. Meheus, Joke (ed.). 2002. Inconsistency in Science. Dordrecht: Kluwer Academic Publishers.
37. Hegel, Georg W.F. 1998. Wissenschaft der Logik (in German), vols. I to III. Schrag, Nrnberg,
181218131816. English edition: Science of Logic (trans: Arnold, V. Miller), Prometheus
Books.
28 1 Contradiction and (in)Consistency

38. Hegel, Georg W.F. 1977. Die Phnomenologie, and des Geistes (in German). Goebhardt,
Bamberg and Wrzburg, 1807. Translated to English by Arnold V. Miller as Phenomenology
of Spirit. Oxford University Press.
39. Carnielli, Walter A. Many-valued logics and plausible reasoning. In [Epstein, George, ed. 1990.
Proceedings of the Twentieth International Symposium on Multiple-Valued Logic, Charlotte,
NC, USA. The IEEE Computer Society Press], pp. 328335.
40. Hunter, Geoffrey. 1973. Metalogic. University of California Press.
41. Carnielli, Walter A. 2011. The single-minded pursuit of consistency and its weakness. Studia
Logica 97(1): 81100.
42. Sartre, Jean-Paul. 2007. Existentialism is a humanism (1946) (trans: Arlette, Elkaim-Sartre).
Yale University Press.
43. Field, Hartry. 1991. Metalogic and modality. Philosophical Studies 62(1): 122.
44. Wojtowicz, Krzysztof. 2001. Some remarks on Hartry Fields notion of logical consistency.
Logic and Logical Philosophy 9: 199212.
45. Bueno-Soler, Juliana, and Walter A. Carnielli. 2014. Experimenting with consistency. CLE
e-Prints, 14(3). http://www.cle.unicamp.br/e-prints/vol_14,n_3,2014.html.
46. Slater, Barry H. 1995. Paraconsistent logics? Journal of Philosophical Logic 24(4): 451454.
47. Bziau, Jean-Yves. Are paraconsistent negations negations? In [Carnielli, Walter A., Marcelo
E. Coniglio, and Itala M.L. DOttaviano, eds. 2002. Paraconsistency: The Logical Way to the
Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP 2000), Vol. 228
of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New York], pp. 465486.
48. Bziau, Jean-Yves. 2006. Paraconsistent logic! (a reply to Slater). Sorites, 1725.
49. Beall, J.C., and Greg Restall. 2006. Logical Pluralism. Oxford University Press.
50. da Costa, Newton C.A. 2008. Ensaio sobre os fundamentos da lgica (Essay on the foundations
of logic, in Portuguese. So Paulo: Hucitec.
51. Hegel, Georg W.F. 1817. Enzyklopdie der philosophischen Wissenschaften (in German). Hei-
delberg: Owald. Trans. and Ed. T.F. Geraets, W.A. Suchting and H.S. Harris. 1991. The
encyclopaedia logic. Hackett Publishing Company.
Chapter 2
A Basic Logic of Formal Inconsistency: mbC

In [1], the study of LFIs starts with mbC, which is basic in the following sense:
it starts with positive classical logic (which is assumed as a reasonable basisin
Chap. 5, some LFIs will be studied which are based on logics other than positive
classical logic) and has a negation and a consistency operator which are added with
minimal properties to satisfy the definition of LFIs.

2.1 Introducing mbC

Since the first chapters of the book are exclusively devoted to propositional logics,
some notation which will be used throughout the rest of the book will now be defined.
From this chapter on, and in the rest of the book, the symbol  is used to mean the
end of a definition, while  is used to indicate the end of a proof, being omitted when
a proof is not given explicitly.

Definition 2.1.1 (Propositional signatures) A propositional signature is a set  of

symbols called connectives, together with the information concerning the arity of
each connective. 

Notation 2.1.2 For the entirety of the book the following symbols will be used
for logical connectives (the intended meaning and the arity of each connective are
included in the list below): (conjunction, binary); (disjunction, binary);
(implication, binary); (weak negation, unary); (consistency operator, unary);
(inconsistency operator, unary); (strong negation, unary); (bottom formula,
0-ary, i.e., a propositional constant).

Springer International Publishing Switzerland 2016 29

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_2
30 2 A Basic Logic of Formal Inconsistency: mbC

Definition 2.1.3 Consider the following propositional signatures that will be used
thereafter:
 = {, , , , };
 = {, , , , };
0 = {, , , };
+ = {, , };
c = {, , , };
1 = {, , , , };
2 = {, , , , };
 = {, , , }.
Let Var = { p1 , p2 , . . .} be a denumerable set of propositional variables (will be
fixed henceforth), and let  be any propositional signature. The propositional lan-
guage generated by  from Var will be denoted by L . 

In this book we will deal exclusively with so-called Tarskian logics (see, for
instance, [2]):

Definition 2.1.4 (Tarskian Logic) A logic L defined over a language L, which has
a consequence relation , is Tarskian if it satisfies the following properties, for every
  {} L:
(i) if  then   ;
(ii) if   and   then   ;
(iii) if   and   for every  then   .
A logic satisfying item (ii) above is called monotonic. A logic L is said to be finitary
if it satisfies the following:
(iv) if   then there exists a finite subset 0 of  such that 0  .
Finally, a L defined over a propositional language L generated by a signature from a
set of propositional variables is called structural if it satisfies the following property:
(v) if   then []  (), for every substitution of formulas for variables.1
A propositional logic is standard if it is Tarskian, finitary and structural (see [2]). 

From now on, a logic L will be represented by a pair L = L,  , where L

and  denote the language and the consequence relation of L , respectively. If L
is generated by a propositional signature  from Var, that is, L = L then we will
write L = ,  .

Notation 2.1.5 Let L = L,  be a logic. Let be a formula in L and let X 1 . . . X n

(for n 1) be a finite sequence such that each X i is either a set of formulas in L or
a formula in L. Then, as usual, X 1 , , X n  will stand for X 1 . . . X n 

1 In this book the following standard notation will be adopted: given a function f and a subset A of
its domain, f [A] will denote the set { f (a) : a A}.
2.1 Introducing mbC 31

where, for each i, X i is X i , if X i is a set of formulas, or X i is {X i }, if X i is a formula.

Thus, for instance, if  and  are sets of formulas and {, , 1 , . . . , n } is a set of
formulas then

,  ;  ; ,   ; 1 , . . . , n 

will stand for

 {}  ; {}  ;    ; {1 , . . . , n } 

respectively.

The main notion of this book can now be defined rigorously: the Logics of Formal
Inconsistency.

Definition 2.1.6 A Tarskian logic L is paraconsistent if it has a (primitive or

defined) negation such that , L for some formulas and in the language
of L . 

If L has a deductive implication , in the sense that it satisfies the Deduction

meta-theorem DMT (see Proposition 2.1.14(i) below), then L is paraconsistent iff
the schema formula ( ) is not valid. That is, the explosion law is not
valid in L with respect to the negation . In other words, the negation is not
explosive.
The idea behind LFIs is to have a paraconsistent logic in which the explosion
law is allowed in a local or controlled way. This is attained by the existence of a
set of formulas ( p), depending exclusively on a single propositional variable p,
such that the set (), together with a contradiction {, } is explosive or logically
trivial, that is: (), , L for every and . A logic satisfying this property
is called gently explosive in [1, 3]. Of course, it must be also required that ()
together with not be trivial, as well as the combination of () with (otherwise,
the principle of gently explosiveness will be redundant). As we shall see, there exists
three ways to introduce LFIs. The original one (proposed in [3]) is the following:

Definition 2.1.7 Let L = ,  be a standard logic. Assume that the signature 

of L contains a negation , and let ( p) be a nonempty set of formulas depend-
ing exactly on the propositional variable p. Accordingly, L is a Logic of Formal
Inconsistency (an LFI, for short) (with respect to and ( p)) if the following holds
(here, () = {() : ( p) ( p)}):
(i) ,  for some and , i.e., L is not explosive w.r.t. ;
(ii) there are two formulas and such that
(ii.a) (),  ;
(ii.b) (),  ;
32 2 A Basic Logic of Formal Inconsistency: mbC

(iii) (), ,  for every and .



When ( p) is a singleton, its element will be denoted by p, where is the con-

sistency operator. Item (i) of the above definition states that an LFI is, by definition,
non-explosive (w.r.t. ). Because of item (iii), L is said to be gently explosive w.r.t.
and ( p).
Observe that clauses (i) and (ii) are existential, while (iii) is universal. The pair of
witnesses required to satisfy (i) and (ii) are possibly different, but (ii) is composed
by two clauses, both of them being satisfied by the same pair (, ). This could be
weakened, obtaining the following weaker notion of LFIs:

Definition 2.1.8 Let L = ,  be a standard logic. Assume that the signature 

of L contains a negation , and let ( p) be a nonempty set of formulas depending
exactly on the propositional variable p. Then L is a weak LFI (with respect to
and ( p)) if the following holds:
(i) ,  for some and ;
(ii) (),  for some and ;
(iii) (),  for some and ; and
(iv) (), ,  for every and .


Observe that any LFI is a weak LFI, but the converse is not necessarily true. The
notion of weak LFI was proposed as an alternative to the original one in [46], since
it is more uniform: the sets {, }, {, } and {, } are not always deductively
trivial, but the set {, , } is always deductively trivial.2 Finally, a stronger notion
of LFIs (which is also more uniform than the original definition) could be proposed:

Definition 2.1.9 Let L = ,  be a standard logic. Assume that the signature 

of L contains a negation , and let ( p) be a nonempty set of formulas depending
exactly on the propositional variable p. Then L is a strong LFI (with respect to
and ( p)) if the following holds:
(i) there are two formulas and such that
(i.a) ,  ;
(i.b) (),  ;
(i.c) (),  ; and
(ii) (), ,  for every and .


2 It should be observed that the weak LFIs investigated in the three references mentioned above are

also strong LFIs in the sense of Definition 2.1.9.

2.1 Introducing mbC 33

Remark 2.1.10 Clearly, any strong LFI is an LFI, but the converse is not necessarily
true. It should be observed that all the LFIs introduced in the previous literature are
strong LFIs. Moreover, if L is a propositional logic then L is a strong LFI whenever
the following holds:
(i) if p and q are two different propositional variables then
(i.a) p, p  q;
(i.b) ( p), p  q;
(i.c) ( p), p  q; and
(ii) (), ,  for every and .
Because of its simplicity, conditions (i) and (ii) above will be used along the rest of
the book in order to prove that a given logic is a strong LFI.

The general definition of LFIs encompasses a wide range of paraconsistent logics.

In order to formally express the properties of consistency, any logic featuring a
consistency connective must present a set of logical axiom schemas and semantic
rules governing this connective. Along these lines, in [1], a fundamental propositional
LFI known as mbC was first introduced. Starting from positive classical logic plus
tertium non datur ( ), mbC is intended to comply with the above definition
in a minimal way: an axiom schema called (bc1) is added solely to describe the
expected behavior of the consistency operator , namely, the gentle explosion law
(see Definition 2.1.12). In what follows, this logic will be described in its original
language along with the statement of soundness and completeness theorems with
respect to paraconsistent valuations.

Definition 2.1.11 (Formula Complexity) The complexity of a given formula

L , denoted by l(), is recursively defined as follows:
1. If = p, where p Var, then l() = 1;
2. If = , then l() = l() + 1;
3. If = , then l() = l() + 2;
4. If = #, where # {, , }, then l() = l() + l() + 1.


Definition 2.1.12 (mbC) The logic mbC is defined over the language L by the
Hilbert calculus:

Axiom schemas:
 
(Ax1)
      
(Ax2)
  
(Ax3)
 
(Ax4)
34 2 A Basic Logic of Formal Inconsistency: mbC
 
(Ax5)
 
(Ax6)
 
(Ax7)
    
( ) ( ) (Ax8)
 
(Ax9)
(Ax10)
  
(bc1)

Inference rule:

(MP)

Axiom (bc1) is called the gentle explosion law. In Theorem 2.3.2, we shall prove
that, as expected, the logic mbC is an LFI.

Definition 2.1.13 Let  {} L be a set of formulas. A derivation of from

 in mbC is a finite sequence 1 n of formulas in L such that n is and, for
every 1 i n, the following holds:
1. i is an instance of an axiom schema of mbC, or
2. i , or
3. there exist j, k < i such that k = j i (and so i follows from j and k
by MP).
We say that is derivable from  in mbC, denoted by  mbC , if there exists a
derivation of from  in mbC. 

Observe that (Ax1)(Ax9) plus MP constitute a Hilbert calculus over the signature
+ = {, , } for positive classical propositional logic CPL+ (the negation-free
fragment of classical propositional logic CPL, see Definition 2.4.3), which is in fact
the basis for mbC and its extensions.
The following meta-theorems of mbC will prove to be quite useful throughout
the entirety of the book.

Proposition 2.1.14 The calculus mbC satisfies the following properties:

(i) , mbC iff  mbC (Deduction meta-theorem, DMT).
(ii) If , mbC and , mbC then , mbC .
(iii) If , mbC and , mbC then  mbC (Proof-by-cases).
2.1 Introducing mbC 35

Proof (i) It is well known that, in order to prove that a Hilbert calculus satisfies
DMT, it suffices to derive axioms (Ax1) and (Ax2), while MP must be the unique
inference rule (see, for instance, [7]).
(ii) Suppose , mbC and , mbC . By DMT, both  mbC and
 mbC . By axiom (Ax8), mbC ( ) (( ) (( ) ))
and so  mbC ( ) by MP twice. Therefore , mbC , by MP.
(iii) This is a consequence of item (ii) and the fact that is a theorem of
mbC. 

2.2 A Valuation Semantics for mbC

In [8, 9], N.C.A.

 da  Costa and E. H. Alves proposed an original valuation semantics
for C1 over 0, 1 . A key feature of these valuations is that, as expected, they are
defined as 2-valued CPL+ -valuations with respect to the binary connectives (con-
junction, disjunction and implication). However, the paraconsistent negation has a
non-deterministic behavior w.r.t. this semantics: in general, if one of such valuations
assigns the value 1 to a formula , then the formula can receive either the value
0 or the value 1 (but not both) under the same valuation. That is: the truth-value of
does not uniquely determine the truth-value of .
This kind of semantics (sometimes called bivaluations) were generalized to sev-
eral LFIs in [1]. Based on that approach, in this section the logic  mbC will be
semantically characterized by a suitable valuation semantics over 0, 1 . The same
will be done for several other LFIs in the next chapters of the book.
 
Definition 2.2.1 (Valuations for mbC) A function v : L 0, 1 is a valuation
for mbC, or an mbC-valuation, if it satisfies the following clauses:
(vAnd) v( ) = 1 v() = 1 and v() = 1
(vOr) v( ) = 1 v() = 1 or v() = 1
(vImp) v( ) = 1 v() = 0 or v() = 1
(vNeg) v() = 0 = v() = 1
(vCon) v() = 1 = v() = 0 or v() = 0.
The set of all such valuations will be designated by V mbC . 

It should be observed that each mbC-valuation v assigns an unique truth-value

(0 or 1) to each formula of mbC. However, because of clauses (vNeg) and (vCon),
the value v(#) is not necessarily
 determined by the value v() of the immediate
subformula , for # , .
For every  {} L , the following semantical consequence relation w.r.t.
the set V mbC of mbC-valuations can be naturally defined:  mbC iff, for every
v V mbC , if v( ) = 1 for every  then v() = 1. The set V mbC constitutes a
sound and complete semantics for the logic mbC, as it will be proved below.
36 2 A Basic Logic of Formal Inconsistency: mbC

Theorem 2.2.2 (Soundness) For every  {} L :

 mbC =  mbC .

Proof The first step is to show the following:

Facts: Let v be an mbC-valuation.
(1) If is an instance of a mbC axiom schema, then v( ) = 1.
(2) If and are formulas such that v() = v( ) = 1, then v() = 1.
In order to prove (1), each mbC axiom schema must be checked. Observe that
the clauses for the connectives , and are the usual ones which character-
ize the truth-tables for these connectives in classical logic. Thus, axioms Ax1Ax9
(corresponding to positive classical logic) are sound w.r.t. this valuation seman-
tics. Concerning axiom Ax10, let be a formula. If v() = 1 then v( ) = 1,
by (vOr). Otherwise, if v() = 0then v() = 1, by(vNeg), whence v( ) = 1,
 
by (vOr). Finally, let = be an instance of axiom bc1. If
v() = 0 then v( ) = 1, by (vImp). Otherwise, if v() = 1 then either v() = 0
or v() = 0. In both cases it is easy to see that v( ) = 1, by (vImp).
Item (2) is an easy consequence of the clause (vImp) from Definition 2.2.1.
Once these facts are proved, the rest of the proof follows by induction on the length
n of a derivation 1 . . . n = in mbC of from . Indeed, it will be proven by
induction on n that, for any mbC-valuation v such that v[] {1}, v(i ) = 1 for
every 1 i n. In particular, v() = 1, showing that  mbC as required.
Thus, if n = 1, then 1 is either an instance of an axiom schema of mbC (and so
the result follows by Facts(1)), or 1  (and so the result follows by hypothesis).
Suppose now that the result holds for every formula admitting a derivation in mbC
from  of length k n (induction hypothesis), and suppose that 1 . . . n+1 = is a
derivation in mbC of from  with length n + 1. By induction hypothesis, v(i ) = 1
for 1 i n (as each i is derived from  in mbC with a derivation of length i).
If n+1  or n+1 is an instance of an mbC axiom schema, the proof is as above.
Otherwise, there exist j, k n such that k = j n+1 and n+1 is obtained from
j and k by MP. By Facts(2), v(n+1 ) = 1. This completes the proof. 

The proof of completeness needs some definitions and results. Recall the notion
of Tarskian Logic from Definition 2.1.4.

Definition 2.2.3 For a given Tarskian logic L over the language L, let  {} L.
The set  is maximal non-trivial with respect to in L if  L but , L
for any
/ . 

Definition 2.2.4 Let L be a Tarskian logic. A set of formulas  is closed in L , or a

closed theory of L , if the following holds for every formula :  L iff .

Lemma 2.2.5 Any set of formulas maximal non-trivial with respect to in L is

closed, provided that L is Tarskian.
2.2 A Valuation Semantics for mbC 37

Proof Let  be a set of formulas maximal non-trivial with respect to in L . If 

then  L , as L is Tarskian. Conversely, if  L , suppose that / . Then,
by Definition 2.2.3, , L . However, given that L is Tarskian, it follows that
 L , contradicting the hypothesis that  is maximal non-trivial with respect to
in L . Then  and so  is a closed theory. 
Now consider the following classical result:
Theorem 2.2.6 (Lindenbaum-os) Let L be a Tarskian and finitary logic over the
language L. Let  {} L be such that  L . There exists then a set  such
that   L with  maximal non-trivial with respect to in L .
Proof We will adapt the proof in [2] (Theorem 22.2). By the well-ordering principle,3
assume that the set L is well-ordered as a transfinite sequence ( )< , where is
an ordinal. By transfinite recursion, a transfinite and increasing sequence of theories
( )< is defined as follows: 0 = , and for every < ,

 if = + 1 and  , L
 =  { } if = + 1 and  , L

<  if is a limit ordinal.


Then  = <  satisfies the requirements. Indeed, observe firstly that  . By
transfinite induction, it is easy to prove that  L , for every < : if = 0 then
 L , by hypothesis. Assuming that  L for every < < , suppose
that = + 1 < . Then  L , by induction hypothesis, and so  L , by
definition of  . If is a limit ordinal, suppose that  L . By finitariness of L ,
there exists a finite subset  fin of  such that  fin L . But   if <
and so  f in  for some < . This means that  L for some < ,
contradicting the induction hypothesis.
By a similar argument,  L is proved. Suppose now that / . Then =
for some < and so / +1 , by definition of . By construction of +1 , it
follows that  , L and so, by monotonicity of L , it follows that , L .
This shows that  is maximal non-trivial with respect to in L . 
Every logic L defined by a Hilbert calculus, where the inference rules are fini-
tary, is Tarskian and finitary, and so Theorem 2.2.6 holds for L . In particular,
Theorem 2.2.6 holds for mbC.
Theorem 2.2.7 Let  {} L , with  maximal non-trivial with respect to in
mbC. The mapping v : L 0, 1 defined by:

v() = 1 

for all L is a valuation for mbC.

3 Since it is well-known, it is equivalent to the Axiom of Choice.

38 2 A Basic Logic of Formal Inconsistency: mbC

Proof It will be proved that v satisfies all the clauses of Definition 2.2.1.
1. Assume that v( ) = 1. Then . Suppose that neither  nor
. Then , mbC and , mbC . So , mbC , by Proposi-
tion 2.1.14(ii). But then  mbC , which is a contradiction. Thus, either  or
 and so either v() = 1 or v() = 1. Conversely, suppose that either v() = 1
or v() = 1. Thus, either  or . Suppose that . As ( ) ,
by axiom (Ax6) and by Lemma 2.2.5, it follows that , by MP. From
this, v( ) = 1. Analogously, if  then v( ) = 1 (now by using axiom
(Ax7)). This shows that v satisfies clause (vOr) of Definition 2.2.1.
2. Assume that v( ) = 1. Then . As ( ) , by axiom
(Ax4) and Lemma 2.2.5, it follows that  by MP. From this, v() = 1. Analo-
gously, v() = 1 is proved by axiom (Ax5). Conversely,
 suppose that v() = 1 and
 
v() = 1. Then  and . But , by axiom (Ax3)
and by Lemma 2.2.5. Then, by applying MP twice, it follows that  and so
v( ) = 1. This shows that v satisfies clause (vAnd) of Definition 2.2.1.
3. Suppose that v( ) = 1. Then . If , then  by MP and
Lemma 2.2.5. Then v() = 1 implies that v() = 1. This shows that either v() = 0
or v() = 1. Conversely, suppose that either v() = 0 or v() = 1. Then either
/  or . If , then  by axiom (Ax1), MP and Lemma 2.2.5.
So v( ) = 1. Now, if /  then, by the maximality of , it follows that
, mbC . Suppose, by contradiction, that / . Then, again by the max-
imality of , it follows that , mbC . Hence , ( ) mbC , by
Proposition 2.1.14(ii). But then  mbC by axiom (Ax9), which leads to a con-
tradiction. Therefore  and so v( ) = 1, showing that v satisfies
clause (vImp) of Definition 2.2.1.
4. Suppose that v() = 0 and, by contradiction, that also v() = 0. Then /
and / . As  is maximal, it follows that , mbC and , mbC . By
Proposition 2.1.14(iii),  mbC , which is a contradiction. Therefore v() = 0
implies that v() = 1, and so v satisfies clause (vNeg) of Definition 2.2.1.
5. Suppose that v() = 1 and, by contradiction, that both v() = 1 and v() = 1.
Then  and both  and . Thus, by axiom (bc1), MP twice and
Lemma 2.2.5,  for every . In particular, , which is a contradiction.
Therefore: v() = 1 implies that either v() = 0 or v() = 0. This means that v
satisfies clause (vCon) of Definition 2.2.1. 

Theorem 2.2.8 (Completeness of mbC w.r.t. valuations) For every  {} L :

 mbC =  mbC

Proof Suppose  mbC and let  be a maximal non-trivial set with respect to
in mbC extending  (using Theorem 2.2.6). By Theorem 2.2.7, there is an mbC-
valuation v, such that v[] {1} (as  ) but v() = 0 (as
/ ). Therefore,
 mbC and the theorem follows by contraposition. 
2.3 Applications of mbC-Valuations 39

2.3 Applications of mbC-Valuations

From the soundness and completeness theorems for mbC w.r.t. valuations proved
in the previous section, some properties of mbC can now be stated semantically,
instead of using the associated Hilbert calculus.
We begin by observing that mbC-valuations can be used to construct truth-tables
which exhibit a non-deterministic character.4 Indeed, the clauses for mbC-valuations
corresponding to the binary connectives , and   (see Definition 2.2.1) define
the usual truth-tables for these connectives over 0, 1 . On the other hand, the para-
consistent negation defines the following diagram:

1 1 v1
0 v2
0 1 v3

This means that there are three mbC-valuations (or scenarios) concerning a propo-
sition and its paraconsistent negation , namely v1 , v2 and v3 . According to
v1 , v1 () = v1 () = 1. According to v2 , v2 () = 1 but v2 () = 0. Finally, in
the third scenario v3 , v3 () = 0 and v3 () = 1. Observe that the fourth scenario,
namely v4 () = v4 () = 0, is not allowed for mbC-valuations, in virtue of the
clause (vNeg): and can be both simultaneously true, but they cannot be simul-
taneously false. This means that mbC is paraconsistent but not paracomplete.5
Concerning the other non-classical connective, the consistency operator , it can
be better understood in terms of the formulas and instead of analyzing or
alone:

Table 2.1

1 0 v1
1 1 v2
0
0 v3
1 v4
0 1
0 v5

4 However, such tables do not correspond to non-deterministic matrices in the sense of Avron and
Lev (see [10, 11]). The relationship between valuations for LFIs and non-deterministic matrices
will be analyzed in Chap. 6.
5 There are logics which are simultaneously paraconsistent and paracomplete, that is, that allow

the fourth scenario in which and are both false. Logics of this kind are frequently called
paranormal by the literature. One example of paranormality is a tetravalent modal logic that can be
associated with Monteiros tetravalent modal algebras, see [12]. This example will be analyzed in
Chap. 5.
40 2 A Basic Logic of Formal Inconsistency: mbC

Observe that, if v() = v() = 1, then v() is forced to be 0. Otherwise, if

v() = v() then the truth-value v() of is arbitrary in mbC.

Remark 2.3.1 Diagrams as the one displayed in Table  2.1 can be naturally associ-
ated to the non-deterministic valuation semantic over 0, 1 for mbC introduced in
Definition 2.2.1. This idea was originally proposed by da Costa  and Alves in [9],
associated to the non-deterministic valuation semantic over 0, 1 for C1 defined
there. It was proved by Fidel (see Theorem 7, p. 627 in [9]) that such diagrams,
called quasi-matrices by Alves in [13],6 provide a decision procedure for testing
tautologies in C1 . It is easy to see that the same holds for mbC and for most of the
LFIs to be analized in this book.

By using the soundness and completeness theorems for mbC w.r.t. valuations, it
is easy to prove the following:

Theorem 2.3.2 Let ( p) = { p}, for a propositional variable p. Then the logic
mbC is a strong LFI (w.r.t. and ( p)), according to Definition 2.1.9.

Proof Assume that p and q are two different propositional variables. By considering
the valuation v1 of Table 2.1 and taking v1 (q) = 0, it follows that p, p mbC q and
clause (i.a) of Remark 2.1.10 thusly is satisfied. Now considering valuation v2 and
taking v2 (q) = 0, we show that p, p mbC q and clause (i.b) of Remark 2.1.10 is sat-
isfied. By considering valuation v4 such that v4 (q) = 0, it follows that p, p mbC q
and clause (i.c) of Remark 2.1.10 is satisfied. Finally, by means of the same table, it
is clear that for no valuation it is the case that , and are simultaneously true.
Thus, clause (ii) of Remark 2.1.10 is satisfied. This shows that mbC is a strong LFI
w.r.t. and . 

The dependency of on {, }, which was highlighted above, explains why

the complexity l() of is defined as l() + 2. That is, the complexity of
is strictly greater than the complexity of and (recall Definition 2.1.11). By
combining these non-deterministic truth-tables with the (deterministic) truth-table
of conjunction, we can compare the consistency with the non-contradiction
( ) on the one hand, and the inconsistency (or non-consistency) with
the contradiction on the other. This produces eight possible-scenarios (or
mbC-valuations), which are depicted below.

6 Itshould be observed that Ivlev (see [14]) and other authors use the term quasi-matrices to refer
to non-deterministic matrices in the sense of of A. Avron and I. Lev.
2.3 Applications of mbC-Valuations 41

( )
1 1 0 1 1 1 v1
0 v2
0 1 1 0 1 v3
0 0 1 v4
0 1 0 1 v5
0 1 1 1 0 1 v6
0 0 1 v7
0 1 0 1 v8

From the table above, and by the adequacy of mbC w.r.t. valuations, the following
becomes clear:
Proposition 2.3.3 In mbC the following holds:
(1) mbC but mbC ;
(2) mbC ( ) but ( ) mbC .
Proof (1) By inspection of the table above, it follows that mbC but
mbC (because of v3 , v5 , v6 and v8 ).
(2) Looking at the table above, it follows that mbC ( ) but
( ) mbC (because of v1 , v5 and v8 ). 
Item (1) of Proposition 2.3.3 shows that, in mbC, the notion of contradiction is
strictly stronger than the notion of inconsistency (or non-consistency). In other words,
in mbC every contradictory formula is inconsistent, but the converse is not always the
case. By its turn, item (2) establishes that the notion of consistency is strictly stronger
than the notion of non-contradiction in mbC. That is, every consistent formula is
non-contradictory, but the converse does not hold in general.
Another simple example of application of mbC-valuations is the following:
Proposition 2.3.4 In mbC the following holds:
(1) mbC but mbC ;
(2) , mbC .
Proof Consider the following diagram:

1 1 1 1 1 v1
0 1 1 v2
1 0 1 1 0 v3
0 1 1 v4
0 1 1 1 1 v5
0 0 1 0 0 v6

Clearly, mbC but mbC , because of valuation

v3 . However, v3 () = v3 () = 1, so v3 () = 0. Accordingly, , mbC
given that v3 , the unique countermodel for the inference mbC
, does not satisfy the premises: v3 [{, }]  {1}. 
42 2 A Basic Logic of Formal Inconsistency: mbC

Finally, the following example, which comes from the same vein, shows that
several contraposition rules for implication do not hold when the paraconsistent
negation is taken into account:

Proposition 2.3.5 In mbC the following holds:

(1) mbC but , mbC ;
(2) mbC but , mbC ;
(3) mbC but , mbC ;
(4) mbC but , mbC .

Proof Items (1) and (2). Consider the following diagram:

1 1 1 1 1 1 1 1 v1
0 1 1 0 1 v2
0 1 1 0 1 0 v3
0 1 1 0 0 v4
1 0 1 1 0 1 1 1 v5
0 1 0 0 1 1 v6
0 1 1 1 1 1 1 1 v7
0 1 1 1 1 v8
0 0 1 1 1 1 1 1 v9

Because of valuation v3 , mbC and mbC .

But v3 () = v3 () = 1, so v3 () = 0. Using an argument similar to the proof
found in Proposition 2.3.4, it follows that , mbC and ,
mbC .
Items (3) and (4). Consider the following diagram:

1 1 1 1 1 1 1 1 v1
0 1 1 0 1 v2
0 1 0 1 1 1 v3
0 0 1 1 1 v4
1 0 1 1 0 1 1 1 v5
0 1 1 1 1 1 v6
0 1 1 1 1 0 1 0 v7
0 1 1 0 0 v8
0 0 1 1 0 0 1 0 v9

Because of valuation v7 , mbC and mbC .

As v7 () = v7 () = 1, v7 () = 0. Therefore, , mbC and
, mbC . 
2.3 Applications of mbC-Valuations 43

Finally, valuation semantics allows one to clearly see that the logic mbC is not
self-extensional in Wjcickis sense (see [15]). We say that a propositional (Tarskian)
logic L satisfies weak replacement if the following holds: given formulas i and i
(for 1 i n) such that 1 1 , , n n , then (1 , . . . , n ) (1 , . . . , n )
for every formula ( p1 , . . . , pn ). Here, is an abbreviation for L and
L . A logic is said to be self-extensional if it satisfies weak replacement.
It is easy to comprehend, by using valuations, that mbC does not satisfy weak
replacement and so it is not self-extensional. Indeed, from , it does not follow
in general that # #, for # {, }. For instance, ( p1 p2 ) ( p2 p1 ). How-
ever, neither ( p1 p2 ) ( p2 p1 ) nor ( p1 p2 ) ( p2 p1 ): it is enough
to consider a mbC-valuation v such that v( p1 p2 ) = 1, v(( p1 p2 )) = 1 (and
so v(( p1 p2 )) = 0), but v(( p2 p1 )) = 0 and v(( p2 p1 )) = 1.
Moreover, since mbC and several of its extensions do not admit a non-trivial
logical consequence, they are not algebraizable even in the wide framework of Blok
and Pigozzi (see [1620]). The question of algebraizability of extensions of mbC
will be analyzed in Chaps. 4 and 6.

2.4 Recovering Classical Logic Inside mbC

The laws governing the operator of mbC define a paraconsistent negation, making
it weaker than the classical negation (which will be represented from now on by the
symbol ). In order to clarify the relationship between mbC and classical proposi-
tional logic (CPL), the notion of translations and conservative translations between
logics, introduced in [21], will be used. In what follows, if is a mapping defined on
formulas and  is a set of formulas, then  = { : }.
def

Definition 2.4.1 (Translation between Logics ([21])) Let L 1 and L 2 be logics

with sets of formulas L1 and L2 , respectively. A mapping : L1 L2 is said to be
a translation from L 1 to L 2 if, for every  {} L1 :

 L1 =  L2 .

And it is said to be a conservative translation if it satisfies the stronger property:

 L1  L2 . 

Recall the notion of standard logic (Definition 2.1.4). A logic satisfying item (ii)
of that definition is called monotonic, while a logic satisfying item (iv) is said to be
finitary. The following result can be stated (see [6]):

Theorem 2.4.2 Let L 1 and L 2 be two logics, where L 1 is finitary and L 2 is

monotonic, such that both logics have implications and  respectively, satis-
fying the Deduction meta-theorem DMT (see Proposition 2.1.14). Suppose that :
L1 L2 is a mapping for which:
44 2 A Basic Logic of Formal Inconsistency: mbC

L1 = L2 ,

and this mapping is such that ( ) =  . Then is a translation from

L 1 to L 2 . If, additionally, both L 1 and L 2 are finitary and monotonic, and
satisfies the stronger property:

L1 L2 ,

then the mapping is a conservative translation.

Proof Suppose that  L1 . By the finitariness of L 1 , there is a finite 0  such

that 0 L1 . Suppose that 0 ={1 , . . . , n } is nonempty.
 Then, given the fact
that satisfies DMT, L1 1 . . . (n ) . . . . From the hypothesis on

, it is the case that:

  
L2 1 . . . (n ) . . .

thus  
L2 1  . . .  (n  ) . . . .

Given the fact that  satisfies DMT,

1 , . . . , n L2

and, given the monotonicity of L2 ,  L2 . The remaining is proved similarly.

The case when 0 is empty is even simpler. 

Definition 2.4.3 Let c be the signature {, , , } (recall Definition 2.1.3). Let

CPL be the Hilbert calculus for Classical Propositional Logic which is defined over
the language Lc , obtained from the Hilbert calculus for CPL+ (see comment after
Definition 2.1.13) by adding the following axiom schemas:

(TND)
 
(exp)

The consequence relation generated by CPL will be denoted by CPL . 

Notice that the two axioms above, (TND) and (exp), are tertium non datur and
the classical explosion law, both w.r.t. the negation .
It will be useful to consider the expansion of CPL to the language generated by
the signature c = c {}, obtained by adding a trivial consistency operator
such that is always a top formula:
2.4 Recovering Classical Logic Inside mbC 45

Definition 2.4.4 Let CPL be the calculus over c obtained by adding to CPL the
axiom schema . 
It is clear that CPL is sound and complete w.r.t. the classical 2-valued truth-
tables for the connectives in c displayed below, with the addition of the following
truth-table for :

where 1 is the unique designated truth-value. That is: for every  {} Lc ,
 CPL iff  |=CPL , where |=CPL is the semantical consequence relation
defined by the logical matrix above (recall the notion of logical matrix in Defini-
tion 4.1.2).
Proposition 2.4.5 Let t : L Lc be the mapping which replaces by .7 Then
t is a translation from mbC to CPL . That is: for every  {} L ,  mbC
implies t[] CPL t (). The mapping t is not a conservative translation.
Proof It is clear that mbC is sound for the truth-tables above which caracterize the
logic CPL , whenever is interpreted by the truth-table for . Indeed, it is enough
to see that every instance of an axiom of mbC is a tautology w.r.t. the truth-tables
for CPL when is interpreted as . This means that, for every  {} L ,
 mbC implies t[] |=CPL t (). As CPL is complete for such semantics, it
follows that, for every  {} L ,  mbC implies t[] CPL t (). This
shows that t is a translation from mbC to CPL . The mapping t is not a conservative
translation: for instance, if p and q are two distinct propositional variables, then
t ( p), t ( p) CPL t (q) (i.e., p, p CPL q) but p, p mbC q.
Recall that 0 denotes the signature {, , , } (see Definition 2.1.3).
Proposition 2.4.6 Let t0 : L0 Lc be the mapping which replaces by .8
Then t0 is a translation from the 0 -fragment of mbC to CPL. More precisely: for
every  {} L0 ,  mbC implies t0 [] CPL t0 (). The mapping t0 is not a
conservative translation.
Proof Observe that t0 is the restriction to L0 of the mapping t of Proposition 2.4.5.
Thus, for every  {} L0 ,  mbC implies t0 [] CPL t0 (). Combining

7 In formal terms, t is recursively defined as follows: t ( p) = p if p V ar ; t () = t (); t () =

t (); and t ( # ) = t () # t () if # {, , }.
8 In formal terms, t is recursively defined as follows: t ( p) = p if p V ar ; t () = t (); and
0 0 0 0
t0 ( # ) = t0 () # t0 () if # {, , }.
46 2 A Basic Logic of Formal Inconsistency: mbC

this with the clear fact that CPL is a conservative extension of CPL it follows
that, for every  {} L0 ,  mbC implies t0 [] CPL t0 (). Thus, t0 is a
translation. The proof that t0 is not a conservative translation is identical to the one
for t given above, by changing CPL by CPL. 
The last result simply shows that the fragment of mbC without is a proper
sublogic of CPL (once we interpret as ). Moreover, Proposition 2.4.5 shows that
the full logic mbC is a proper sublogic of CPL , the (inessential) expansion of CPL
obtained by adding as a top formula. Indeed, some interactions valid in CPL
between the negation and the other connectives are missing in mbC, for instance
the ones described in Propositions 2.3.4(1) and 2.3.5. Of course, the fact that is
explosive while is not is the first evidence that the negation of mbC is weaker
than the negation of CPL.
However, mbC is not as weak as it seems: actually, CPL can be fully interpreted
inside mbC, as we shall see. Moreover, mbC can be regarded as an expansion of
CPL obtained by adding a consistency operator and a paraconsistent negation
(see Sect. 2.5). In this sense, mbC can be seen both as a subsystem of CPL and as
a conservative extension of CPL.
There are two natural ways to reproduce CPL inside mbC: one is to consider con-
servative translations, while the other is to state a Derivability Adjustment Theorem
(or DAT) between CPL and mbC as follows:
Theorem 2.4.7 Let t  : Lc L0 be the mapping which replaces with .9 Then
the following holds: for every  {} Lc ,

 CPL iff , t  [] mbC t  () for some  L0 ,

where  = { : }.10
Proof For the only if part, suppose that  CPL and let be a derivation 1 n
of from  in CPL. The sequence t  ( ) provided by t  (1 ) . . . t  (n ) is not, in gen-
eral, a derivation of t  () from t  [] in mbC because some instances of the explo-
sion law (exp) could occur in . But the sequence t  ( ) can be transformed into a
derivation in mbC as follows: for each i = i (i i ) occurring in as an
instance of axiom (exp) (not occurring, therefore, as an hypothesis or as the conse-
quence of an application of MP), replace the formula t  (i ) = t  (i ) (t  (i )
t  (i )) in t  ( ) by the following sequence: t  (i )(t  (i ) t  (i ))t  (i ). Observe
that the latter sequence is a derivation of t  (i ) from {t  (i )} in mbC, as t  (i )
t  (i ) is an instance of axiom (bc1). After completing this procedure, a new sequence
 of formulas will be obtained from t  ( ). Let  L0 be the set of all the formulas
t  (i ) used in the procedure described above. In this way,  is clearly a derivation
in mbC of t  () from  t  [], and consequently , t  [] mbC t  ().

9 In formal terms, t  is defined recursively as follows: t  ( p) = p if p V ar ; t  () = t  (); and

t  ( # ) = t  () # t  () if # {, , }.
10 Notice that the mapping t  is necessary only because we consider different signatures for mbC

and CPL.
2.4 Recovering Classical Logic Inside mbC 47

For the if part, suppose that , t  [] mbC t  () for some  L0 . By
Proposition 2.4.5 and the soundness of CPL w.r.t. its truth-tables, it follows that
t[],  |=CPL , given that t (t  ()) = for every Lc . However, every for-
mula in t[] is a tautology in CPL . So,  |=CPL , by definition of the semantics
of CPL . Given the completeness of CPL w.r.t. its truth-tables,  CPL . 

By employing a conservative translation, another form to reproduce CPL inside

mbC can be obtained. Before defining the translation mapping, it is necessary to
observe that, as a consequence of axiom (bc1), a bottom formula is always
definable in mbC from a given formula as = ( ):
def

1 1 0 0 v1
0 1 0 v2
0 0 v3
0 1 1 0 v4
0 0 v5

which is represented here in a more compact and natural way:

1 0
0 0

Now, by combining with the implication connective (as is done for instance,
in intuitionistic logic or even in classical logic, see Sect. 2.5), a new unary operator
(namely, a negation) can be defined as follows. Let ( p) be a formula which depends
exclusively on the propositional variable p, and let ( p) be the unary operator
defined as follows: ( p) p = p ( p) . By substitution, it is clear that () =
def

() , for every formula . To simplify the notation, we will write and

instead of () and () , respectively, for every formula . Observe that the new
operator produces the following table:

1 1 0 0 v1
1 0 0 0 v2
0 1 0 1 v3
0 0 0 1 v4
48 2 A Basic Logic of Formal Inconsistency: mbC

or, in a more compact way,

1 0
0 1

Observe that satisfies the basic properties of a Boolean negation:

1 0 0 1
0 1 0 1

and is semantically equivalent to (observe the abuse of notation in the

last formula: indeed, it should be written as (() ) () ). Because of these
properties, the operator defined in mbC as above for every formula ( p), is
called a strong negation. In order to simplify the notation a bit more, from now on
(and when there is no risk of confusion), we will write and instead of and
, respectively.
The bottom and the strong negation satisfy the following properties:

Proposition 2.4.8 The following holds in mbC, for every  {} L :

(i) mbC .
(ii) If , mbC then  mbC .
(iii) mbC , and so mbC .

Proof (i) It is a consequence of the definition of and axiom (bc1).

(ii) Assume that , mbC . By item (i), mbC and then , mbC .
Since is always the case that , mbC , it follows that  mbC , by Propo-
sition 2.1.14(iii).
(iii) By definition of and by MP, , mbC . Then mbC , by item (ii).
Finally, mbC by DMT. 

More properties of the strong negation will be proved in Chap. 7, Proposition 7.2.2.
As was done in da Costas logic C1 , another negation operator in mbC could be
defined as follows:  = . Despite  being explosive, it does not behave
def

exactly as expected, as tertium non datur is not satisfied (see valuation v5 below):

  
1 1 0 0 0 1 v1
0 1 0 0 1 v2
0 0 0 1 v3
0 1 1 1 0 1 v4
0 0 0 0 v5
2.4 Recovering Classical Logic Inside mbC 49

Additionally,  does not satisfy the law of double negation: take an mbC-valuation
v such that v( p) = 1 for a given propositional variable. From this, it follows that
v( p) = 0 and so v( p) = 1. Suppose that v( p) = 0. Then v( p) = 0,
showing that mbC ( p  p). Now, consider an mbC-valuation v such that
v ( p) = 0, v ( p) = 0 (and so v ( p) = 1), and v ( p) = 1. From this,
v ( p) = 1 and therefore mbC ( p p).
It is possible to correct all the failures of the operator  by requiring the following
additional property to the valuations:

v() = 0 = v() = 1.

It is easy to prove that this strategy corresponds to adding to mbC the following
axiom schema:

The following proposition shows that tertium non datur is indeed equivalent to the
law above, and it shows also that the law of double negation will be automatically
satisfied by the alternative negation :
Proposition 2.4.9 Let L be an extension of mbC.
(i) L  iff L .
(ii) If L  then L  and L  .
Proof (i) Observe that  = ( ) is equivalent to ( )
( ) (by distributivity), which is in turn equivalent to (given that
is a theorem of L ). This proves the result.
(ii) As observed above, an extension L proving the theorem , is character-
ized by mbC-valuations which must satisfy: for every , if v() = 0 then v() = 1.
It is easy to see, then, that the unique counterexamples of the law of double negation
in both directions (see comment above) are now forbidden, as is now logically
equivalent to  in L . 
This point will be resumed in Sect. 3.
Now, profiting from the strong negation definable in mbC (as discussed above),
we show that the mapping that replaces the classical negation , with the defined
operator , is a conservative translation from CPL to mbC. This constitutes an
alternative way to recover classical logic inside mbC.
Proposition 2.4.10 Fix a given formula ( p1 ) L which depends exclusively on
the propositional variable p1 (for instance, = p1 ). The mapping  : Lc L is
recursively defined as follows:

p  = p, if p V ar ;
() = (  );
( # ) =  #  if # {, , }.
50 2 A Basic Logic of Formal Inconsistency: mbC

In this way  is a conservative translation from CPL to mbC. That is: for every
 {} Lc ,
 CPL iff   mbC  .

Proof By Theorem 2.4.2, it suffices to prove that CPL iff mbC  , for every
formula Lc .
For the only if part, recall that CPL is obtained from CPL+ by adding axiom
schemas (TND) and (exp). As mbC is an extension of CPL+ , and the translation of
all the instances of axiom schemas and rules of CPL+ produces instances of axiom
schemas and rules of CPL+ (and so of mbC), the only cases to be analyzed are the
two axiom schemas (TND) and (exp). Thus, let and ( ) be
instances of (TND) and (exp), respectively. They are translated by  as the formulas
 (  ) and  ( (  )  ), respectively. But, as observed above,
is always true and is always false for the mbC-valuations, for
every . Being so,  (  ) and  ( (  )  ) are always true for
mbC-valuations. Thus, they are derivable in mbC, by completeness. This shows that
if CPL then mbC  .
For the if part, suppose that mbC  . By Proposition 2.4.5 it follows that
CPL t (  ). However, in CPL , the formula t ( ) is equivalent to t (), for
every . As a consequence, it is easy to prove by induction on the complexity of
that t (  ) is equivalent to in CPL . Hence CPL and so, by the fact that
CPL is a conservative extension of CPL, it follows that CPL , which completes
the proof. 

2.5 Reintroducing mbC as an Expansion of CPL

The results of the previous section, in particular Proposition 2.4.10, suggest that a
different language based on the signature  = {, , , } could be used for mbC
(recall Definition 2.1.3). This is the proposal found in [6], on which this section is
based.
The notion of complexity of a formula in L is defined analogously to Defini-
tion 2.1.11:
Definition 2.5.1 (Formula Complexity in L ) The complexity of a given formula
L , denoted by l(), is recursively defined as follows:
1. If = p, where p Var {}, then l() = 1;
2. If = , then l() = l() + 1;
3. If = , then l() = l() + 2;
4. If = , then l() = l() + l() + 1.

As observed in the previous section, any formula defines a bottom constant
in mbC as follows: = ( ) and then CPL can be recovered within
def
2.5 Reintroducing mbC as an Expansion of CPL 51

mbC by means of a conservative translation (see Proposition 2.4.10). This suggests

that it is possible to consider from the start a 0-ary connective and the axiom
schemas for CPL in the signature  , as well as the corresponding axiom schemas
for the paraconsistent negation and the consistency operator , to obtain a new
axiomatization of mbC in the signature  .
Besides the simplification it achieves (for instance, when doing proofs by induc-
tion on the complexity of a formula), a justification for the use of this new language
is that , being so important in the context of LFIs, is usually defined with respect
to a formula () as () , and so there is an infinitude of such bottom formu-
las. The same observation applies to the classical negation (), which is defined as
() = () , and so there are infinitely many classical negations inside
def

mbC, with the index () varying with . For that reason, the inclusion of bottom
to the signature allows to define a distinguished classical negation as expected:
Definition 2.5.2 (Classical Negation in L ) Let p be a propositional variable. The
classical negation is defined in signature  by means of the following formula:

p = p .
def


Hence, if is any formula in L then = , by substitution. This
def

constitutes a significant simplification with respect to the signature , where the

index of the negation varies with the formula being negated. From this, and can be
considered the canonical choices for bottom and the classical negation inside mbC.
Moreover, the new presentation of mbC in this signature is equivalent to consider this
logic as an expansion of classical propositional logic CPL (this time defined in the
signature {, }) by adding a paraconsistent negation and a consistency operator.
This allows mbC to be seen as a kind of bimodal logic based on CPL.

2.5.1 The New Presentation mbC of mbC

Definition 2.5.3 (mbC ) The calculus mbC is defined over the language L by
the following Hilbert calculus:
Axiom schemas:

( ) (Ax1)
      
(Ax2)
()
()

 
(bc1 )
52 2 A Basic Logic of Formal Inconsistency: mbC

Inference rule:

(MP)

As was argued for mbC in Proposition 2.1.14, the calculus mbC satisfies the
Deduction meta-theorem DMT:

Theorem 2.5.4 (Deduction meta-theorem) For every  {, } L :

 {} mbC  mbC .

The following result is well-known in the literature. See a proof, for instance, in
Chap. 1 of Churchs book [22] (where the system CPLW below is called P1 ).11

Proposition 2.5.5 Let W be the signature {, }. Consider the logic CPLW given
by the Hilbert calculus over the signature W which is formed by the axiom schemas
(Ax1), (Ax2) and () plus the rule (MP). In this way CPLW constitutes a sound
and complete axiomatization of CPL in the signature W (by taking as in Defin-
ition 2.5.2).

Remark 2.5.6 The axiomatization CPLW of CPL in the signature {, } was firstly
proposed in 1939 by Wajsberg (see [23]).

The next technical lemma is required for establishing the completeness theorem
in the following section.

Lemma 2.5.7 All the following formulas are theorems of mbC :

1.
 
2. ( ) 
  
3. ( ) ( ) ( )
  

4. ( ) ( ) ( )

Proof It follows from the fact that all these formulas are classic tautologies in the
signature {, } and they can therefore be derived in mbC by Proposition 2.5.5. 

11 Not to be confused with Settes P1, see Sect. 4.4.4. To be more precise P1 contains, besides MP,
the inference rule of Uniform Substitution, since the axioms are presented by using propositional
variables instead of schema formulas.
2.5 Reintroducing mbC as an Expansion of CPL 53

2.5.2 Valuation Semantics for mbC

Now a valuation semantic, adapted from that for mbC, will be proposed for mbC .
 
Definition 2.5.8 (Valuations for mbC ) A function v : L 0, 1 is a valuation
for mbC , or an mbC -valuation, if it satisfies the following clauses:
(vBot) v() = 0
(vNeg) v() = 0 = v() = 1
(vCon) v() = 1 = v() = 0 or v() = 0.
(vImp) v( ) = 1 v() = 0 or v() = 1

The set of all such valuations will be designated by V mbC . 

A technical result is given below, whose demonstration will be used latter on,
more specifically in the proof of Theorem 2.5.20.
 
Lemma 2.5.9 Let v0 : Var 0, 1 be a mapping. Then there exists a valuation

v V mbC extending v0 .

Proof For L , the truth-values of v() are defined by induction on l(). To

begin, if is such that l() = 1, then either Var or = . In the first case,
the valuation is defined as v() = v0 (). In the second case, we define v() = 0.
Now, suppose that the valuation v is already defined for all  such that l(  ) < n,
where n > 1 (induction hypothesis), and let such that l() = n. According to the
main connective of , the definition goes as follows:
1. If = , then:

v( ) = 1 v() = 0 or v() = 1

2. If = , then:

1 if v( ) = 0, or
v( ) =
arbitrary otherwise

3. If = , then:

0 if v( ) = v( ) = 1, or
v( ) = .
arbitrary otherwise

It is worth noting that v is well-defined, by Definition 2.5.1 of complexity. Clearly,

v is an mbC -valuation extending v0 . The easy details are left to the reader. 
54 2 A Basic Logic of Formal Inconsistency: mbC

Remark 2.5.10 The reader should notice that, different to the case of valuations
over logical matrices, which are homomorphisms
  between algebras (see Sect. 4.1 in
Chap. 4), any mapping v0 : Var 0, 1 can be extended to more than one mbC -
valuation (in fact, there are infinite mbC -valuations extending v0 ). This is a conse-
quence of the non-determinism inherent to valuations of this kind, in contrast with the
determinism imposed by homomorphic valuations. The question of non-determinism
versus determinism in the context of semantic approaches to LFIs will be analyzed
with more details in Chap. 6.
Now we will prove that the new logic mbC is sound and complete for its seman-
tics of valuations.
Theorem 2.5.11 (Soundness for mbC ) For every  {} L :

 mbC =  mbC .

Proof The proof of this theorem is entirely analogous than that of Theorem 2.2.2
for mbC. The only axioms to be analyzed are the new ones. Firstly observe that, for
every mbC -valuation v and for every formula , v() = 1 iff v() = 0. Hence,
axiom () is clearly valid. Let = be an instance of axiom (), and
let v be an mbC -valuation. If v() = 0 then v( ) = 1, by (vImp). Otherwise, if
v() = 0 then v() = 0 and so v() = 1, by  clause (vNeg).
 From this, v( ) = 1
by (vImp). Finally, let be an instance of axiom (bc1 ), and let
v be an mbC -valuation. If v() = 0 then v() = 1, by clause (vImp). Otherwise, if
v() = 1 then either v() = 0 or v() = 0, by (vCon). If v() = 0 then v() =
v() = 1 and so v( ) = 1. If v() = 0 then clearly v( ) = 1
too. In both cases v() = 1. 
The proof of completeness will be also analogous to that of mbC.
Theorem 2.5.12 Let  {} L , with  maximal non-trivial
 with respect to
in mbC (recall Definition 2.2.3). The mapping v : L 0, 1 defined by:

v() = 1 

for all L is a valuation for mbC .

Proof Let L be an arbitrary formula.
1. = . Suppose, by contradiction, that . As mbC (by
Lemma 2.5.7(1)) then , by Lemma 2.2.5. By MP, and using Lemma 2.2.5
again, it follows that , which is a contradiction. Therefore /  and so
v() = 0.
2. = . Suppose /  and, by contradiction, also
/ . As  is maximal,
it follows that , mbC and , mbC . By the deduction meta-theorem,
 mbC and  mbC . Now, by Lemma 2.5.7(4),  mbC
( ) . However, ( ) is an instance of Axiom (),
and then (by MP)  mbC , which is a contradiction. Therefore:
2.5 Reintroducing mbC as an Expansion of CPL 55

v() = 0 = v() = 1.

3. = . Suppose  and, by contradiction, that both  and .

Then, by Axiom (bc1 ) and Lemma 2.2.5, . By the definition of and
by MP, it follows that . But then, as a consequence of Lemma 2.5.7(1), it can
be inferred that , which is a contradiction. Therefore, either
/  or
/ .
That is:
v() = 1 = v() = 0 or v() = 0.

4. = . Suppose . If  then  by MP and Lemma 2.2.5.

Therefore:
v( ) = 1 = v() = 0 or v() = 1.

Now, suppose /  or . If  then  by Axiom (Ax1), MP and

/  then, by the maximality of , it follows that , mbC .
Lemma 2.2.5. If
Now, suppose by contradiction that / . Then , mbC . By the
deduction meta-theorem, both
   ( ) and  
 mbC . By
mbC

Lemma 2.5.7(3),  mbC ( ) and, by item 2 of that
lemma,  mbC , which is a contradiction. Therefore:

v() = 0 or v() = 1 = v( ) = 1.

The completeness of mbC is then an immediate consequence of Theorems 2.5.12

and 2.2.6 (which clearly holds for mbC ):

Corollary 2.5.13 (Completeness of mbC ) For every  {} L :

 mbC =  mbC

Proof Suppose  mbC and let  be a maximal non-trivial set with respect to
in mbC extending  (see Theorem 2.2.6). By Theorem 2.5.12, there is a valuation
for mbC satisfying  (as  ) but not (as
/ ). Therefore,  mbC and
the theorem follows by contraposition. 

2.5.3 Equivalence Between mbC and mbC

In this section, mbC will be shown to be equivalent to its counterpart mbC.

By induction on the formula complexity, two mappings will now be defined which
later on will be shown to be conservative translations.
56 2 A Basic Logic of Formal Inconsistency: mbC

Definition 2.5.14 Fix an arbitrary propositional variable in Var, for instance p1 . The
mapping  : L L is defined inductively for all L as follows:

q  = q, if q Var;
 = p1 ( p1 p1 );
(#) = #(  ) for # {, };
  
( ) = .

Definition 2.5.15 The mapping : L L is defined by induction on l(), for

all L as follows:
q = q, if q Var;
(#) = #( ) for # {, };
( ) = ;
 
( ) = ;
 
( ) = ( ) .

The injectivity of these mappings needs to be established, in order to be possible

to properly define the valuations of Theorem 2.5.20. But, first, an intermediary result
shall be given:

Lemma 2.5.16 There are no formulas , L which satisfy the following

equation:
=

Proof Suppose, by contradiction, there is a solution in L for the above identity

and let and be such a solution with a minimum value of l() + l(). Observe
now that is not in the image of , and so = and = for any
of these would imply = . Therefore, the only way to get the image of
 to
be is with = 1 2 . Therefore = 1 (2 ) ,
and so = 1 (2 ).
There are two cases to be analyzed:
1. = 1 2 . Therefore = 1 2 , 1 = 1 and 2 = 2 .
2. = 1 2 . Therefore = (1 ) 2 , 1 = 1 and 2 =
2 .
In both cases (2 , 2 ) is a solution to the equation in question with l(2 ) + l(2 ) <
l() + l(), a contradiction. 

Theorem 2.5.17 The mappings  : L L and

: L L from Defini-
tions 2.5.14 and 2.5.15 are injective.
2.5 Reintroducing mbC as an Expansion of CPL 57

Proof Let , L be such that  =  . By induction on l() + l() it is easy

to prove that = . This is a consequence of the fact that, by Definition 2.5.14,
there are not two different equations having, to the right, formulas with the same
main connective.
Now, let , L be such that = . The proof is given by induction,
analogous to  . However, the induction step for which the main connective of both
sides of the above equation is proves to be a bit more complicated. In fact, there
are three equations in Definition 2.5.15 having, to the right, a formula with as
the main connective. So, let =   = . Hence, there are the following
possibilities:
(a) = and = . Therefore, = = = . By
unique readability, it follows that = and = . The result is then obtained
by the induction hypothesis: = and = , which implies = .
(b) = and = . Therefore, = = ( ) =
. By unique readability, = , which is impossible by Lemma 2.5.16.
(c) = and = . This is impossible, for it would imply = .
(d) = and = . Then, as in item (a), = and = , given the fact
that = and = . Then = .
(e) = and = . This is impossible, for it would imply = .
(f) = and = . Then ( ) = ( ), which
implies = and = . Therefore = . 
Corollary 2.5.18
1. Let = # L , with # {, }. If I m( ) = { : L }, there exists
a unique formula L such that = (#) .
2. Let = # L , with # {, }. If I m( ) = {  : L }, there exists
a unique formula L such that = (#) .
Proof This result is a direct consequence of the injectivity and the very definition of
the mappings and  . 
Lemma 2.5.19  
1. Let v V mbC . Then the mapping v : L 0, 1 defined by v () = v(  ) is
def

such that v V mbC .  

2. Let v V mbC . Then the mapping v : L 0, 1 defined by v () = v( ) is
def

such that v V mbC .

Proof 1. Let L be an arbitrary formula. We will prove that v satisfies the
clauses from Definition 2.5.8.
(a) = . Then  = p1 ( p1 p1 ) and so v(  ) = 0 for any valuation for
mbC. Therefore v () = v(  ) = 0.
(b) = . Then  = (  ). Therefore, if v () = 0, then v((  )) = 0 (by
definition of v ). Now, as v is a valuation for mbC, it follows that v(  ) = 1, and so
v () = 1.
(c) = . Then  = (  ). Therefore, if v () = 1, then v((  )) = 1. Now,
as v is a valuation for mbC, v(  ) = v () = 0 or v((  )) = v () = 0.
58 2 A Basic Logic of Formal Inconsistency: mbC

(d) = . Then  =   . Therefore, v ( ) = 1 if, and only if,

v(   ) = 1. But the last occurs exactly when v(  ) = 0 or v(  ) = 1, that
is, exactly when v () = 0 or v () = 1.
2. Let L be an arbitrary formula. We will prove that v is an mbC-valuation. If
is of the form , , or , the proof is similar to that of item 1 above. Now,
for the remaining cases:

(a) =  . Then = ( ) and therefore, v ( ) = v (
) . Since and are interpreted as in propositional classical logic, it fol-
lows that v ( ) = v( ) = 
 1 iff v () = v( ) = 1 or v () = v( ) = 1.

(b)
 = . Then =  ( ) and therefore, v ( ) =

v ( ) . In line with the argument in the previous item, it fol-
lows that v ( ) = v( ) = 1 iff v () = v( ) = 1 and v () = v( ) = 1. 
The next lemma establishes that when there is a model (or counter-model) for a
formula in mbC , there also exists a model (or counter-model) for  in mbC.
Similarly, given a model (or counter-model) for a formula in mbC, there also exists
a model (or counter-model) for in mbC . As it will become clear later on, this
result suffices to prove that the translations in question are conservative ones.
Lemma 2.5.20

1. Let v V mbC . Therefore, there exists v V mbC such that v ( ) = v(), for
every L .

2. Let v V mbC . Therefore, there exists v V mbC such that v (  ) = v(), for
every L .
 
Proof 1. Let v be a valuation for mbC. Define a mapping v : L 0, 1 by
induction on the complexity of the formula L as follows:
If = q Var, then v (q) = v(q).
If = , then v () = 0.
If = , then v ( ) = 1 iff v () = 0 or v ( ) = 1.
If = , then

1 if v ( ) = 0

v ( ) = v() if = ()

arbitrary otherwise.

If = , then

0 if v ( ) = v ( ) = 1

v ( ) = v() if = ()

arbitrary otherwise.

Using Corollary 2.5.18, it is easy to prove by induction on the complexity of

formulas that v is well-defined and v ( ) = v() for every L . Additionally,

v V mbC by the proof of Lemma 2.5.9.
2.5 Reintroducing mbC as an Expansion of CPL 59
 
2. Let v be a valuation for mbC . Consider a mapping v : L 0, 1 defined by
induction as follows:
If = q Var, then v (q) = v(q).
If = , then v ( ) = 1 iff v () = v ( ) = 1.
If = , then v ( ) = 0 iff v () = v ( ) = 0.
If = , then v ( ) = 1 iff v () = 0 or v ( ) = 1.
If = , then

1 if v ( ) = 0

v ( ) = v() if = ()

arbitrary otherwise.

If = , then

0 if v ( ) = v ( ) = 1

v ( ) = v() if = ()

arbitrary otherwise.

Using Corollary 2.5.18, one can simply prove by induction on the complexity of
formulas that v is well-defined and v (  ) = v() for every L . Moreover,
v V mbC by the proof of Lemma 2.5.9. 

The equivalence between these logics in the different languages can then be estab-
lished by the following theorem:

Lemma 2.5.21 The functions : L L and : L L satisfy the
following:

mbC mbC 
mbC mbC .

Proof As a consequence of the completeness of the logics in both languages, the

lemma can be proved by using valuation semantics, namely:

V mbC V mbC 

V mbC V mbC

or equivalently by contraposition:

v V mbC : v() = 0 v V mbC : v(  ) = 0

v V mbC : v() = 0 v V mbC : v( ) = 0.
60 2 A Basic Logic of Formal Inconsistency: mbC

For the first equivalence, suppose that there exists v V mbC such that v() = 0.
By item 2 of Lemma 2.5.20, there exists v V mbC such that v (  ) = v() = 0.
Conversely, if v(  ) = 0 for some v V mbC then, by item 1 of Lemma 2.5.19, there

exists v V mbC such that v () = v(  ) = 0.
Now, suppose that there exists v V mbC such that v() = 0. Because of item 1 of

Lemma 2.5.20, there exists v V mbC such that v ( ) = v() = 0. Conversely, if
mbC
v( ) = 0 for some v V then, by Lemma 2.5.19, item 2, there exists v V mbC

such that v () = v( ) = 0. 

Theorem 2.5.22 The mapping : L L is a conservative translation from
mbC to mbC. The mapping
: L L is a conservative translation from
mbC to mbC .

Proof This result is a direct consequence of Theorem 2.4.2 and Lemma 2.5.21. 

It is worth noting that the disjunction inside mbC must be defined exactly
as it is proposed here: if disjunction is interpreted as it usually is, in terms of the
implication, the resulting mapping is no longer a conservative translation:

Proposition 2.5.23 Let : L L be the translation mapping of Definition

 of the clause for , which is replaced by the following:
2.5.15 except in the case
( ) = . Then the mapping , thus defined, it is not a conser-
vative translation even though it is a translation from mbC to mbC .

Proof First observe that both formulas and ( ) are translated into
the same formula:
   
( ) = = ( ) ,

and thus the translation is not injective. Moreover, there is a way to choose a formula
whose translation under is a theorem, while there is some other formula translated in
the same theorem which is not a theorem of the source logic. Consider, for instance,
the formula () in L and let be the formula () , that is:
 
= ( ) ( ) ( ) .

It is easy to see that is a theorem of mbC , and is also a theorem of mbC. But
now consider the following formula in L :
   
= ( ) ( ) ( ) .
2.5 Reintroducing mbC as an Expansion of CPL 61

It is a straightforward task to prove that = , however, is not a theorem in the

source logic. This shows that

mbC = mbC .

This illustrates the consequences of a logic not being self-extensional (recall the
end of Sect. 2.3) and draws our attention to the care required when dealing with this
kind of logic. The last proposition shows that the right translation of disjunctions
inside mbC is done through a schema formula that uses and . As the example
above shows, the other formulation of disjunction (just using ) simply does not
work, despite both formulations being equivalent in mbC .
Although mbC and mbC are equivalent in the sense of being inter-translatable,
they are not the same logic: there are some subtle distinctions between them, that
will be explored in Sect. 3.4.

References

1. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Incon-
sistency. In Handbook of Philosophical Logic, 2nd. edn. vol. 14, ed. by Dov M. Gabbay and
Franz Guenthner, 193. Springe. doi:10.1007/978-1-4020-6324-4_1.
2. Wjcicki, Ryszard. 1984. Lectures on propositional calculi. Wroclaw, Poland: Ossolineum.
http://www.ifispan.waw.pl/studialogica/wojcicki/papers.html.
3. Carnielli, Walter A., and Joo Marcos. A taxonomy of C-systems. In [Carnielli, Walter A.,
Marcelo E. Coniglio, and Itala M.L. DOttaviano, eds. 2002. Paraconsistency: The Logical
Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP
2000), Vol. 228 of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New
York], pp. 194.
4. Coniglio, Marcelo E., and Luiz H. Silvestrini. 2014. An alternative approach for quasi-truth.
Logic Journal of the IGPL 22(2): 387410. doi:10.1093/ljigpal/jzt026.
5. Coniglio, Marcelo E., Francesc Esteva, and Llus Godo. 2014. Logics of formal inconsistency
arising from systems of fuzzy logic. Logic Journal of the IGPL 22(6): 880904. doi:10.1093/
jigpal/jzu016.
6. Coniglio, Marcelo E., and Tarcsio G. Rodrigues. 2014. Some investigations on mbC and mCi.
In Tpicos de lgicas no clssicas. Nel-Lgica, vol. 1, ed. by Czar A. Mortari, 1170,
Florianpolis: SC, Brazil, 2014. NEL/UFSC. http://nel.ufsc.br/logica.html.
7. Mendelson, Elliot. 1987. Introduction to Mathematical Logic, 3rd edn. New York: Chapman
& Hall.
8. da Costa, Newton C.A., and Elias H. Alves. 1976. Une smantique pour le calcul C1 (in French).
Comptes Rendus de lAcadmie de Sciences de Paris (A-B), 283: 729731.
9. da Newton, C.A. 1977. Costa and Elias H. Alves. A semantical analysis of the calculi Cn . Notre
Dame Journal of Formal Logic 18(4): 621630.
10. Avron, Arnon, and Iddo Lev. 2001. Canonical propositional Gentzen-type systems. In Proceed-
ings of the 1st International Joint Conference on Automated Reasoning (IJCAR 2001). LNAI,
vol. 2083, ed. by Rajeev Gore, Alexander Leitsch, and Tobias Nipkow, 529544. Springer.
11. Avron, Arnon, and Iddo Lev. 2005. Non-deterministic multi-valued structures. Journal of Logic
and Computation 15(3): 241261.
62 2 A Basic Logic of Formal Inconsistency: mbC

12. Coniglio, Marcelo E., and Martn Figallo. 2014. Hilbert-style presentations of two logics asso-
ciated to tetravalent modal algebras. Studia Logica 102(3): 525539. doi:10.1007/s11225-013-
9489-0.
13. Alves, Elias H. 1976. Lgica e inconsistncia: um estudo dos clculos Cn , 1 n (Logic
and inconsistency: A study of the calculi Cn , 1 n , in Portuguese). Masters thesis,
FFLCH, State University of So Paulo.
14. Ivlev, Yury V. 2000. Quasi-matrix logic as a paraconsistent logic for dubitable information.
Logic and Logical Philosophy 8: 9197.
15. Wjcicki, Ryszard. 1979. Referential matrix semantics for propositional calculi. Bulletin of
the Section of Logic 8: 170176.
16. Blok, Willem J., and Don Pigozzi. 1989. Algebraizable Logics. Memoirs of the American
Mathematical Society, vol. 77(396). Providence, RI, USA: American Mathematical Society.
17. Czelakowski, Janusz. 2001. Protoalgebraic logics. Trends in Logic Series, vol. 10. Dordrecht:
Kluwer Academic Publishers.
18. Font, Josep Maria. 2003. Ramn Jansana, and Don Pigozzi. A survey of abstract algebraic
logic. Studia Logica 74(12): 1397.
19. Font, Josep Maria, and Ramn Jansana. 2009. A General Algebraic Semantics for Sentential
Logics. Vol. 7 of Lecture Notes in Logic, 2nd ed. Ithaca, NY, USA: Association for Symbolic
Logic.
20. Font, Josep Maria. 2016. Abstract algebraic logic: An introductory textbook. Vol. 60 of Math-
ematical Logic and Foundations Series. London: College Publications.
21. da Silva, Jairo J., Itala M.L. DOttaviano, and Antonio M.A. Sette. 1999. Translations between
logics. In Models, algebras and proofs. Selected papers of the X Latin American symposium
on Mathematical Logic held in Bogot. Lecture Notes in Pure and Applied Mathematics, vol.
203, ed. by Xavier Caicedo and Carlos H. Montenegro, 435448, New York: Marcel Dekker.
22. Church, Alonzo. 1956. Introduction to Mathematical Logic - Volume 1. Princeton mathematical
series, vol. 17. Princeton University Press.
23. Wajsberg, Mordchaj. 1967. Metalogische Beitrge II (in German). Wiadomosci Matematyczne,
47:119139, 1939. Trans. by S. McCall as Contributions to Metalogic II. In Polish Logic
19201939, ed. by Storrs McCall, 319334. Oxford: Clarendon Press.
Chapter 3
Some Extensions of mbC

Clearly, the logic mbC is the minimal extension of CPL+ having a consistency
operator and a paraconsistent negation , and such that it is an LFI. In this chapter,
several extensions of mbC will be proposed, taking into account different features
of mbC that can be strengthened or expanded.1

3.1 A Wider Form of Truth-Functionality for Consistency

By observing the quasi-matrices presented in Sect. 2.3 of Chap. 2, it becomes clear

that not every connective of mbC is truth-functional (w.r.t. valuations): the tables
for and show that non-determinism exists, to a certain extent, in the evaluation
process. This aspect is not a defect of the valuations for mbC: it can be proven
that mbC (and several extensions of it) are not characterizable by finite matrices,
adapting J. Dugundjis argument for modal logics. This topic will be briefly analyzed
in Sect. 4.2 of Chap. 4.
Returning to the valuation semantics for mbC, it is clear that the truth-value of a
formula partially determines the truth-value of : if is false then must be
true, but when is true the truth-value of cannot be determined: it can be either
true or false. Similarly, the truth-values of and partially determine the truth-
value of in mbC: if they are both true then must be false, but if one of {, }
is false and the other is true (that is, if the truth-values of and are different), the
truth-value of is undetermined: it can be either true or false. In [1] an extension
of mbC called mCi was researched, which guarantees that the truth-values of and
are sufficient to determine the truth-value of . However, this extension also

1 In general, names of logic systems are acronyms for the names of the axioms involved.

Springer International Publishing Switzerland 2016 63

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_3
64 3 Some Extensions of mbC

requires that every formula of the form n must be consistent, for n 0. That
is, n is a theorem for every n 0. In particular, is a theorem. As we shall
see below, it is possible to consider extensions of mbC which are weaker than mCi,
where the truth-values of and are completely determined by the truth-values
of and (see Remarks 3.1.5 and 3.1.11). The logic mCi will be analyzed later
on.
The first extension of mbC proposed in this section, called mbCciw, consists of
the minimum extension capable of guaranteeing that the truth-values of and
will completely determine the truth-value of .
Thus, consider mbC-valuations additionally satisfying the following clause:
(vCiw) v() = 0 or v() = 0 = v() = 1.
Observe that (vCiw) is the converse of (vCon). Let V mbCciw be the set of such
valuations. Then V mbCciw will be the set of mbC-valuations such that

v() = 1 v() = v().

This means that, in V mbCciw , the truth-values of and are sufficient to determine
the truth-value of , as desired.
Notice that, by contraposition, (vCiw) is equivalent to the following: if v() = 0
then v() = v() = 1. From this, it is immediate to see that the axiom schema
corresponding to clause (vCiw) goes as follows:

( ) (ciw)

It is clear that adding (ciw) to mbC is equivalent to adding the following two
axioms:

(d1)
(d2)

Definition 3.1.1 Let mbCciw be the logic obtained from mbC by adding the axiom
schema (ciw). The elements of V mbCciw will be called mbCciw-valuations.


Remark 3.1.2 The two axioms (d1) and (d2) were first studied by Avron in [2] as a
weaker version of (ci), one of the axioms included in mbC extensions such as mCi
and Ci (see [1, 3]). Afterwards, Avron and his collaborators considered axiom (k),
which coincides with axiom (ciw), and system BK, which coincides with mbCciw.
They argued that BK (and not mbC) should be taken as the basic LFI (see [47]).
When analyzing in Sect. 3.6 the propagation of consistency, the system mbCciw will
be used as a basis instead of mbC, for convenience.
3.1 A Wider Form of Truth-Functionality for Consistency 65

The proof of soundness and completeness of mbCciw with respect to V mbCciw

follows straightforwardly from that made for mbC. First, a technical result is
necessary:
Theorem 3.1.3 Let  {} L, with  maximal non-trivial with respect to in
mbCciw. The mapping v : L 0, 1 which is defined by:

v() = 1 

for all L , is a valuation for mbCciw.

Proof By means of Theorem 2.2.7, we know that v is an mbC-valuation. It is suf-
ficient to prove that v satisfies clause (vCiw). Thus, let be a formula such that
either v() = 0 or v() = 0. Then v( ) = 0. However, by axiom (ciw),
mbCciw ( ) and then ( )  by Lemma 2.2.5. Thus, v(
( )) = 1 and v() = 1, accordingly. This shows that v is an mbCciw-
valuation. 
Theorem 3.1.4 (Soundness and Completeness of mbCciw w.r.t. valuations) For
every  {} L :

 mbCciw  mbCciw .

Proof (Soundness) From the soundness of mbC w.r.t. valuations, it is enough to

prove that mbCciw-valuations validate any instance of the axiom (ciw). Thus, let
be an instance ( ) of (ciw) for a given formula , and let v V mbCciw .
If v() = 1, then v() = 1. Otherwise, if v() = 0, it then follows by (vCiw) that
v() = v() = 1 and so v( ) = 1. Therefore, v() = 1.
(Completeness) The proof is similar to the proof of completeness for mbC, by using
Theorem 3.1.3 instead of Theorem 2.2.7. 
The situation in mbCciw concerning consistency, inconsistency, contradiction
and non-contradiction is displayed in Table 3.1.

Table 3.1
( (
1 v1
1 0 1 1
0 v2
1
1 0 1 v3
0 1
0 0 1 v4
1 0 1 v5
0 1 1
0 0 1 v6
66 3 Some Extensions of mbC

Remark 3.1.5 From Table 3.1 it can be seen that, in mbCciw, the relationship
between contradiction and non-consistency, as well as the relationship between non-
contradiction and consistency, are the same than in mbC. Namely, contradiction
is a sufficient (but not necessary) condition for inconsistency. On the other hand,
non-contradiction is a necessary (but not sufficient) condition for consistency.
As observed above, mbCciw is the minimal extension of mbC which guarantees
that the truth-values of and completely determine the truth-value of . How-
ever, because of valuations v3 and v5 , the truth-value of the inconsistency of
cannot be determined from the truth-values of and .
Corollary 3.1.6 The logic mbCciw is a (strong) LFI.
In order to obtain the truth-value of from the truth-values of and , we
can consider the stronger axiom (ci), introduced in [3]:

( ) (ci)

Definition 3.1.7 Let mbCci be the logic obtained from mbC by adding the axiom
schema (ci). We say that an mbC-valuation v is an mbCci-valuation if it also satisfies
the following clause:
(vCi) v() = 1 = v() = 1 and v() = 1.

The set of mbCci-valuations will be denoted by V mbCci .


Theorem 3.1.8 Let  {} L , with
  maximal non-trivial with respect to in
mbCci. The mapping v : L 0, 1 , defined by:

v() = 1 

for all L , is a valuation for mbCci.

Proof The mapping v is an mbC-valuation by virtue of Theorem 2.2.7. It is
enough to prove that v satisfies clause (vCi). Thus, let be a formula such that
v() = 1. Since mbCci ( ) by (ci), then ( )  by
Lemma 2.2.5, and v( ( )) = 1, accordingly. Therefore, v( ) =
1 by (vImp) and v() = v() = 1 by (vAnd). This shows that v is an mbCci-
valuation. 
Theorem 3.1.9 (Soundness and Completeness of mbCci w.r.t. valuations)
For every  {} L :

 mbCci  mbCci .

Proof It is analogous to that for mbCciw, but uses Theorem 3.1.8 instead of
Theorem 3.1.3. 
3.1 A Wider Form of Truth-Functionality for Consistency 67

Proposition 3.1.10
(1) The axiom (ciw) is derivable in mbCci, but (ci) is not derivable in mbCciw. That
is: mbCci ( ) for every , but mbCciw ( ) for some .
Therefore, the logic mbCci is strictly stronger than mbCciw.
(2) The axiom schema
(cc)

is derivable in mbCci.
(3) The logic mbCci is equivalent to adding the axiom schema (cc) to mbCciw.
That is, mbCci is equivalent to the extension of mbCobtained by adding axiom
schemas (ciw) and (cc).

Proof In virtue of the respective soundness and completeness theorems, this proof
will be done semantically by means of valuations.
(1) Let v be an mbCci-valuation such that v() = 0. Then v() = 1 by (vNeg),
and so v( ) = 1 by (vCi) and (vAnd). This shows that any instance (
) of (ciw) is validated by v and so it is valid in mbCci, being therefore demonstra-
ble in mbCci. On the other hand, let be an instance ( ) of (ci) such
that is a propositional variable. Let v be the mbCciw-valuation v3 or v5 considered
in the Table 3.1. Then v() = 0, showing that mbCciw ( ). By means
of soundness, mbCciw ( ).
(2) Let v be an mbCci-valuation and a formula such that v() = 1 and
v() = 1. By clauses (vCon) and (vNeg), v() = v(). But, by clause (vCi), we
also have v() = v(), which is an absurd. Then either v() = 0 or v() = 0
and so v() = 1, by (vCiw) (which is satisfied by v, see item 1). In virtue of the
completeness of mbCci, it follows that mbCci for every .
(3) Let v be an mbCciw-valuation such that v() = 1 for every , and let be a
formula such that v() = 1. By the hypothesis on v and (vCon), it follows that
either v() = 0 or v() = 0, and so v() = 0. By (vCiw) and contraposition,
v() = v() = 1, showing that v satisfies (vCi). The converse is a consequence of
Items 1 and 2. 

Remark 3.1.11 As mbCci extends mbCciw, the truth-value of is entirely deter-

mined by the truth-values of and in mbCci. Additionally, the truth-value of
is also determined by the truth-values of and in mbCci. Indeed, for every
mbCci-valuation v, the following holds:

v() = 1 iff v() = v();

v() = 1 iff v() = v() .

The situation in mbCci concerning contradiction, (in)consistency and non-

contradiction is therefore as follows:
68 3 Some Extensions of mbC

( )
1 1 0 1 1 1 v1
0 v2
0 1 0 0 1 v3
0 1 1 0 0 1 v4

Observe that, in mbCci, is equivalent to . On the other hand, is not

equivalent to ( ), because of valuation v1 (the countermodel v1 also holds
for mbCciw).
Corollary 3.1.12 The logic mbCci is a (strong) LFI.
The fact that is not equivalent to ( ) in mbCci can be seen from two
perspectives: there is a more pessimistic point of view which would consider this to be
a bad feature, since it shows the weakness of mbCci. There is also a more optimistic
point of view which considers this to be a positive feature, as it demonstrates the
ability of mbCci to separate the notions of consistency and non-contradictoriness,
while also identifying non-consistency (or inconsistency) with contradictoriness. In
Sect. 3.3, some LFIs will be analyzed in which the consistency operator can be
defined in terms of other connectives of its language.

3.2 A Hidden Consistency Operator

As was discussed in Sect. 2.4 of Chap. 2, there is a hidden classical negation within
mbC: for each formula ( p) which depends exclusively on the propositional vari-
able p, the formula () = () behaves like a classical negation in mbC,
where = ( ) is a bottom formula, for every formula . In Sect. 2.5
of Chap. 2, the possibility of considering a single (and so a single ) was ana-
lyzed, arriving to the logic mbC , which is an alternative presentation of mbC in the
signature {, , , }. In this section we will show that there is another interest-
ing hidden operator in mbC: an alternative consistency operator ( p) , one for each
formula ( p) which depends exclusively on the propositional variable p. As we
shall see in Sect. 3.4, this operator will establish an important conceptual difference
between mbC and mbC .
Definition 3.2.1 Let ( p) L be a formula which depends exclusively on the
propositional variable p, and recall the strong negation ( p) defined through it in
mbC. The strong ( p)-consistency operator, denoted by ( p) (or simply by ), is
defined in mbC by means of the following formula:

( p) p = ( p p) ( p p) .
def


3.2 A Hidden Consistency Operator 69

Given that ( p) p = ( p p) ( p p) , it is clear that, for every L :

() = () ( ) = ( ) () .

As was done in Sect. 2.4 in order to simplify notation, from now on we will write
, and instead of () , () and () , respectively, for every formula
. Thus, we can write = ( ).
The semantics of and its relationship with the original consistency operator
and with other related formulas is displayed in Table 3.2.
Table 3.2
( )
1 0 v1
1 0 1
0 0 v2
1
1 0 1 1 v3
0
0 0 1 1 v4
1 0 1 1 v5
0 1
0 0 1 1 v6

Theorem 3.2.2 Let be a formula which depends exclusively on the propositional

variable p.
(i) Let ( p) ( p) = {( p) p}, for a propositional variable p. The logic mbC is a
strong LFI (w.r.t. and ( p) ( p)) (recall Definition 2.1.9).
(ii) The logic mbC satisfies axiom (ciw ), namely: mbC ( ) for every
formula in L . That being so, the truth-value of is determined by the truth-
values of and as follows: v( ) = 1 iff v() = v() (iff either v() = 0 or
v() = 0), for every mbC-valuation v.
(iii) mbC but in general mbC . Thus, the consistency oper-
ator is different from the original .
(iv) mbC ( ) but in general mbC ( ) .
(v) mbC ( ) but in general mbC ( ).
Proof The proof will be done by means of mbC-valuations, taking into account the
soundness and completeness theorems for mbC w.r.t. valuations.
(i) The proof is entirely analogous to that of Proposition 2.3.2, but now by analyzing
Table 3.2. The details of the proof are left as an exercise.
(ii) Let = ( ) be an instance of (ciw ) and let v be a valuation. If
v( ) = 1 then v() = 1, by (vOr). Otherwise, if v( ) = 0 then v( ) = 1,
by the definition of and the properties of . Hence v() = 1, by (vOr).
(iii) and (iv): By observing Table 3.2 it is clear that the first part of each item is true.
Now, assume that is a propositional variable. The second part of item (iii) holds
when considering valuation v4 on the same table. In turn, valuation v1 shows that the
second part of item (iv) holds.
70 3 Some Extensions of mbC

(v) In Table 3.2, observe that v1 is the unique model for , and v1 ( ) = 0.
Then, by clause (vNeg) for mbC-valuations, v1 ( ) = 1. From this, it fol-
lows that mbC ( ) . For the second part, consider again Table 3.2
(where is a propositional variable) and extend any valuation v within v3 v6 by
imposing that v( ) = 1 (observe that this kind of extension is always possi-
ble for mbC-valuations). Then v( ) = 1 but v( ) = 0, so mbC
( ). 

The theorem above states that there are alternative consistency operators in
mbC, all of them inequivalent to . It is also clear that, for every ( p) and  ( p),
 for every . However, this logical equivalence is not as strong as it
seems, recalling the fact that mbC is not self-extensional (see the end of Sect. 2.3
of Chap. 2): for instance, ( p) p   ( p) p, if ( p) =  ( p). On the other hand,
it is easy to see that by considering the extension mbCciw of mbC introduced in
Sect. 3.1, equivalence between and can be obtained:

Proposition 3.2.3 The formulas and are equivalent in mbCciw, for every
and ( p). Moreover, mbCciw is the minimal extension of mbC such that and
are equivalent.2

Proof The proof can be done by using mbCciw-valuations because of

Theorem 3.1.4. In virtue of Theorem 3.2.2, mbC and hence this holds in
mbCciw also. Now, let v be an mbCciw-valuation such that v( ) = 1. Then, by
Definition 3.2.1 and the properties of , v( ) = 0 and then either v() = 0 or
v() = 0. From clause (vCiw), v() = 1 whence mbCciw , as required.
Now, let L be an extension of mbC such that and are equivalent. Then
L ( ) and, by contraposition (which is satisfied by the classical
negation ), it follows that L ( ). But L (
) ( ), and then L ( ). From this, and by observing
that is equivalent to in mbC (and so in L ), it follows that
L ( ). This shows that L is an extension of mbCciw. 

However, and ( ) are still inequivalent in mbCciw. As was done

with mbCci, this situation can be fixed by adding a suitable axiom.

Definition 3.2.4 Let be a formula which depends exclusively on the propositional

variable p. Let mbCciwci be the logic obtained from mbCciw by adding the fol-
lowing axiom schema:
( ) (ci )

An mbCciw-valuation v is said to be an mbCciwci -valuation if it also satisfies the

following clause:

2 As it was done above, to simplify notation, the index will kept fixed in the statement and in the

proof of this proposition. Of course this abuse of notation does not affect the validity of the claims,
because of the properties of () , () and () .
3.2 A Hidden Consistency Operator 71

(vCi) v( ) = 1 = v() = 1 and v() = 1.



Theorem 3.2.5 (Soundness and Completeness of mbCciwci w.r.t. valuations) For

every  {} L :

 mbCciwci  mbCciwci .

Proof Left as an exercise to the reader. 

Theorem 3.2.6 Let be a formula which depends exclusively on the propositional

variable p. The formulas and are equivalent in mbCciwci , for every formula
. Moreover, the formula is equivalent to in mbCciwci , for every
formula .

Proof It is enough to observe the following table obtained for mbCciwci :

1 1 0 1 0 1 v1
0 1 0 1 0 v2
0 1 1 0 1 0 v3


Corollary 3.2.7 Let be a formula which depends exclusively on the proposi-

tional variable p, and let  ( p) = { p}. The logic mbCciwci is a strong LFI
(w.r.t. and  ( p)).

In the following, the properties of and in the different extensions of

mbC will be analyzed under the perspective of dC-systems, which will be introduced
in the next section.

3.3 Consistency and Inconsistency as Derived Connectives

In his famous hierarchy Cn (for 1 n < ) of paraconsistent systems, da Costa

defined what he called well-behavedness operators (nowadays identified as one of
the many senses of consistency) for each system Cn in terms of the paraconsistent
negation and conjunction. For instance, in C1 , the well-behavedness (or consis-
tency) operator is given by = ( ) (see Sect. 3.7). The use of an operator
def

to recover the law of explosion (and so, the whole classical logic) in a controlled
way is a distinctive feature of da Costas systems, which boldly differentiates his
approach to paraconsistency from the others.
The original proposal of LFIs in [3] intended to deeply generalize da Costas
approach by introducing weaker systems in which the consistency operator is, in
72 3 Some Extensions of mbC

principle, primitive instead of being defined in terms of the others as in the case of
the well-behavedness operators. The consistency operator of mbC, as well as its
axiomatic extensions studied in the literature (in particular, in this book), is closed
related to that of da Costas system C1 . In Sect. 3.1, the relationship between and
( ) in the systems mbC, mbCciw and mbCci was analyzed, showing that
both formulas are inequivalent, while the former is logically stronger than the latter.
By turn, and are inequivalent in mbC and mbCciw, but the latter implies
the former, being equivalent in the strongest system mbCci. The main goal of this
section is to find minimal extensions of mbC in order to guarantee the equivalence
between consistency and non contradiction, on the one hand, and inconsistency and
contradiction, on the other. This is related to the general problem of considering
(minimal) LFIs in which the consistency and/or the inconsistency connectives are
definable in terms of the others, that is, they are derived connectives.
In [3], an special type of LFIs called dC-systems was introduced, which is char-
acterized by the fact that the consistency operator can be defined in terms of the other
primitive operators. The systems Cn are the canonical examples of dC-systems (see
Theorem 3.7.8). In this section some dC-systems will be briefly analyzed, showing
in Sect. 3.4 that, surprisingly, the logic mbC (the version of mbC in a different
language which was analyzed in Sect. 2.5 of Chap. 2) can be seen as a dC-system in
a weaker sense (see Propositions 3.4.6, 3.4.7 and 3.4.8).
The notion of dC-system is based on the concept of C-system. Both notions were
proposed in [3] and reformulated in [1] (Definition 27, p. 23 and Definition 32, p. 27,
respectively). A new version of both definitions will be proposed below, which is
more appropriate for the purposes of this section.

Definition 3.3.1 Let L2 = 2 , 2  be a Tarskian and structural logic (recall Def-

inition 2.1.4), and let 1 be a nonempty signature contained in signature 2 . The
logic L2 is said to be a reducible to 1 if, for every connective # 2 1 there
exists a formula # ( p1 , . . . , pn ) L1 (where n is the arity of #) such that:
(i) #( p1 , . . . , pn ) 2 # ( p1 , . . . , pn ) and # ( p1 , . . . , pn ) 2 #( p1 , . . . , pn );
(ii) let t : L2 L1 be a mapping recursively defined as follows:
1. t ( p) = p if p V ar ,
2. t (#(1 , . . . , n )) = #(t (1 ), . . . , t (n )), if # 1 ,
3. t (#(1 , . . . , n )) = # (t (1 ), . . . , t (n )), if # 2 1 .
Then for every  {} L2 ,

 2 = t[] 2 t () .

Remark 3.3.2 Let L2 be a logic reducible to 1 , and let

 
1 = 2 (L1 ) L1 .
def
3.3 Consistency and Inconsistency as Derived Connectives 73

Then L1 = 1 , 1  is a Tarskian and structural logic such that L2 is a conservative

extension of L1 , and t is a translation from L2 to L1 . This means that L2 can be
interpreted in L1 via t.
It should be observed that, in general, L2 has a greater expressive power than L1 ,
in the sense that it can express more (inequivalent) formulas. This is the case when the
logic L2 is not self-extensional. See an example of this situation in Remark 3.3.43.

Example 3.3.3
(1) Let c = {, , , } be the signature of classical propositional logic CPL,
and let c be any proper subset of c containing and at least another connective
of c . Then CPL is reducible to c . Since CPL is self-extensional, the expressive
power of CPL is the same than that of any of the fragments in which it can be reduced.
This is a consequence of the fact that the mapping t of Definition 3.3.1 produces, in
this case, a conservative translation.
(2) The logic mbC is not reducible to + = {, , }. Otherwise, it would be
semantically characterized by a 2-valued logical matrix (namely, the characteristic
logical matrix of CPL+ ), contradicting Theorem 4.2.10 in Chap. 4, which states that
mbC is not characterizable by a finite matrix semantics.
(3) By a similar argument, is proved that the logic mbC is not reducible to W =
{, }. Observe that, by adapting the proof of Theorem 4.2.10 in Chap. 4, it can be
proved that mbC cannot be characterized by a finite matrix semantics.

Definition 3.3.4 Let Li = i , i  (i = 1, 2) be two Tarskian and structural logics

(recall Definition 2.1.4) such that 2 extends 1 , and 2 contains a unary connec-
tive not belonging to 1 . The logic L2 is said to be a C-system based on L1
w.r.t. and p (in short, a C-system) if:
(i) L2 is a conservative extension of L1 ,
(ii) L2 is an LFI with respect to and the set ( p) = { p},
(iii) cannot be defined in L1 , that is: there is no formula ( p) L1 such that
p 2 ( p) and ( p) 2 p,
(iv) L1 is non-trivial.


The logic L1 can be seen as a kind of positive basis (w.r.t. ) for the logic L2 ,
seen as an LFI (w.r.t. and some formula p). As an example of C-systems, it will be
proven that mbC is a C-system based on CPL+ , while mbC is a C-system based on
CPL (in Theorems 3.3.10 and 3.3.13 below, respectively). The notion of dC-system
can, thus, be formalized (with some differences with respect to the notions proposed
in [1, 3]) as follows:

Definition 3.3.5 Let Li = i , i  (i = 1, 2) be two Tarskian and structural logics.

We say that L2 is a dC-system based on L1 w.r.t. and p (or simply a dC-system)
if the following conditions hold:
74 3 Some Extensions of mbC

(i) L2 is a C-system based on L1 w.r.t. and p;

(ii) if p is defined by means of a primitive connective of 2 then L2 is reducible
to 2 { };
(iii) if p is not defined by a primitive connective of 2 :
(iii.a) there is no strict subformula ( p) of p such that L2 is an LFI
w.r.t. and ( p);
(iii.b) let 0 = {# 2 : # occurs inp} be the signature of the formula p;
def

hence, if there is some proper subset 0 of 0 and some set = ( p) L0

such that L2 is an LFI w.r.t. and ( p), then ( p) 2 p and p 2
for every ( p) (that is, p and ( p) are logically equivalent in L2 ).


This means that dC-systems are C-systems where either:

(1) p is defined by means of a primitive connective , which can be alternatively
defined in terms of the connectives in 2 { }; or
(2) p is not defined by means of a primitive connective, but the formula p is
non-degenerate in the sense of items (iii.a) and (iii.b) above.

Remark 3.3.6 At this point it is important to observe that, given a logic L with a
paraconsistent negation , L could be regarded as an LFI w.r.t. several sets ( p). In
particular, several consistency connectives could be defined over the same logic w.r.t.
the same paraconsistent negation (of course, several paraconsistent negations could
also be considered, varying together with the sets ( p)). This multiplicity of options
is naturally transferred to the notions of C-systems and dC-systems. In view of such
circumstances, Item (iii) of Definition 3.3.5 intends to avoid some pathological cases,
as we shall see below.
The idea behind the proposed notion of dC-systems is that, given a paraconsistent
negation in a logic L , if L is an LFI w.r.t. then there is a canonical or intended
way to define the notion of consistency, for instance by means of an specific connec-
tive (in the case of mbC) or by means of an specific schema formula (in the case of
the systems Cn , see Theorem 3.7.8). The latter constitutes a typical example of dC-
systems, while the former is a typical example of a logic which is not a dC-system
(with respect to its canonical structure of LFI).

To illustrate this discussion, let p = ( p) p = ( p p) ( p p), for a given

def

formula ( p) depending exclusively on the propositional variable p. This defines

a new consistency operator in mbC, as discussed in Sect. 3.2. However, because
of Item (iii.b) of Definition 3.3.5, mbC is not a dC-system w.r.t. this consistency
operator, given that the original consistency connective occurs (among others) in
p. That is, the formula p, defined in a signature properly contained in the signature
of p, is able to define a consistency operator in mbC which is inequivalent to p,
as proved in Theorem 3.2.2(iii).
A more drastic example can be given by considering  p = p p ( p = p
def def

p, respectively) in a C-system L where p is a consistency operator and is a

3.3 Consistency and Inconsistency as Derived Connectives 75

conjunction ( is a disjunction, respectively). Clearly, each of the two formulas above

defines an alternative consistency operator. However, and as expected, L is not a
dC-system w.r.t.  or w.r.t.  , because of Item (iii.a) of Definition 3.3.5: the strict
subformula p of  p defines a consistency operator in L w.r.t. . Analogously for
 p.
The three examples above ( p,  p and  p) can be viewed as degenerate forms
of defining consistency in terms of the other connectives. On the other hand, Item (iii)
of Definition 3.3.5 does not rule out the following case as a dC-system (related, but
essentially different, to the case of  p mentioned above): suppose that L is an LFI
w.r.t. and p = 1 p 2 p such that is a conjunction. Suppose also that
def

does not occur in ( p) = {1 p, 2 p}. Then ( p) is an alternative way to define

def

consistency in L w.r.t. in a signature strictly contained in that of p. However,

the existence of ( p) does not preclude the possibility of L being a dC-system,
since ( p) is logically equivalent to p in L .
A variant of the last example is as follows: suppose that the conjunction connective
, which does not occur in {1 p, 2 p}, can be alternatively defined in L by means of
a formula ( p1 , p2 ) such that all the connectives in already occur in {1 p, 2 p}.
This means that  p = (1 p, 2 p) is an alternative way to define a consistency
def

operator in L w.r.t. in a signature strictly contained in that of p. However, p

and  p are logically equivalent in L and so this logic still has the possibility of
being a dC-system, by Item (iii.b) of Definition 3.3.5.
In another variant of this example, suppose now that p = (1 p, 2 p) is a
def

consistency operator in L , where ( p1 , p2 ) is an alternative way to define in L the

conjunction in terms of connectives which do not occur in {1 p, 2 p}. Suppose
also that the conjunction does occur in {1 p, 2 p}.3 Then p = 1 p 2 p is an
def

alternative way to define a consistency operator in L in terms of a proper subset

of the connectives occurring in p. However, p is equivalent to p and so, by
Item (iii.b) of Definition 3.3.5, L still can be a dC-system w.r.t. and p.
It is easy to prove the following property of dC-systems:
Proposition 3.3.7 Let L2 be a dC-system based on L1 w.r.t. and p such that is a
primitive connective of 2 . Assume that L2 , and p satisfy conditions (i) and (ii)
 
of Remark 2.1.10. Let 3 = 2 { }, and let 3 = 2 (L3 ) L3 .
def def

Then L3 = 3 , 3  is a C-system based on L1 w.r.t. and ( p), where ( p) is

a formula in L3 associated to as in Definition 3.3.1.
Proof By the very definitions, it is immediate that L3 is a (strong) LFI
w.r.t. and ( p) satisfying conditions (i) and (ii) of Remark 2.1.10. Thus, condi-
tion (ii) of Definition 3.3.4 is verified. On the other hand, since 1 3 , it follows
that L1 L3 and so
      
1 = 2 (L1 ) L1 = 2 (L3 ) L3 (L1 ) L1 .

3 As a concrete example of this situation, consider the expansion C2 of da Costas system C2 (see
Sect. 3.7) obtained by adding a primitive connective for the classical negation , where 1 p and
2 p are formulas which only use and , and where the formula ( p1 , p2 ) is ( p1 p2 ).
76 3 Some Extensions of mbC
  
Now, 3 = 2 (L3 ) L3 by definition, whence 1 = 3 (L1 )
L1 . This shows that L3 is a conservative extension of L1 and so condition (i)
of Definition 3.3.4 holds. Clearly 3 1 , and cannot be defined in L1
according to condition (iii) of Definition 3.3.4. 

In order to prove that mbC is a C-system based on CPL+ , some previous results
must be obtained:

Proposition 3.3.8 The logic mbC is a conservative extension of CPL+ . That is: for
every  {} L+ ,

 CPL+  mbC .

Proof Let  {} L+ . If  CPL+ then clearly  mbC , since mbC extends
CPL+ . Now, suppose that  mbC . Then t[] CPL t (), by Proposition 2.4.5.
By definition of t, t () = for every L+ and so  CPL . Since CPL is a
conservative extension of CPL+ , it follows that  CPL+ . 

Proposition 3.3.9 The paraconsistent negation of mbC cannot be defined in

CPL+ , the + -fragment of mbC. That is: there is no formula ( p) L+ such
that p mbC ( p) and ( p) mbC p.

Proof The proof is by reductio ad absurdum. Let p be a propositional variable. Sup-

pose that there is a formula ( p) L+ such that p mbC ( p) and ( p) mbC
p. Let v be an mbC-valuation such that v( p) = 1. Since the only connectives occur-
ring in are , and then v(( p)) = 1. By completeness of mbC w.r.t. valu-
ations it follows that p mbC ( p). Since ( p) mbC p then p mbC p. Given
that p mbC p, it follows that mbC p, by Item (iii) of Proposition 2.1.14. Thus
v( p) = 1 for every mbC-valuation v. In particular, this holds for any CPL-valuation
v in which v( p) = 0, a contradiction. 

Theorem 3.3.10 The logic mbC is a C-system based on CPL+ w.r.t. and p.

Proof It is a consequence of Theorem 2.3.2, Propositions 3.3.8 and 3.3.9 and the
obvious fact that CPL+ is not trivial. 

Using a similar argument, it is possible to prove that mbC is a C-system based on

CPL (presented in a suitable signature), see Theorem 3.3.13 below. This establishes
an important conceptual diference between mbC and mbC , which will be analyzed
in Sect. 3.4.
Recall from Proposition 2.5.5 the axiomatization CPLW of CPL in the signature
W = {, }. The following holds:

Proposition 3.3.11 The logic mbC is a conservative extension of CPLW . That is:
for every  {} LW ,

 CPLW  mbc .
3.3 Consistency and Inconsistency as Derived Connectives 77

Proof Let  {} LW . If  CPLW then clearly  mbC , since mbC is an

extension of CPLW . Conversely, suppose that  mbC . Let CPLW be the system

defined over W = {, , } obtained from CPLW by adding the axiom schema
. Clearly, CPLW is sound and complete w.r.t. the classical 2-valued truth-tables for
the connectives in W with the addition of the truth-table for considered for CPL
after Definition 2.4.4. Now, let t  : L LW be the mapping which just replaces
by (that is, t  () = t  () ). By adapting the proof of Proposition 2.4.5,
it follows that t  is a translation from mbC to CPLW . Thus, t  [] CPLW t  (). By
definition of t  , t  () = for every LW , hence  CPLW . It is easy to prove,
by using its semantics, that CPLW is a conservative extension of CPLW . From this,
 CPLW . 

Proposition 3.3.12 The paraconsistent negation of mbC cannot be defined in

CPLW , that is: there is no formula ( p) LW such that p mbC ( p) and ( p)
mbC p.

Proof Let p be a propositional variable. Suppose by contradiction that there is a

formula ( p) LW such that p mbC ( p) and ( p) mbC p. Given that
the mbC -valuation semantics restricted to L  W coincides
 with the classical 2-
valued truth-tables, there exists a function F : 0, 1 0, 1 such that v(( p)) =
F(v( p)) for every mbC -valuation v. Suppose that F(1) = 1. This means that, for
every mbC -valuation v, if v( p) = 1 then v(( p)) = 1 and so, by completeness of
mbC , p mbC ( p). From this, p mbC p. On the other hand, p mbC p,
by () and MP. This means that, for every mbC -valuation v, v( p) = 1. In
particular, for every valuation v for CPL (in which is interpreted as ) such
that v( p) = 0, it holds that v( p) = 1, a contradiction. Therefore F(1) = 0 and so
v( p) = 0 whenever v( p) = 1. Equivalently, if v( p) = 1 then v( p) = 0. Thus,
v( p) = 1 implies that v( p) = 1. That is, p mbC p and so p and p are
equivalent in mbC , a contradiction. 

Theorem 3.3.13 The logic mbC is a C-system based on classical logic CPLW
w.r.t. and p.

Proof Observe that, since mbC is equivalent (up to translations) to mbC, then it is
a (strong) LFI w.r.t. and p. The rest of the proof follows from Propositions 3.3.11
and 3.3.12, and the obvious fact that CPLW is not trivial. 

On the other hand, neither mbC nor mbC are dC-systems, as it will be proved
below in Propositions 3.3.20 and 3.3.8.
To prove this, recall from Definition 2.1.3 the signature 0 = {, , , } and let
mbC be the 0 -fragment of mbC obtained by dropping axiom schema (bc1) while
considering all the other axiom schemas and rules of mbC defined over the signature
0 . It is worth observing that mbC is nothing else than logic PI, introduced by
Batens in [8], and afterwards renamed as CLuN by the author (this logic will be
taken up again in Sect. 5.4). In [8] (Theorems 1 and 2, p. 205) a result equivalent to
the following was proved:
78 3 Some Extensions of mbC

Proposition
 3.3.14
 The logic mbC is sound and complete w.r.t. valuations v :
L0 0, 1 satisfying clauses (vAnd), (vOr), (vImp) and (vNeg) of
Definition 2.2.1.

The interested reader can prove the result above by adapting the proof of Theo-
rems 2.2.2 and 2.2.8.
In order to prove that mbC conservatively extends mbC , some previous defini-
tions and results must be established.
Consider a new complexity measure l : L N over the set L defined as fol-
lows:

(i) l() =1if V ar ;
1 if L0
(ii) l() =
l() + 1 otherwise

1 if L0
(iii) l() =
l() + 2 otherwise
(iv) for # {, , }, 
l(#) = 1 if # L0
.
l() + l() + 1 otherwise

Observe that l() = 1 if L0 . A suitable variant of Lemma 2.5.9 can be stated:
 
Lemma 3.3.15 Let v0 : L0 0, 1 be an mbC -valuation. Then there exists a
valuation v for mbC extending v0 .

Proof For L , the values v() are defined by induction on l() as follows:

if l() = 1 then the valuation v is defined as v() = v0 (), if L0 , and v()
is arbitrary, if V ar . Suppose now that the valuation v is already defined for
 ) < n, where n > 1 (induction hypothesis), and let such that
all  such that l(
l() = n (thus / L0 V ar ). Then the definition of v goes as follows:
1. If = , then: 
1 if v( ) = 0, or
v( ) =
arbitrary otherwise

2. If = , then:

0 if v( ) = v( ) = 1, or
v( ) =
arbitrary otherwise

3. If = #, then v() is defined according to clauses (vAnd), (vOr) and (vImp)

of Definition 2.2.1.
It is easy to prove that v
It is worth noting that v is well-defined, by definition of l.
is an mbC-valuation extending v0 . Details are left to the reader. 

Proposition 3.3.16 The logic mbC is a conservative extension of mbC .

3.3 Consistency and Inconsistency as Derived Connectives 79

Proof Let  {} L0 . If  mbC then obviously  mbC . Conversely, if

 mbC then  |=mbC . Let v be an mbC -valuation such that v[] {1}. By
Lemma 3.3.15, the mbC -valuation v can be extended to an mbC-valuation v ,
whence v [] {1}. From this, v () = 1 and so v() = 1, since v extends v. This
means that  |=mbC and so, by Proposition 3.3.14,  mbC . 

A fundamental property of mbC is that this system is unable to define a bottom

formula, that is , for no formula ( p) over 0 depending exclusively on a proposi-
tional variable p, it is the case that ( p) mbC for every (see Proposition 3.3.18
below). The proof employs a technique analogous to that of Lemma 3.3.15.
Given a propositional variable p, let L0p be the set of formulas in L0 which
depends exclusively on a propositional variable p. A new complexity measure l :
L0 N over the set L0 can be defined as follows:

(i) l() =1 if V ar ;
1 if L0p
(ii) l() =
l() + 1 otherwise
(iii) for # {, , }, 
l(#) = 1 if # L0p
.
l() + l() + 1 otherwise

Observe that l() = 1 if L0p . A variant of Lemma 3.3.15, which will be
useful for our purposes, can now be obtained:
 
Lemma 3.3.17 Let v0 : L0p V ar 0, 1 be a mapping such that v0 () = 1 for
any L0p , where p is a propositional variable. Then there exists a valuation v
for mbC extending the mapping v0 .

Proof For L0 , the values v() are defined by induction on l() as follows:

if l() = 1 then the valuation v is defined as v() = v0 (). Suppose now that the
 ) < n, where n > 1 (induc-
valuation v is already defined for all  such that l(

tion hypothesis), and let such that l() = n (hence / L0p V ar ). Then the
definition of v goes as follows:
1. If = , then: 
1 if v( ) = 0, or
v( ) =
arbitrary otherwise

2. If = #, then v() is defined according to clauses (vAnd), (vOr) and (vImp)

of Definition 2.2.1.
It is easy to prove that v
It is worth noting that v is well-defined, by definition of l.
is an mbC -valuation extending v0 . Details are left to the reader. 

Proposition 3.3.18 It is not possible to define a bottom formula in mbC . That is:
there is no formula ( p) L0 (depending exclusively on a propositional variable
p) such that ( p) mbC , for every L0 .
80 3 Some Extensions of mbC

Proof Let ( p) be a formula in L0p , for a given propositional variable p, and let q
be a propositional variable different from p. Consider a mapping v0 : L0p V ar
 
0, 1 such that v0 () = 1 for any L0p , and v0 ( p  ) = 0, for any p  V ar { p}.
 
By Lemma 3.3.17, the mapping v0 can be extended to a valuation v : L0 0, 1
for mbC . The valuation v obtained in such a way shows that ( p) |=mbC q and
so, by Proposition 3.3.14, ( p) mbC q. 

Proposition 3.3.19 The consistency operator cannot be defined in mbC over the
signature 0 , and so mbC is not reducible to 0 .

Proof Suppose that ( p) L0 is a formula over 0 depending on the propositional

 
variable p such that ( p) is equivalent to p in mbC. Let v0 : L0p V ar 0, 1
be a mapping such that v0 () = 1 for any L0p , and v0 ( p  ) = 0, for every p 

 Lemma 3.3.17, the mapping v0 can be extended to a valuation v :
V ar {p}. By
L0 0, 1 for mbC . Let q be a propositional variable different from p. Since
p, p, p mbC q then p, ( p), p mbC q. By Proposition 3.3.16, it follows that
p, ( p), p mbC q and so p, ( p), p |=mbC q. But the mbC -valuation v
constructed above contradicts the latter assertion. Therefore, p cannot be defined
in mbC over the signature 0 . 

Proposition 3.3.20 The logic mbC is not a dC-system based on CPL+ w.r.t. and
p.

Proof It follows from Proposition 3.3.19. 

Concerning mbC , let W = {, , } and let mbC be the W -fragment of

mbC obtained by dropping axiom schema (bc1 ) while considering all the other
axiom schemas and rules of mbC (recall Definition 2.5.3) written over the signature

W .

Proposition
 3.3.21
 The logic mbC is sound and complete w.r.t. valuations v :
LW 0, 1 satisfying clauses (vBot), (vImp) and (vNeg) of Definition 2.5.8.

Proof The proof is left to the reader. 

Proposition 3.3.22 The logic mbC is a conservative extension of mbC

.

Proof The proof is analogous to that of Proposition 3.3.16, and is left as an exercise
to the reader. 

In order to prove that the consistency operator cannot be defined in mbC over

the signature W , a technique concerning extensions of valuations, similar to that
used in the proof of Lemma 3.3.15, will be used. Thus, consider a new complexity
measure l : L N over the set L defined as follows:

(i) l() =1
if V ar {};
1 if LW
(ii) l() =
l() + 1 otherwise
3.3 Consistency and Inconsistency as Derived Connectives 81

1 if LW
(iii) l() =

l() + 2 otherwise

) = 1 if LW
(iv) l( l() + l()
.
+ 1 otherwise

Observe that l() = 1 if LW . A variant of Lemma 3.3.15 will be stated now:
 
Lemma 3.3.23 Let v0 : LW 0, 1 be an mbC -valuation. Then there exists a
valuation v for mbC extending v0 .

Proof The proof is similar to the one given in Lemma 3.3.15, but now using induction
on the complexity measure l. There is just one additional case to consider: if =
then v() = 0. Clearly, v is well-defined, by definition of l.
def
It is easy to prove that
v is an mbC -valuation extending v0 . Details are left to the reader. 

Proposition 3.3.24 The consistency operator cannot be defined in mbC over the

signature W , and so mbC is not reducible to W .

Proof Suppose that ( p) LW is a formula over W depending on the proposi-
tional variable p such that ( p) is equivalent to p in mbC . Since p, p mbC q
(if q V ar is different from p) then p, ( p) mbC q and so p, ( p) mbC q,
by Proposition 3.3.22. By Proposition 3.3.21, there exists an mbC -valuation v
such that v( p) = 1, v(( p)) = 1 but v(q) = 0. Given that p, p, p mbC q then
p, ( p), p mbC q and so p, ( p), p mbC q, by Proposition 3.3.22. Using
again Proposition 3.3.21 it follows that v( p) = 0, by the hypothesis about v. By
Lemma 3.3.23, the mbC
-valuation v can be extended to an mbC -valuation v .

  
Given that v ( p) = 1 and v ( p) = 0 then the truth-value v ( p) can be defined
arbitrarily; thus, define v ( p) = 0. Taking into account that v extends v, it follows
that v (( p)) = 1. This shows that ( p) |=mbC p. By completeness of mbC ,
( p) mbC p, a contradiction. Therefore, p cannot be defined in mbC over the

signature W . 

Proposition 3.3.25 The logic mbC is not a dC-system based on CPLW w.r.t.
and p.

Proof It is a consequence of Proposition 3.3.24. 

Despite neither mbC nor mbC being dC-systems, the latter can be regarded as
a dC-system under a somewhat different notion, as Sect. 3.4 will show.
From Proposition 3.3.20 it follows that the canonical structure of mbC as an LFI
fails to produce a dC-system. An interesting question is to determine the minimal
extension of mbC which is a dC-system w.r.t. its original LFI structure. We begin by
semantically analyzing which restrictions applicable to mbC would suffice to obtain
the equivalence between and ( ):
82 3 Some Extensions of mbC

( )
1 1 0 1 1 v1
0 v2
0 1 0 1 v3
0 0 1 v4
0 1 1 0 1 v5
0 0 1 v6

Observe that v1 , v4 and v6 are the only models in which ( ) does not entail
(recalling that entails ( ) in mbC). Thus, it is clear that adding the
requirement ( ) to mbC suffices to establish the equivalence between
( ) and and blocks the three valuations mentioned above. It is interest-
ing to note that the proposed condition imposes restictions on the valuations with
respect to the negation , on the one hand, and to the consistency , on the other.
Indeed, in order to eliminate v1 , we can argue as follows: if v() = v() (and so
v() = 0) then v( ) = 1 and hence v(( )) should be 0. Cleary, this
requirement concerns the negation . And, in order to eliminate v4 and v6 , note that if
v() = v() (and so v(( )) = 1) then v() should be 1. This constitutes
a requirement for the consistency operator .
This justifies the following:

Definition 3.3.26 Let mbCcl be the logic obtained from mbC by adding the fol-
lowing axiom schema:
( ) (cl)

An mbCcl-valuation is an mbC-valuation v which also satisfies the following clause:

(vCl) v(( )) = 1 = v() = 1.

Let V mbCcl be the set of mbCcl-valuations.



Theorem 3.3.27 Let  {} L, with

  maximal non-trivial with respect to
in mbCcl. The mapping v : L 0, 1 defined by:

v() = 1 

for all L , is a valuation for mbCcl.

Proof By Theorem 2.2.7, the mapping v is an mbC-valuation. Thus, it is enough

to prove that v also satisfies clause (vCl). Let be a formula such that v((
)) = 1. Given that mbCcl ( ) by (cl), then ( ) ,
by Lemma 2.2.5 and v(( ) ) = 1 accordingly. From this, v() = 1
by (vImp), and then v is an mbCcl-valuation. 

Theorem 3.3.28 (Soundness and Completeness of mbCcl w.r.t. valuations) For

every  {} L :
3.3 Consistency and Inconsistency as Derived Connectives 83

 mbCcl  mbCcl .

Proof This proof is analogous to that for mbC, but it now requires Theorem 3.3.27
instead of Theorem 2.2.7. 

Next result is the first step towards proving that mbCcl is the least extension of
mbC which is a dC-system w.r.t. its canonical LFI structure:

Corollary 3.3.29 For every mbCcl-valuation v and every formula :

v(( )) = v() .

This implies that ( ) and are logically equivalent in mbCcl. Moreover,

this is the minimal extension of mbC which satisfies this property.

The consistency operator of the system mbCcl enjoys an additional and funda-
mental property:

Corollary 3.3.30 For every formula in L : mbCcl ( ). Thus, the

logic mbCcl extends the logic mbCciw. Moreover, mbCciw is properly contained
in mbCcl.

Proof Given the soundness and completeness of both logics with respect to their
valuations, it is enough to prove that any mbCcl-valuation is an mbCciw-valuation.
Thus, let L and v be an mbCcl-valuation such that v() = 0. Accordingly
to the last corollary, this means that v(( )) = 0 and then, by clause (vNeg),
v( ) = 1. This shows that v is an mbCciw-valuation, and so mbCciw is con-
tained in mbCcl.
In order to see that the inclusion is proper, let p be a propositional variable
and let v be an mbCciw-valuation such that v( p) = v( p) = 1. Then v( p) = 0
and v( p p) = 1. However, nothing forbids having v(( p p)) = 1, and so
v(( p p) p) = 0. This shows that (cl) is not a valid schema in mbCciw,
whence mbCciw is properly contained in mbCcl. 

The situation in mbCcl is as follows:

( )
1 1 0 1 1 0 0 1 v1
0 1 0 1 1 1 1 v2
0 v3
0 1 1 1 v4
0 v5
0 1 1 0 1 1 1 1 v6
0 v7
0 1 1 1 v8
0 v9
84 3 Some Extensions of mbC

Corollary 3.3.31 The logic mbCcl is a (strong) LFI.

From the table above, it is easy to see that, in mbCcl, any consistency operator
of the form p, as well as p itself, collapse to ( p p):
Corollary 3.3.32 In mbCcl the formulas , and ( ) are logically
equivalent for every formulas and ( p).
The next step is proving that mbCcl is a C-system based on CPL+ . The fol-
lowing two results can be proved in an analogous way to the case for mbC, see
Propositions 3.3.8 and 3.3.9. Details are left to the reader.
Proposition 3.3.33 The logic mbCcl is a conservative extension of CPL+ .
Proposition 3.3.34 The paraconsistent negation of mbCcl cannot be defined in
CPL+ , the + -fragment of mbCcl. That is: there is no formula ( p) L+ such
that p mbCcl ( p) and ( p) mbCcl p.
As a consequence of this:
Theorem 3.3.35 The logic mbCcl is a C-system based on CPL+ w.r.t. and p.
Finally, in order to prove that mbCcl can be reduced to its consistency-less frag-
ment (proving so that it is a dC-system), it is necessary to define the counterpart of
the system mbCcl in the signature 0 which does not use the consistency operator
, which bring us to a very interesting system.
Definition 3.3.36 (mbCcl ) Recall the signature 0 = {, , , }. The calculus
mbCcl , defined over the language L0 , is obtained from mbC by replacing axiom
schema (bc1) with the following:
  
( ) (bc1 )

while the other axiom schemas and rules are considered over L0 . 
Equivalently, mbCcl is the system obtained from CPL+ by adding axiom
schemas (Ax10) and (bc1 ).  
The semantics of mbCcl -valuations is formed by valuations v : L0 0, 1
satisfying all the clauses of Definition 2.2.1 with the exception of (vCon), which is
replaced by the following clause:
(vCon ) v(( )) = 1 = v() = 0 or v() = 0.
It is now immediate to prove the following:
Theorem 3.3.37 (Soundness and Completeness of mbCcl w.r.t. valuations) For
every  {} L0 :

 mbCcl  mbCcl .
3.3 Consistency and Inconsistency as Derived Connectives 85

A translation mapping from mbCcl to mbCcl will now be defined according to

Definition 3.3.1, Item (ii):

Definition 3.3.38 The mapping : L L0 is recursively defined as follows:

p = p if p V ar ;
(#) = (( )#( )) for # {, , };
() = ( );
() = (( ) ( )).


Theorem 3.3.39 The function is a translation from mbCcl to mbCcl . That is:
if  mbCcl then  mbCcl .

Proof The proof will be analogous to that of Theorem 3.5.11. Thus, it is easy to see
that the only instance of an axiom of mbCcl, such that is not an instance of an
axiom of mbCcl , is when

= ( )

is an arbitrary instance of axiom (cl). Then,

= (( ) ( )) (( ) ( ))

which is not an instance of an axiom of mbCcl . However, it is obviously a theorem

of mbCcl , given that is derivable in mbCcl for any formula . 

Additionally, we have the following:

Theorem 3.3.40 The system mbCcl is a conservative extension of mbCcl . That

is: for every  {} L0 ,

 mbCcl  mbCcl .

Proof (Only if part:) Observe that every axiom schema of mbCcl , with the
exception of (bc1 ), is an axiom schema of mbC and thusly of mbCcl (all these sys-
tems have the same inference rule, MP). Concerning axiom schema (bc1 ), consider
the arbitrary instance

= ( ) ( ( ))

where , L0 . Consider now the following derivation in mbCcl:

86 3 Some Extensions of mbC

1. ( ) (hyp)
2. (hyp)
3. (hyp)
4. ( ( )) (bc1)
5. ( ) (cl)
6. (MP 1, 5)
7. ( ) (MP 6, 4)
8. (MP 2, 7)
9. (MP 3, 8)

This shows that {( ), , } mbCcl and so, by the deduction meta-

theorem, is derivable in mbCcl.
From here, it follows by induction on the length of the derivation of from  in
mbCcl , that: if  mbCcl , then  mbCcl .
(If part:) Let  {} L0 be such that  mbCcl . By Theorem 3.3.39,
 mbCcl . But it is immediate to see that if L0 , then = . There-
fore,  mbCcl . 

Corollary 3.3.41 For every  {} L :

 mbCcl implies  mbCcl .

Proof Suppose that  mbCcl . Therefore,  mbCcl in accordance with The-

orem 3.3.39. From this,  mbCcl , in accordance with Theorem 3.3.40. 

Proposition 3.3.42 The logic mbCcl is reducible to 0 =  { }.

Proof Item (i) of Definition 3.3.1 follows from Corollary 3.3.29. Item (ii) is just
Corollary 3.3.41. 

Remark 3.3.43 It is worth noting that the mapping is not a conservative translation
from mbCcl to mbCcl : let p be a propositional variable, and = ( p p)
( p p). Clearly, mbCcl . But = for = p ( p p), and
then mbCcl . On the other hand, mbCcl , which is easily checked by means of
mbCcl-valuations. This means that, despite being reducible to 0 , the logic mbCcl
has a greater expressive power than its 0 -fragment mbCcl : the former can express
the logical inequivalence between = p and = ( p p). By its turn,
def def

= is, consequently, logically indistinguishable in mbCcl .

Theorem 3.3.44 The system mbCcl is a dC-system based on CPL+ w.r.t. and
p.

Proof It is a consequence of Theorem 3.3.35 and Proposition 3.3.42. 

Corollary 3.3.45 The system mbCcl is a dC-system based on CPL+ w.r.t. and
( p p).
3.3 Consistency and Inconsistency as Derived Connectives 87

Proof Straightforward. Details are left to the reader. 

Remark 3.3.46 The results contained in Theorems 3.3.39 and 3.3.40 and in Corol-
lary 3.3.41 were obtained in Theorem 106 of [1] for the system Cil (to be introduced
in Definition 3.5.17 below). Subsequently, they constitute an improvement of that
theorem, since mbCcl is a considerably weaker system than Cil.

Clearly, mbCcl is the least extension of mbC defined over  which is a dC-system
based on CPL+ . Moreover, system mbCcl is the least dC-system defined over 0
based on CPL+ such that consistency can be defined by ( p p): starting from
positive classical logic expanded with a negation that satisfies tertium non datur,
the explosion law is controlled by the consistency operator defined in terms of these
connectives by means of the non-contradiction law:

, mbCcl in general, but , , ( ) mbCcl is always the case.

This is the basic principle of da Costas system C1 , when the term well-behaviorness
is used instead of consistency for denoting such operator. The only difference with
C1 is that the axiom schema and the propagation of consistency (or well-
behaviorness) axioms are dropped in mbCcl . The relationship in mbCcl between
contradiction and the (defined) consistency operator is shown in the table below.

( )
1 1 1 0
0 0 1
0 1 0 1

Finally, the dual question of defining inconsistency (meaning non-consistency,

where non stands for the paraconsistent negation) in terms of the other connectives
will be briefly analyzed.
Recall that axiom (ci) (see Definition 3.1.7) is the least requirement added to
mbC to obtain equivalence between and . Thinking about as an
inconsistency connective, the latter would identify inconsistency with contradiction.
This would constitute a dual (or complementary) notion of a dC-system, in which
the inconsistency operator (instead of that of consistency) can be expressed in terms
of the others. The relevance of defining an inconsistent operator instead of its dual
(a consistency operator) will be addressed in Sect. 3.5.
Recall the system mbCci introduced in Definition 3.1.7. Then, as a consequence
of the already stated definitions and results, the following is immediate:

Theorem 3.3.47 The logic mbCci is the least extension of mbC in which
and ( ), for every formulas and .
88 3 Some Extensions of mbC

Still concerning axiom (ci), another system that could be looked at this point is
the following:

Definition 3.3.48 Let mbCcicl be the logic obtained from mbCcl by adding the
axiom schema (ci) or, equivalently, the extension of mbCci obtained by
 adding
 the
axiom schema (cl). An mbCcicl-valuation is a mapping v : L 0, 1 which
is simultaneously an mbCci-valuation and an mbCcl-valuation. The set V mbCci
V mbCcl of mbCcicl-valuations will be denoted by V mbCcicl . 

The proof of the soundness and the completeness of mbCcicl w.r.t. valuations is
as usual:

Theorem 3.3.49 Let  {} L ,with maximal non-trivial with respect to

in mbCcicl. The mapping v : L 0, 1 , defined by:

v() = 1 

for all L , is a valuation for mbCcicl.

Theorem 3.3.50 (Soundness and Completeness of mbCcicl w.r.t. valuations) For

every  {} L :

 mbCcicl  mbCcicl .

Theorem 3.3.51 The logic mbCcicl is the least extension of mbC in which, for
every formula and , ( ) and ( ).

The situation in mbCcicl is as follows:

( )
1 1 0 1 1 0 0 1 v1
0 1 0 0 1 1 1 v2
0 v3
0 1 1 0 0 1 1 1 v4
0 v5

Corollary 3.3.52 The logic mbCcicl is a (strong) LFI.

Finally, we can again consider axiom (ci) with respect to the consistency operator
, that is, axiom (ci ) (recall Definition 3.2.4), in order to obtain two new systems:

Definition 3.3.53 Let be a formula which depends exclusively on the propositional

variable p.
3.3 Consistency and Inconsistency as Derived Connectives 89

(1) Let mbCcici be the system obtained from mbCciwci (recall Definition 3.2.4)
by adding axiom schema (cc) (recall Proposition 3.1.10) or, equivalently, the system
obtained from mbCci by adding axiom schema (ci ). The set of valuations associated
with mbCcici is V mbCciwci V mbCci .
(2) Let mbCcici cl be the system obtained from mbCcici by adding axiom
schema (cl) or, equivalently, the system obtained from mbCcicl by adding axiom
schema (ci ). The set of valuations associated with mbCcici cl is V mbCcici
V mbCcicl . 

Theorem 3.3.54 (Soundness and Completeness of mbCcici and mbCcici cl w.r.t.

valuations) Let L {mbCcici , mbCcici cl}. For every  {} L :

 L  L .

Theorem 3.3.55 Let be a formula which depends exclusively on the propositional

variable p.
(1) The logic mbCcici is the least extension of mbC in which, for every formula ,
and ( ).
(2) The logic mbCcici cl is the least extension of mbC in which, for every formula
, ( ) and ( ).

The situation in mbCcici is as follows:

( )
1 1 0 1 1 0 1 1 v1
0 v2
0 1 0 0 1 0 1 v3
0 1 1 0 0 1 0 1 v4

Accordingly, the situation in mbCcici cl is as follows:

( )
1 1 0 1 1 0 0 1 v1
0 1 0 0 1 1 0 v2
0 1 1 0 0 1 1 0 v3

Corollary 3.3.56 The logics mbCcici and mbCcici cl are (strong) LFIs.
90 3 Some Extensions of mbC

The relationship between the different extensions of mbC over the signature 
introduced thus far is displayed below, where the name of each edge corresponds to
the property being added.

mbCcici cl
}> _@@
} }} @@
@@
}} @@
() }
}}} @@
@@
}} @@
}} @@
}}
mbCcici mbCcicl
? ]<
}}> `AA
A ~~ <<<
} AA ~~ <<
}} AA ~~ <<
}} A ~
}} AA
A ~~~ <<
}}} AA ~~ () <<
<<
} } AA
A ~~~ <<
}} A ~ ~
mbCciwci mbCci
O mbCcl
hPPP
PPP oo oo7
PPP oo
PPP oo ooo
PPP oo
PPPP ooo()

PPP oo o
PPP ooo
P ooo
mbCciw O

mbC

The properties enjoyed by each system are displayed in the following table (Table 3.3):

Table 3.3

mbC N N N N N N

mbCciw Y N N N N N

mbCciwci Y N N N N Y

mbCci Y N N N Y N

mbCcl Y Y Y N N N

mbCcici Y N N Y Y Y

mbCcicl Y Y Y N Y N

mbCcici cl Y Y Y Y Y Y
3.4 Some Conceptual Differences Between mbC and mbC 91

3.4 Some Conceptual Differences Between mbC and mbC

Despite having defined a consistency operator in terms of the other connectives,

neither mbC nor even mbCciwci are dC-systems, in virtue of Item (iii) of Defini-
tion 3.3.5. Indeed, as observed after Remark 3.3.6, the construction of the operator
, involved in the definition of , uses the consistency operator itself. However,
the situation changes if we consider the signature  of Sect. 2.5 of Chap. 2 and the
presentation mbC of mbC in that signature: it can be proved that this system is a
dC-system in a weaker sense. This and other conceptual differences between mbC
and mbC constitute the subject of this section.
Recall from Definition 2.5.3 that the calculus mbC is defined over the signature
 = {, , , } as an alternative (and equivalent) presentation of mbC. The
Hilbert calculus mbC is an axiomatic extension of CPL+ , while mbC extends
CPLW , the Hilbert calculus for CPL over the signature W = {, } introduced
by Wajsberg. Thus, it is clear that both logics, mbC and mbC , departs from different
logical basis, the latter strictly stronger than the former. This difference can be made
more explicit when considering the notion of C-systems (recall Definition 3.3.4),
which takes into account more aspects and relationships between the involved logics.
In Sect. 3.3, it was proved that, while mbC is a C-system based on CPL+ (see
Theorem 3.3.10), mbC is a C-system based on CPLW (see Theorem 3.3.13).
Let us return to the definition of an alternative consistency operator in mbC. In the
signature  of mbC , the consistency operator introduced in Definition 3.2.1 is
no longer indexed by , as there is a constant included in the signature  . Thus,
it is possible to define the following alternative consistency operator in mbC :

Definition 3.4.1 The strong consistency operator, denoted by , is defined in

mbC by means of the following formula:

p = p ( p ) .
def

Thus, = ( ), for every L .

Remark 3.4.2 Arguably, the operator is the version (over the signature  ) of
the operator ( p) , up to logical equivalence. Indeed, if the formula is rewritten
by using some abbreviations and logical equivalences from mbC , the following
is obtained: is ( ) (by the definition of strong negation in mbC ),
which is logically equivalent in mbC to ( ), which, by the definition of
conjunction in mbC , is the formula ( ). By identifying with () ,
the latter corresponds to the formula () defined over .

The following result is similar to Theorem 3.2.2, but behind it lies an important
conceptual difference which sets it apart from the former (see Remark 3.4.9 below):
92 3 Some Extensions of mbC

Theorem 3.4.3
(i) Let  ( p) = { p} for a propositional variable p. The logic mbC is a (strong)
LFI (w.r.t. and  ( p)) according to Definition 2.1.9.
(ii) With the LFI structure of Item (i), the logic mbC satisfies axiom (ciw ), namely:
for every formula in L , mbC ( ). That being so, the truth-value
of is determined by the truth-values of and as follows: for every mbC -
valuation v, v( ) = 1 iff v() = v() (iff either v() = 0 or v() = 0).
(iii) mbC , but in general mbC . Thus, is a consistency
operator different from the original .
(iv) mbC ( ) but in general mbC ( ) .
(v) mbC ( ) but in general mbC ( ).

Proof Straightforward. 

At this point, it will be convenient to introduce an alternative notion of dC-system:

Definition 3.4.4 Let Li = i , i  (i = 1, 2) be two Tarskian and structural logics.

We say that L2 is a weak dC-system based on L1 w.r.t. (or simply a weak dC-
system) if the following conditions hold:
(i) L2 is a conservative extension of L1 ,
(ii) cannot be defined in L1 , that is: there is no formula ( p) L1 such that
p 2 ( p) and ( p) 2 p,
(iii) let 3 = 1 { }; then, there exists a formula ( p) L3 such that L2 is
an LFI w.r.t. and ( p),
(iv) L1 is non-trivial.


From the previous definition, the following result can be easily proved. Details
are left to the reader.

Proposition 3.4.5 Let L2 be a weak dC-system based on L1 w.r.t. , and let ( p)

be a consistency operator definable in L2 according to Item (iii) of Definition 3.4.4.
Then L2 is a C-system based on L1 w.r.t. and ( p).

A relevant distinction between mbC and mbC can now be established: while
the former is not a weak dC-system based on CPL+ , the latter is a weak dC-system
based on CPLW .

Proposition 3.4.6 The logic mbC is not a weak dC-system based on

CPL+ w.r.t. .

Proof By adapting the proof of Proposition 3.3.19, it is easy to prove that Item (iii)
of Definition 3.4.4 cannot be satisfied by mbC. 

Proposition 3.4.7 The logic mbC is a weak dC-system based on CPLW w.r.t.
, where the consistency operator is defined as p = p ( p ) (recall
def

Definition 3.4.1).
3.4 Some Conceptual Differences Between mbC and mbC 93

Proof It is an immediate consequence of Propositions 3.3.11 and 3.3.12, and Theo-

rem 3.4.3(i). 

Moreover, the following result can be obtained (see Theorem 3.4.10 for a more
general result):

Proposition 3.4.8 The logic mbC is a dC-system based on CPLW w.r.t. , where
the consistency operator is given by p = p ( p ).
def

Proof The proof is left as an exercise to the reader. 

Remark 3.4.9 (An important difference between mbC and mbC ) The last results
prove that, as in mbC, it is possible to define in mbC an alternative consistency oper-
ator , which is not equivalent to the primitive : the former satisfies axiom (ciw ),
while the latter do not. However, there is an important conceptual difference between
Theorem 3.2.2 for mbC and Theorem 3.4.3 concerning mbC : in the case of mbC,
the strong consistency operator ( p) is defined in terms of some primitive connec-
tives of the signature , including the original consistency operator (recall that the
consistency operator occurs in ( p) ). But, in the case of mbC , the situation is
completely different: the consistency operator is defined in terms of those con-
nectives of  other than . This diference can be formalized by means of the notion
of weak dC-system, and it was stated in Propositions 3.4.6 and 3.4.7.
This phenomenon can be explained by the fact that, when considering as a
primitive operator in mbC , part of the meaning of is already included: it is not
possible (in the original formulation of mbC) to define a bottom formula without
using the consistency operator (recall Proposition 3.3.18).
From the last observations, it is clear that mbC and mbC have a different nature
in a certain sense: the signature  allows us to consider the consistency-less fragment
mbC of mbC which is unable to define a bottom formula, and so it is not an LFI
w.r.t. . On the other hand, by considering the signature  , the consistency-less
fragment mbC
of mbC extends CPLW and, moreover, is capable of defining an
alternative consistency operator for mbC . Moreover, mbC is, itself, a dC-
system, as stated in Proposition 3.4.8.

From the last remark, an interesting and very general result concerning LFIs based
on classical logic can be obtained:

Theorem 3.4.10 Let LCPL be a Hilbert calculus for the classical propositional logic

CPL defined over any suitable signature, and let LCPL be the system obtained from
LCPL by adding a new unary connective to the signature of LCPL together with
the axiom schema
.

Then LCPL is a dC-system based on LCPL w.r.t. and the consistency operator p
defined by
p = ( p p),
def
94 3 Some Extensions of mbC

where denotes the classical negation (defined or primitive) in LCPL . Moreover,

 

LCPL is semantically characterized by (classical) LCPL -valuations over 0, 1
which additionally satisfy the following clause:
(vNeg) v() = 0 = v() = 1.

Additionally, LCPL derives axiom schema (ciw) and then the truth-value of under
a valuation v of such semantics is determined by the truth-values of and under
v as follows:
v() = 1 iff v() = 0 or v() = 0 .

Proof This is an easy consequence of the previous results and definitions. We leave
the details to the reader. 

Of course, the logic mbC belongs to the family of systems of the form LCPL
considered in the last theorem. Indeed, Proposition 3.4.8 is a particular instance of
Theorem 3.4.10.
As occurs with mbC, it is possible to extend mbC in a minimal way to obtain
equivalence between and :

Definition 3.4.11 Let mbC ciw be the logic obtained from mbC by adding the
axiom schema
( ) (ciw )

and by considering a semantics of mbC -valuations (recall Definition 2.5.8) which

additionally satisfy the clause
(vCiw) v() = 0 or v() = 0 = v() = 1. 

Theorem 3.4.12 (Soundness and Completeness of mbC ciw w.r.t. valuations) For
every  {} L :

 mbC ciw  mbC ciw .

Proof In order to prove soundness, it is enough to observe that, for every mbC ciw-
valuation v, v( ) = 1 iff v() = 0 or v() = 0. Completeness is proved
as usual. 

Corollary 3.4.13 The logic mbC ciw is a (strong) LFI.

Theorem 3.4.14 The logic mbC ciw is the least extension of mbC which is a
dC-system such that can be expressed by the formula = ( ).
def

Proof It is enough to observe that (ciw ) is the converse of the axiom schema (bc1 )
of mbC . Details are left to the reader. 

The last result shows that mbC ciw is the least extension of mbC such that
and where the mapping t which replaces by t () is a translation from
mbC ciw to mbC ciw. This property highlights the intrinsic difference between
3.4 Some Conceptual Differences Between mbC and mbC 95

mbC and mbC observed in Remark 3.4.9: it is easier to obtain a (genuine) dC-
system from mbC than from mbC. Indeed, adding axiom schema (ciw) to mbC
is not enough to produce a dC-system: it is necessary to add the stronger axiom
schema (cl).
Finally, in order to also define the inconsistency operator in terms of the
others, the following system should be taken into consideration:

Definition 3.4.15 Let mbC ci be the logic obtained from mbC by adding the
axiom schema (ci) below.
( ) (ci)

The semantics of mbC ci is given by all the valuations for mbC which additionally
satisfy the clause (vCi) (recall Definition 3.1.7). 

The next theorem is proved as usual:

Theorem 3.4.16 (Soundness and Completeness of mbC ci w.r.t. valuations) For

every  {} L :

 mbC ci  mbC ci .

Corollary 3.4.17 The logic mbC ci is a (strong) LFI.

Clearly, system mbC ci is the least extension of mbC such that both and
can be defined in terms of the other connectives: ( ) and
( ). Again, this shows a difference between mbC and mbC: in the latter,
and are definable in terms of the other connectives only in the stronger system
mbCcicl, where ( ) and ( ).

3.5 Inconsistency Operators and Double-Negations

Thus far, we have just considered the (formal) notion of consistency, represented by
the operator . With regard to inconsistency, in the previous sections we spoke loosely
about the possibility of considering the derived operator as defining the (formal)
inconsistency of . The absence of an inconsistency operator in the LFIs studied up
to now is somewhat surprising, since these systems are called Logics of Formal
Inconsistency, which suggests that the inconsistency instead of the consistency
operator should be stressed. This is why Logics of Formal Consistency might
possibly be a better name for this kind of logic.
Returning to the aforementioned inconsistency operator = , its truth-
def

table in the basic system mbC is as follows:

96 3 Some Extensions of mbC

1 1 0 1 1
0 1 0 1
0
0 0 1
0 1 1 0 1
0
0 0 1

or, more succintly:

1 1 1 1
0 1 0
0 0
0 1 1 0
0 0

If we consider instead of as the primitive connective, the basic system of incon-

sistency characterized by the table above can be axiomatized as follows: recall from
Definition 2.1.3 the signature  obtained from  by replacing by , and let L be
the calculus over L obtained from mbC by replacing axiom schema (bc1) by

( ) (Inc )
 
The logic L is clearly characterized by valuations v : L 0, 1 , satisfying
all the clauses of Definition 2.2.1 with exception of (vCon), which is accordingly
replaced by the following clause:
(vInc ) v() = 1 and v() = 1 = v() = 1.
However, this logic is too weak to be viewed as an LFI representing mbC in the
signature  , since it is not possible to define a bottom formula
 within L . This is
a consequence of the fact that the mapping v : L 0, 1 such that v() = 1 for
every , is a valuation for this logic (we invite the reader to check the details).
Observe that should have the intended meaning of a (derived) consistency
operator. Accordingly, axiom (Inc ) and the corresponding clause (vInc ) must be
strengthened in order to obtain an alternative presentation of mbC being defined
in terms of the inconsistency operator . We accordingly arrive at the following
definition:

Definition 3.5.1 (mbC ) The calculus mbC is defined over the language L by
adding the following axiom schemas to CPL+ (that is, to the system formed by
Ax1-Ax9 plus MP, recall Definition 2.1.12):
3.5 Inconsistency Operators and Double-Negations 97

(Ax10)
 
(Inc)
 
The semantics of mbC -valuations is formed by all the mappings v : L 0, 1
which satisfy all the clauses of Definition 2.2.1 with the exception of (vCon), which
is replaced by the following clause:
(vInc) v() = 1 and v() = 1 = v() = 0. 

  maximal non-trivial with respect to in

Theorem 3.5.2 Let  {} L , with
mbC . The mapping v : L 0, 1 defined by:

v() = 1 

for all L is a valuation for mbC .

Proof The definitions of mbC , mbC -valuations and Theorem 2.2.7 suffice to prove
that v satisfies clause (vInc). Thus, assume that is a formula such that v() =
v() = 1. Then  and . Suppose by contradiction that v() = 1,
and then . However, ( ( )) is a theorem of mbC given
axiom (Inc). From this, and using MP, it follows that  mbC , a contradiction.
Thus, v() = 0 and v accordingly satisfies (vInc). This shows that v is an mbC -
valuation, as required. 
Theorem 3.5.3 (Soundness and Completeness of mbC w.r.t. valuations) For every
 {} L :
 mbC  mbC .

Proof Soundness is immediate. Concerning Completeness, the proof is analogous

to that of mbC, but now employing Theorem 3.5.2 instead of Theorem 2.2.7. The
details are left as an exercise to the reader. 
If a consistency operator is defined in mbC as = , then the following
def

table is obtained:

( )
1 1 1 1 1 0 v1
0 0 v2
0 1 0 1 1 v3
0 v4
0 0 1 1 v5
0 1 1 0 1 1 v6
0 v7
0 0 1 1 v8

Observe that the derived consistency operator of mbC has the same properties
as the primitive one of mbC.
98 3 Some Extensions of mbC

However, mbC and mbC are not inter-translatable, as one might expect (see
Proposition 3.5.5 below):

Definition 3.5.4
(1) The mapping + : L L is recursively defined as follows:
p + = p if p V ar ;
(#)+ = (( + )#( + )) for # {, , };
()+ = ( + );
()+ = ( + ) if = for every ;
()+ = ( + ).
(2) The mapping : L L is recursively defined as follows:
p = p if p V ar ;
(#) = (( )#( )) for # {, , };
() = ( );
() = ( ) if = for every ;
() = ( ).


Proposition 3.5.5
(1) The function + is not a translation from mbC to mbC .
(2) The function is not a translation from mbC to mbC.

Proof (1) Let be the formula ( ( )) in L . Then + is

the formula ( + ) (( + ) (( + ) + )) in L . On the one hand,
mbC , since it is an instance of axiom (bc1). But, on the other hand, the formula
+ is not valid in mbC . To see this, it is enough to consider and as two different
propositional variables and an mbC -valuation v such that v( ) = v( ) =
v( ) = 1 but v() = 0. (Observe that this is possible, as is a propositional vari-
able: from v( ) = v( ) = 1, it follows that v( ) = 1, and nothing forbids
having v( )) = 1 as well). This shows that the function + fails to be a trans-
lation from mbC to mbC .
(2) Let be the formula ( ( )) in L . Then is the
formula ( ) (( ) (( ) )) in L . Observe that mbC ,
as it is an instance of axiom (Inc). However, the formula is not a valid for-
mula in mbC (in general). Indeed, suppose that and are two different proposi-
tional variables and consider an mbC-valuation v such that v( ) = 1, v( ) = 0,
v( ) = v( ) = 1, v( ) = 0 and v( ) = 1 (which is always possible).
Finally, as is a propositional variable different from , it is possible to take
v() = 0. This valuation is such that v( ) = 0, and hence is not a theorem of
mbC, by completeness. Therefore the mapping is not a translation from mbC to
mbC. 

From the last result, it is clear that in order to obtain the equivalence between
mbC and mbC up to language, both systems must be improved.
3.5 Inconsistency Operators and Double-Negations 99

Definition 3.5.6 (mbCcew) The calculus mbCcew is defined over the language
L by adding the following axiom schema to mbC:

(cew)

The semantics of mbCcew-valuations is formed by mbC-valuations which addition-

ally satisfy the following clause:
v() = 1
(vCew) = v() = 1. 

Definition 3.5.7 (mbC cew) The calculus mbC cew is defined over the language
L by adding the following axiom schema to mbC :

(cew )

The semantics of mbC cew-valuations is formed by mbC -valuations which addi-

tionally satisfy the following clause:
(vCew ) v() = 1 = v() = 1. 

The proof of the following results is now routine:

Theorem 3.5.8 (Soundness and Completeness of mbCcew w.r.t. valuations) For

every  {} L :

 mbCcew  mbCcew .

Corollary 3.5.9 The logic mbCcew is a (strong) LFI.

Theorem 3.5.10 (Soundness and Completeness of mbC cew w.r.t. valuations) For
every  {} L :

 mbC cew  mbC cew .

The envisaged equivalence (up to language) between a suitable extension of mbC

(defined in terms of consistency) and mbC (defined in terms of inconsistency) can
now be obtained:

Theorem 3.5.11
(1) The function + is a translation from mbCcew to mbC cew. That is: if
 mbCcew then  + mbC cew + .
(2) The function is a translation from mbC cew to mbCcew. That is: if
 mbC cew then  mbCcew .

Proof (1) The strategy of the proof is as follows: assuming that  mbCcew , given
a derivation of from  in mbCcew, a derivation of + from  + = { + : } is
constructed in mbC cew, showing that  + mbC cew + . Moreover, this derivation is
obtained just by checking that, for any instance of any axiom schema of mbCcew,
100 3 Some Extensions of mbC

the formula + is derivable in mbC cew (note that the translation of any application
of MP in mbCcew is an application of MP in mbC cew, because + preserves the
implication ).
It should be firstly observed that, by analyzing all the axiom schemas of mbCcew
along with the definition of the function +, there are exactly three cases in which an
instance of an axiom schema of mbCcew is such that + is not an instance of an
axiom schema of mbC cew. The three cases are as follows:
Case1: is an instance of axiom schema (bc1) of the specific form

= ( ( )) .

Then
+ = ( + ) (( + ) (( + ) ( + ))),

which is not an instance of an axiom in mbC cew. However, the following can be
derived in mbC cew: the sequence

1. ( + ) (hyp)
2. ( + ) (hyp)
3. ( + ) (hyp)
4. ( + ) (( + ) (( + ) ( + ))) (Inc)
5. ( + ) (( + ) ( + )) (MP 1, 4)
6. ( + ) ( + ) (MP 2, 5)
7. ( + ) ( + ) (cew)
8. ( + ) (MP 3, 7)
9. + (MP 8, 6)

shows that {( + ), ( + ), ( + )} mbC cew + . Using the deduction meta-

theorem DMT (used here, three times), it follows that mbC cew + .
Case2: is an instance of axiom schema (Ax10) of the specific form

= .

Then
+ = ( + ) ( + ),

which is not an instance of an axiom in mbC cew. Since mbC cew

(which can be easily checked by the reader), it follows by the axiom (Ax10) of
mbC cew and the property above that mbC cew + .
Case3: is any instance of axiom schema (cew):

= .

Then
+ = ( + ) ( + ),
3.5 Inconsistency Operators and Double-Negations 101

which is not an instance of an axiom in mbC cew. However, it can derived in

mbC cew, as mbC cew for every .
(2) The proof is entirely analogous to that of item (1). There are also only three cases
to be analized:
Case1: is an instance of axiom schema (Inc) of the specific form

= ( ( ))

and then,

= ( ) (( ) (( ) ( ))),

which is not an instance of an axiom in mbCcew.

Case2: is an instance of axiom schema (Ax10) of the specific form

= ,

and so

= ( ) ( ),

which is not an instance of an axiom in mbCcew.

Case3: is any instance of axiom schema (cew ):

= ,

and therefore

= ( ) ( ),

which is not an instance of an axiom in mbCcew.

In the three cases, the proof is analogous to that of item (i). Details are left to the
reader. 

The last theorem means that mbCcew and mbC cew are equivalent up to lan-
guage, and it is now possible to present the extension mbCcew of mbC in terms
of an inconsistency operator. It should be clear that mbCcew and mbC cew are the
least extensions of mbC and mbC satisfying this property. Observe that, in order
to obtain the equivalence between the presentations of minimal LFIs in terms of
consistency and inconsistency, respectively, some (weak) form of double-negation
law is required, namely axioms (cew) and (cew ).
Also with the aim of obtaining equivalent presentations of LFIs in terms of con-
sistency and inconsistency operators, a much stronger system called mCi was studied
in [1]. This system can be recast as follows.
102 3 Some Extensions of mbC

For every n 0 and any formula , let the formula n be defined recursively in
the following way: 0 = , and n+1 = n .
def def

Definition 3.5.12 Let mCi be the calculus defined over the language L by adding
the following axiom schema to mbCci, for each n 0:

n+2 n (n )

The semantics of mCi is given by mbC-valuations v which additionally satisfy the

clauses (vCi) (recall Definition 3.1.7) and
(vNegn ) v(n+2 ) = 1 = v(n ) = 1,
for every n 0. 

As usual, we have the following:

Theorem 3.5.13 (Soundness and Completeness of mCi w.r.t. valuations) For every
 {} L :
 mCi  mCi .

Corollary 3.5.14 The logic mCi is a (strong) LFI.

A important result concerning mCi can now be proved:

Theorem 3.5.15 In mCi, it holds that:

(i) mCi n for every n 0 and every .
(ii) mCi n n+2 for every n 0 and every . As a consequence,
n+2 n in mCi for every n 0 and every .
(iii) The logic mCi can be characterized as the extension of mbCci obtained by
adding the axiom schema

n (ccn )

for each n 0.

Proof (i) By induction on n. The case n = 0 holds given Proposition 3.1.10 item 2,
since mbCci and mCi extends mbCci. Now, suppose that mCi k for
every 0 k n and every (induction hypothesis); let be a formula. Since mCi
extends mbCciw, it follows that, for every , v() = 1 iff v() = v(). Thus, if
v(n+2 ) = 0, then v(n+1 ) = 1, given (vNeg), and then v(n+1 ) = 1 given
the observation above. On the other hand, if v(n+2 ) = 1, then v(n ) = 1 given
clause (vNegn ). However, mCi n by induction hypothesis, so v(n ) = 1, by
Theorem 3.5.13. In light of the observation above, it follows that v(n+1 ) = 0. But
then, by the same observation, it follows that v(n+1 ) = 1. Using Theorem 3.5.13
again, we infer that mCi n+1 , as required.
(ii) According to item (i) and axiom (bc1) it follows that n , n+1 mCi for
every , as mCi n . In particular, n , n+1 mCi n+2 . But always
3.5 Inconsistency Operators and Double-Negations 103

n , n+2 mCi n+2 . Using proof-by-cases (Proposition 2.1.14 item (iii)),

n mCi n+2 and so, by DMT (Proposition 2.1.14 item (i)), the result follows.
(iii) According to item (i), each instance of axiom schema (ccn ) is derivable in mCi
for every n 0. Conversely, let mCi be the system obtained from mbCci by adding
axiom schema (ccn ) for every n 0. As is the case in the proof of item (ii), it is easy to
prove that n+2 mCi n using proof-by-cases and mCi n+1 . Employing
DMT, the result follows. 

Remark 3.5.16 The logic mCi was originally axiomatized as the system described
in item (iii) of Theorem 3.5.15 (see [1, 9]). Observe, again, that some form of double-
negation law is present in this system.

Consider, finally, the two usual laws of double-negation:

(cf)
(ce)

The respective clauses for valuations validating these axioms, respectively, are as
follows:
(vCf ) v() = 1 = v() = 1.
(vCe) v() = 1 = v() = 1.
Several LFIs satisfying one (or both) of the laws above were considered in the litera-
ture. In [1, 3], a wide variety of mbC extensions were studied, accepting axiom (cf)
and/or axiom (ce) in combination with other additional axioms:

Definition 3.5.17 Consider the following extensions of mbC (see [1, 3]):
1. The system bC is obtained from mbC by adding axiom (cf).
2. The system Ci is obtained from mCi by adding axiom (cf).
3. The system Cil is obtained from Ci by adding axiom (cl).
4. The system mbcCe is obtained from mbC by adding axiom (ce).


Remark 3.5.18 The system Ci can be alternatively presented as the extension

obtained from mbC by adding axioms (ci) and (cf). That being the case, the family
of axioms (ccn ) is derived in Ci (see Theorem 102 in [1]).

It is easy to prove that all the systems of the last definition are strong LFIs. The
system bC was the basis for the taxonomy of systems introduced in the pioneering
study of LFIs presented in [3]. The system Cil was studied in [1] as the basic dC-
system in which ( ) (see Theorem 106, p. 65 in [1]). However, in
Theorem 3.3.44 it was proved that this feature can also be obtained in the weaker
system mbCcl (recall Definition 3.3.26). A similar situation happend with mCi, an
extension of mbC introduced in [1] with the aim of obtaining a duality between the
consistency and the inconsistency operators by means of the paraconsistent negation
. From this, mCi is an LFI which can be presented either in terms of consistency or
104 3 Some Extensions of mbC

inconsistency operators up to translations (see Theorem 98, p. 60 in [1]). As proved

in Theorem 3.5.11, the weaker system mbCcew (recall Definition 3.5.6) enjoys the
same feature.

3.6 Propagating Consistency

One of the features that da Costa required for his Inconsistent Formal Systems
(see [10]), was the propagation of well-behaviorness (or, in the terminology of LFIs,
the propagation of consistency). Basically, this means that a complex formula whose
immediate subformulas are assumed to be consistent (or well-behaved, according to
da Costas terminology), is itself consistent. This is equivalent to requiring the propa-
gation of consistency for each connective. In this section, we will analyze extensions
of mbC which enjoy the propagation of consistency, characterizing this property
in terms of other connectives instead of the consistency operator. That is, we will
show that it is possible, starting from a suitable extension of mbC, namely mbCciw,
to express the propagation of consistency of each connective (with exception of )
exclusively in terms of the other connectives.

Definition 3.6.1 (Consistency propagation) Let L be an extension of mbC defined

over the signature . The propagation of consistency of each connective of  in L
is given by the following axiom schemas:

(ca )
(ca )
( ) ( ) (ca )
( ) ( ) (ca )
( ) ( ) (ca )

An immediate consequence of the property of propagation of consistency is the

following:

Proposition 3.6.2 Let L be an extension of mbC defined over  which satisfies

the propagation of consistency for every connective of , in accordance with Defi-
nition 3.6.1. Then:
p1 , . . . , pn L

for every formula L whose propositional variables belong to { p1 , . . . , pn }.

Proof By induction on the complexity of (see Definition 2.1.11). We leave the

details to the reader. 
3.6 Propagating Consistency 105

Remark 3.6.3 Axioms (ca ), (ca ) and (ca ) were included in the original formu-
lation of the systems Cn for n 1 (see Definition 3.7.4), with being a defined,
not primitive connective. The other propagation axioms (ca ) and (ca ) were not
required since they are derivable (actually, is a theorem for each Cn ). The ver-
sion of C1 in the signature  is Cila, which is obtained from mbC by adding axiom
schemas (ci), (cl), (cf), (ca ), (ca ) and (ca ) (see [1, 3]). As in C1 , axioms (ca )
and (ca ) are derived. It is easy to prove that Cila is a strong LFI.

From the semantical point of view, a valuation semantics for a logic L , as in

Definition 3.6.1, must satisfy the following clauses:
(vCa ) v() = 1 = v() = 1.
(vCa ) v() = 1 = v() = 1.
(vCa ) v() = 1 and v() = 1 = v(( )) = 1.
(vCa ) v() = 1 and v() = 1 = v(( )) = 1.
(vCa ) v() = 1 and v() = 1 = v(( )) = 1.

In order to analyze the minimum requirements for consistency propagation, we

start not from mbC, since it is too weak, but from its first extension considered here,
system mbCciw (recall Definition 3.1.1). This is the least extension of mbC such that
the truth-values of and completely determine the truth-value of : for every
mbCciw-valuation v, v() = 1 iff v() = v(). Compared to the basic system
mbC, this feature drastically simplifies an analysis of the propagation of consistency
in mbCciw.
We begin by analyzing the case of axiom (ca ), which states the propagation
of consistency for conjunction. Thus, given arbitrary formulas and , there are
exactly two scenarios in mbCciw, namely v1 and v2 , in which both and are
simultaneously true and, at the same time, is also true:

( ) ( )
1 1 0 0 1 1 1 1 0 v1
0 1 v2

Observe that these are the only relevant cases for analyzing (ca ): in any other case,
either v() = 0 or v() = 0 (and then (ca ) is vacuously true) or v( ) =
0 (and so v(( )) = 1, which implies that v(( )) = 1, again satisfying
axiom (ca )). It is clear that valuation v1 should be blocked, as it violates (ca ).
Thus, the following clause should be required for the valuations:

v() = 0 and v() = 0 implies v(( )) = 0 .

(Note that, following clause (vNeg), v() = 0 is equivalent to v() = 1 and

v() = 0, with the same applying to .) However, this is equivalent, by contrapo-
sition, to the following:
106 3 Some Extensions of mbC

(vDM 1 ) v(( )) = 1 = v() = 1 or v() = 1.

The axiom schema corresponding to this clause is clearly the following:

( ) ( )

This axiom, that we will call (neg1 ), is included in the system LFI1 (to be
studied in Sect. 4.4.7, see Definition 4.4.41) as well as in the Hilbert calculus for
Nelsons paraconsistent logic N4, which will be briefly described in Chap. 5 (see
Definition 5.1.24).
This lead us to the following characterization of the propagation of consistency
through conjunctions, without using the consistency operator:
Theorem 3.6.4 The extension of mbCciw obtained by adding axiom schema (ca )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schema:
( ) ( ) (neg1 )

Proof We leave it to the reader as an exercise. 

Now, let us analyze the case of axiom (ca ), which states the propagation of
consistency for disjunction. Given arbitrary formulas and , there are now six
relevant scenarios in mbCciw in which both and are simultaneously true and,
at the same time, is also true:

( ) ( )
1 1 0 0 1 1 1 1 0 v1
0 1 v2
1 0 0 1 1 1 1 1 0 v3
0 1 v4
0 1 1 0 1 1 1 1 0 v5
0 1 v6

In this case, the valuations v1 , v3 and v5 must be eliminated. Following the same
reasoning as above, the following clause for the valuations can be obtained:

(vDM 1 ) v(( )) = 1 = v() = v() = 1.

The axiom schema associated to this clause is the following:

( ) ( ) (neg1 )

As in the case of (neg1 ), axiom (neg1 ) is included in the system LFI1 (see Def-
inition 4.4.41) as well as in Nelsons paraconsistent logic N4 (see Definition 5.1.24).
However, this condition is sufficient but not necessary in order to characterize con-
sistency propagation for disjunctions:
3.6 Propagating Consistency 107

Theorem 3.6.5 The extension of mbCciw obtained by adding axiom schema (neg1 )
is strictly stronger than the extension of mbCciw obtained by adding axiom
schema (ca ).

Proof It is easy to see that any mbCciw-valuation satisfying (vDM 1 ) also satisfies
clause (vCa ). Thus, assuming soundness and completeness, (ca ) can be derived
from (neg1 ) in mbCciw. On the other hand, given two different propositional vari-
ables p and q, it is possible to define a valuation v for mbCciw satisfying (vCa ) such
that v( p) = 1 and v( p) = 0 (so v( p) = 1); v(q) = v(q) = 1 (so v(q) = 0);
and v(( p q)) = 1 (so v(( p q)) = 0, since v( p q) = 1). Observe that, under
these additional assumptions, v does not violate (vCa ). However, v does not satisfy
(vDM 1 ) since v( p q) = 0. Therefore, assuming soundness and completeness,
(neg1 ) cannot be derived from (ca ) in mbCciw. 

A characterization of consistency propagation for disjunctions can be obtained

by considering instead the following clauses for valuations, which also eliminate the
valuations v1 , v3 and v5 mentioned above:
(vCa1 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCa2 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCa3 ) v(( )) = 1 = v() = 1 or v() = 1.
We thus obtain a characterization of the propagation of consistency through dis-
junctions without using the consistency operator:

Theorem 3.6.6 The extension of mbCciw obtained by adding axiom schema (ca )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schemas:

( ) ( ) (ca1 )
( ) ( ) (ca2 )
( ) ( ) (ca3 )

Proof Again, we leave it to the reader as an exercise. 

For axiom (ca ), which defines the propagation of consistency for implication,
there are also six relevant scenarios in mbCciw in which both , and
are simultaneously true:

( ) ( )
1 1 0 0 1 1 1 1 0 v1
0 1 v2
0 1 1 0 1 1 1 1 0 v3
0 1 v4
0 0 1 1 1 1 1 1 0 v5
0 1 v6
108 3 Some Extensions of mbC

The valuations to be avoided are v1 , v3 and v5 . Reasoning as above, one is tempted

to consider the following clause for valuations:
(vCIp1 ) v(( )) = 1 = v() = v() = 1.
This condition can be associated to the following axiom schema:

( ) ( ) (neg 1 )

Once again, axiom (neg 1 ) is included in the system LFI1 (see Defini-
tion 4.4.41) as well as in Nelsons paraconsistent logic N4 (see Definition 5.1.24).
But, as in the case of axiom (neg1 ) proposed above for disjunctions, this condition
is sufficient but not necessary in order to characterize consistency propagation for
implications:

Theorem 3.6.7 The extension of mbCciw obtained by adding axiom schema

(neg1 ) is strictly stronger than the extension of mbCciw obtained by adding axiom
schema (ca ).

Proof The proof is similar to that for Theorem 3.6.5 presented above. In order to
prove that (neg 1 ) cannot be derived from (ca ) in mbCciw, let p and q be two dif-
ferent propositional variables. It is possible to define a valuation v for mbCciw satis-
fying (vCa ) such that v( p) = 0 and v( p) = 1 (so v( p) = 1); v(q) = v(q) = 1
(so v(q) = 0); and v(( p q)) = 1 (so v(( p q)) = 0, since v( p q) = 1).
It is worth noting that, under these additional assumptions, v does not violate (vCa ).
However, v does not satisfy (vCIp1 ) since v( p q) = 0. 

In order to avoid the valuations v1 , v3 and v5 mentioned above, the following

clauses for valuations can be taken into account instead of (vCIp1 ):
(vCa1 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCa2 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCa3 ) v(( )) = 1 = v() = 1 or v() = 1.
In this way, the propagation of consistency through implications can also be
characterized in terms of an axiom without using the consistency operator:

Theorem 3.6.8 The extension of mbCciw obtained by adding axiom schema (ca )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schemas:

( ) ( ) (ca1 )
( ) ( ) (ca2 )
( ) ( ) (ca3 )

Proof Left to the reader once more. 

3.6 Propagating Consistency 109

Finally, concerning the propagation of consistency w.r.t. the unary connectives

and , the following holds:

Theorem 3.6.9
(i) The extension of mbCciw obtained by adding axiom schema (ca ) is equivalent
to the extension of mbCciw obtained by adding axiom schema (cc), that is, mbCci
(by Proposition 3.1.10(3)).
(ii) The extension of mbCciw obtained by adding axiom schema (ca ) is equivalent
to the extension of mbCciw obtained by adding axiom schema (cf).

Proof The proof will be done using valuations.

(i) Suppose that (ciw) and (ca ) were added to mbC, and let v be a valuation for
the resulting system. Given a formula and assuming that v() = 1, it follows
that v() = 1 by (ca ). By (ciw), the latter implies that v() = v() and then
v() = 0. Thus: v() = 1 implies v() = 0 or, equivalently, either v() = 0
or v() = 0 for every . This means that, by (ciw), v() = 1 for every , and
hence the logic also satisfies axiom schema (cc). The converse is obvious.
(ii) Let v be a valuation for the system obtained from mbC by adding (ciw)
and (ca ), such that v() = 0 for a given formula . Then v() = 1 and v() = 1
by (ciw). Thus, v() = 1 by (ca ). This implies that v() = v(), whence
v() = 0. Thus: v() = 0 implies that v() = 0 for every . By contraposi-
tion, v() = 1 implies that v() = 1, showing that axiom (cf) is valid in this logic.
Conversely, consider the extension of mbC obtained by adding axiom schemas (ciw)
and (cf), and let v be a valuation for this logic such that v() = 1, for a given
formula . Then, by (ciw), v() = v(). Thus, if v() = 1, then v() = 0 and
v() = 1 accordingly. On the other hand, if v() = 0, then v() = 0 by (cf)
and v() = 1 accordingly. In both cases, v() = v(), so v() = 1. Thus:
v() = 1 implies that v() = 1 for every , showing that axiom (ca ) is valid
in this logic. 

In virtue of Proposition 3.1.10(3) and Theorems 3.6.4, 3.6.6, 3.6.8 and 3.6.9(i),
the following consequence is easily obtained:

Corollary 3.6.10 The least extension of mbC which satisfies the propagation of
consistency (see Definition 3.6.1) is the system mbC1 over  obtained from Ci (see
Remark 3.5.18) by adding axioms (neg1 ), (ca1 ), (ca2 ), (ca3 ), (ca1 ), (ca2 ) and
(ca3 ). The system mbC1 is a strong LFI.

The name mbC1 suggests an analogy to the system C1 which, besides satisfy-
ing (cf), enjoys consistency propagation for all the connectives of its signature.
As was observed above, in conjunction with Proposition 3.6.9(ii), it is worth noting
that the axioms added to mbCciw which characterize the propagation of consistency
for the binary connectives and the negation do not use the consistency operator to
express this property. Then:
110 3 Some Extensions of mbC

Corollary 3.6.11 The least extension of mbC which satisfies the propagation of
consistency for all the connectives of 0 = {, , , } is the system mbCw1 which
is obtained from mbCciw by adding axioms (neg1 ), (ca1 ), (ca2 ), (ca3 ), (ca1 ),
(ca2 ), (ca3 ) and (cf).

This means that the least extension of mbC satisfying propagation of consistency
for all the connectives of 0 can be characterized from mbCciw by means of axioms
over the signature 0 . Again, the name given to the system mbCw1 suggests an analogy
to the system C1 . The letter w stands for weak since, in fact, this system is obtained
from mbC1 by replacing (ci) by (ciw). Because of the propagation of the consistency
operator for the connectives in 0 , the DAT for mbC presented in Theorem 2.4.7 can
be improved for mbCw1 : in order to reproduce the classical reasoning inside mbCw1 ,
it is sufficient to add the the additional hypothesis for consistency of the relevant
propositional variables.

Theorem 3.6.12 Let t  : Lc L0 be the mapping which replaces with . Then
the following holds: for every  {} Lc ,

 CPL iff { p : p V ar  } t  [] mbCw1 t  (),

where V ar  is the set of propositional variables occurring in  {}.

Finally, recall from Theorem 3.4.10 that LCPL denotes the system obtained from
LCPL (classical propositional logic CPL defined over any suitable signature) by
adding a negation which only satisfies the third-excluded law . This is an
LFI where p = ( p p) (here, denotes the primitive or definable negation
def

of CPL). Moreover, it is a dC-system based on LCPL which additionally satisfies

axiom schema (ciw).

Definition 3.6.13 Let CPL be the Hilbert calculus for classical propositional logic
defined over the signature c = {, , , } introduced in Definition 2.4.3, and let
CPL be the logic over the signature 1 = {, , , , } obtained from CPL by
adding, as in Theorem 3.4.10, axiom schema (Ax10). The system C1s defined over
the signature 1 is obtained from CPL by adding axiom schemas (neg1 ), (ca1 ),
(ca2 ), (ca3 ), (ca1 ), (ca2 ), (ca3 ), (cf) and (dneg):

(dneg)


Theorem 3.6.14 The logic C1s is a dC-system based on CPL satisfying consistency
propagation for every connective of its signature.

Proof Given the already obtained results, it suffices to prove that axiom (dneg)
characterizes consistency propagation for , namely . But this is obvious,
after observing the following table which describes all the relevant scenarios in
CPL :
3.6 Propagating Consistency 111

1 1 0 0 1 1 v1
0 1 0 1 1 v2
0 1 1 1 1 0 v3
0 1 v4

Clearly, the propagation of consistency for the strong negation is equivalent to

forbidding the valuation v3 . And, the latter is equivalent to requiring the validity of
axiom (dneg). 

From the previous table, it is immediate to see that C1s is a strong LFI. The letter
s stands for strong C1 since, in fact, this system can be viewed as a strong version
of da Costas logic C1 . Indeed, analogous to the original system C1 , this system sat-
isfies the double-negation elimination with respect to the paraconsistent negation ,
and it contains classical logic (but now explicitly, in contrast to C1 ). Moreover, con-
sistency (which is defined as ( p p) instead of ( p p)) propagates through
all the connectives of the signature, as in C1 . The fact that consistency is expressed
by the strong negation of contradiction instead of the weak negation of contradic-
tion, justifies the name of the calculus. Of course the original laws for consistency
propagation could be directly used in the formulation of C1s , provided that is just
an abbreviation in the system:

Corollary 3.6.15 The logic C1s can be alternatively axiomatized by using CPL
plus the axioms (ca ), (ca ), (ca ) and (ca ) found in Definition 3.6.1, together
with
(ca )

taking into account that p is now an abbreviation for the formula ( p p).

3.7 da Costas Hierarchy and Consistency Propagation

This section is dedicated to formally defining the well-known hierarchy of calculi

Cn (for 1 n ) introduced by da Costa in [10], informally mentioned in several
places of this book, and to elucidate how da Costas systems can be regarded as
particular cases of LFIs.
For the readers convenience, we change the order in which the systems were orig-
inally presented, starting by the weakest of them, namely C . Recall the signatures
+ = {, , } and 0 = {, , , }. Let us begin by recalling the definition
of positive intuitionistic logic IPL+ .

Definition 3.7.1 (Positive intuitionistic logic, IPL+ ) The calculus IPL+ for positive
intuitionistic logic is defined over the signature + by dropping axiom schema (Ax9)
from CPL+ (recall Definition 2.1.12). That is, IPL+ is obtained by the following
112 3 Some Extensions of mbC

Hilbert calculus:
Axiom schemas:
 
(Ax1)
      
(Ax2)
  
(Ax3)
 
(Ax4)
 
(Ax5)
 
(Ax6)
 
(Ax7)
    
( ) ( ) (Ax8)

Inference rule:

(MP)



Some properties of IPL+ will be studied in Sect. 5.1 of Chap.5.

Definition 3.7.2 (System C , [10]) The system C is defined over the signature 0
by adding to positive intuitionistic logic IPL+ the following axiom schemas:

(Ax10)

(cf)

Loparic presented in 1976 (though the results wereonlypublished in 1986, see

[11]) a sound and complete valuation semantics over 0, 1 for C . By his turn, a
characterization of C by a Kripke-type semantics was offered by Baaz in [12].
The notion of well-behavedness operators, precursor of the concept of consis-
tency operators, was defined by means of schema formulas combining the conjunc-
tion and negation connectives. Each calculus Cn has its own well-behavedness
operator, defined inductively as follows:

Definition 3.7.3 Let be a formula in L0 , and consider the following abbreviations:
(1) = ( );
def

(2) 0 = , and n+1 = ( n ) , for 0 n < ;

def def

(3) (1) = , and (n+1) = (n) n+1 , for 1 n < .

def def

3.7 da Costas Hierarchy and Consistency Propagation 113

Thus, for instance, 1 = (1) = . On the other hand, 2 = and (2) =

, while 3 = and (3) = ( ) .

Definition 3.7.4 (System Cn , [10]) Let 1 n < . The system Cn is defined over
the signature 0 by adding to C the following axiom schemas:

(n) (( ) (( ) )) (Axwb)
 (n)   
(n) ( )(n) ( )(n) ( )(n) (Axwbp)


The consequence relation in the calculus Cn will be denoted by Cn .

Remark 3.7.5 By defining (0) = for every formula , C0 in Definition 3.7.4 is

def

precisely a formulation of classical propositional logic over signature 0 .

It is an easy task to prove that the schema formula ( ) is derivable in
every calculus Cn , for n < . Hence, Cn (for 1 n < ) is a proper extension of
IPL+ and, moreover, also a proper extension of positive classical propositional logic
CPL+ .  
The valuation semantics over 0, 1 for C1 introduced in [13, 14] was extended
in [15] to the whole hierarchy Cn (for 1 n < ) as follows:

Definition
 3.7.6 (Valuations for Cn , [15]) Let 1 n < . A function v : L0
0, 1 is a valuation for Cn , or a Cn -valuation, if it satisfies clauses (vAnd), (vOr),
(vImp) and (vNeg) from Definition 2.2.1, (vCf ) (introduced after Remark 3.5.16),
plus the following clauses:
(vWB1 ) v( n1 ) = v(( n1 )) v( n ) = 0
(vWB2 ) v() = v() v(( )) = 1
(vWBp ) v()
 = v() and v() = v() = v(#) = v((#)), for #
, , .
The consequence relation w.r.t. Cn -valuations will be denoted by |=Cn . 
Theorem 3.7.7 (Soundness and Completeness of Cn w.r.t. valuations, [15]) Let 1
n < . For every  {} L0 :

 Cn  |=Cn .

Thanks to this valuation semantics, it is possible to prove that da Costas hierarchy

is organized as follows:

C Cn+1 Cn C2 C1 C0  CPL

where the inclusion are strict.

The following result can be easily proved by employing the usual techniques. The
proof is left to the reader.
114 3 Some Extensions of mbC

Theorem 3.7.8 For each 1 n < , the logic Cn is a (strong) LFI w.r.t. and
n p = p (n) . Moreover, each Cn is a dC-system based on CPL+ w.r.t. and n p
def

(recall Definition 3.3.5).

Remark 3.7.9 It should be noticed that, by the properties of the conjunction, each Cn
def  
(for 1 n < ) can also be seen as a strong LFI w.r.t. and n ( p) = p 1 , . . . , p n .
This shows that the calculi Cn , for 2 n < , are examples of LFIs in which the con-
sistency operator can be defined (in a non-trivial way) by means of a set of formulas
which is not a singleton. Because of this, the following holds:

, C1 in general, but always , , 1 C1 ;

, , 1 C2 in general, but aways , , 1 , 2 C2 .

Moreover, for 1 n < ,

, , 1 , . . . , n1 Cn in general, but always , , 1 , . . . , n Cn .

The last remark suggests that each calculus Cn (for n < ) has a paraconsistency
degree which can be described as follows: given a formula in L0 , define, for every
0 k < , the k-depth contradiction of as being the formula ck () = k ( k ).
def

Observe that c0 () is just .

By axiom (cf), it is easy to see that

( k ) Cn ck1 () and so ck () Cn ck1 ()

for every 1 k, n < . From this, the following properties hold in da Costas hier-
archy of paraconsistent logics:

c0 () C1 in general, but always c1 () C1 ;

c1 () C2 in general, but always c2 () C2 .

Moreover, for 1 n < ,

cn1 () Cn in general, but always cn () Cn .

This shows that, in general, just contradictions of depth n (or higher) trivialize the
calculus Cn . Thus, the hierarchy is increasingly paraconsistent in a certain way, and
the degree of paraconsistency of Cn is, according to this, precisely n. In this way,
if n > m then Cn has a higher paraconsistence degree than Cm , what amounts to say
that Cn has a higher tolerance to contradictions than Cm . Under this perspective, C0
(that is, classical logic) has no tolerance at all to any contradictions.
A natural question is: which is the logic possessing the highest tolerance to con-
tradictions? It would be of course a kind of limit of the hierarchy. Recalling that C
3.7 da Costas Hierarchy and Consistency Propagation 115

is strictly contained in every other Cn , and taking into account that the hierarchy is
strictly decreasing, C is a natural candidate to the limit of the hierarchy. Indeed, as
da Costa himself wrongly assumed, in terms of consequence relations one is tempted
to conjecture that:
C = Cn .
n<

As it was proved in [16], however, this is not the case. In that paper, the above limit of
the calculi Cn , for 1 n < is referred to as CLim , and semantically characterized
in terms of possible-translations (see Sect. 6.8 for a definition of possible-translation
semantics). By the very definition of CLim , it is immediate to see that, in general,

, , 1 , . . . , n CLim

for every fixed 1 n < , but always

, , 1 , . . . , n , n+1 , . . . CLim .

That is, CLim is a non-finitary logic which could be considered as a (strong) LFI in
a wider sense, where consistency is given by the infinite set
 
Lim ( p) = pn : 1 n < .
def

Indeed, there are formulas and (for instance, by taking and as two dif-
ferent propositional variables) such that cn () CLim , for every 0 n < , but
, , Lim () CLim , for every and .

3.8 A Stronger Consistency Propagation

As firstly observed by Bziau in [17], a stronger property of consistency propagation

can be perceived: instead of requiring every immediate subformula of a complex
formula to be consistent in order to have the consistency of that formula, we could
require that some immediate subformulas be consistent. If the signature of the logic
under analysis only has unary and binary connectives (as in the case of ), the only
propagation rules affected are those concerning the binary connectives. In formal
terms:

Definition 3.8.1 (Strong consistency propagation) Let L be an extension of mbC

defined over the signature  which satisfies the propagation of consistency for
and (see Definition 3.6.1). The strong propagation of consistency of each binary
connective of  in L is provided by the following axiom schemas:
116 3 Some Extensions of mbC

( ) ( ) (co1 )
( ) ( ) (co2 )
( ) ( ) (co3 )


From the semantical point of view, a valuation semantics for a logic L , as in
Definition 3.8.1, must satisfy the following clauses:
(vCa ) v() = 1 = v() = 1.
(vCa ) v() = 1 = v() = 1.
(vCa1 ) v() = 1 or v() = 1 = v(( )) = 1.
(vCa2 ) v() = 1 or v() = 1 = v(( )) = 1.
(vCa3 ) v() = 1 or v() = 1 = v(( )) = 1.

Remark 3.8.2 Axioms (co1 )(co3 ) (where is a derived connective and not a primi-
tive connective) were proposed by Bziau in [17] for C1 : specifically, he proposes to
substitute in C1 the three axioms of consistency propagation for binary connectives
with their stronger version. This variant of C1 was additionally studied in [18], under
the name C1+ . The counterpart of C1+ in signature  (that is, the system obtained
from mbC by adding axiom schemas (ci), (cl), (cf) and (co1 )(co3 ) is called Cilo
(see [1, 3]). Observe that Cilo satisfies both (ca ) and (ca ), because of (ci) and (cf),
respectively. Clearly, Cilo is a strong LFI.
Since a logic L satisfying the requirements of Definition 3.8.1 automatically
satisfies the requirements of Definition 3.6.1 (but the converse does not hold), we
can speak of a stronger propagation property.
Proposition 3.8.3 Let L be an extension of mbC defined over the signature . If
L satisfies strong propagation of consistency (see Definition 3.8.1) then it satisfies
propagation of consistency (see Definition 3.6.1). The converse statement, in general,
is not true.
Proof Suppose that L satisfies strong consistency propagation. As L
for every formula and , it satisfies consistency propagation.
Now, let L be the logic Cila, which corresponds to C1 in signature  (recall
Remark 3.6.3), and consider the logic Cilo, which is the counterpart of C1+ in sig-
nature  (see Remark 3.8.2). If Cila, which satisfies consistency propagation, also
satisfies strong consistency propagation, then it contains Cilo. However, by adapt-
ing an argument in [18] for C1+ and C1 , the formula ( p1 p2 ) ( p1 p2 )
is a theorem in Cilo but not in Cila. This shows that Cila does not satisfy strong
consistency propagation, despite satisfiying consistency propagation. 
Of course, in order to obtain consistency of a given complex formula, the require-
ments in the second definition are weaker than those in the first: consistency of
just one subformula is enough, as the following proposition shows (compare with
Proposition 3.6.2):
3.8 A Stronger Consistency Propagation 117

Proposition 3.8.4 Let L be an extension of mbC defined over  which satisfies the
consistency propagation for the unary connectives of , and the strong propagation
of consistency for the binary connectives of . Then:

if is any subformula of L . Equivalently: if p V ar occurs in , then

p L .

Proof By induction on the complexity of . The details of the proof are left to the
reader. 
As was done with respect to consistency propagation, the next step is to determine
the minimum requirements on the connectives , and for achieving strong
consistency propagation, starting again from mbCciw.
Let us begin with axiom (co1 ), which concerns strong propagation of consistency
for conjunction. Given arbitrary formulas and , there are exactly six relevant
scenarios in mbCciw to be analyzed, in which either or are true and, at the
same time, is also true:

( ) ( )
1 1 1 0 0 1 1 1 0 v1
0 1 v2
0 1 1 0 1 1 0 v3
0 1 v4
0 1 1 1 1 0 v5
0 1 v6

Observe that valuations v1 , v3 and v5 should be blocked, as they violate (co1 ). With
respect to v1 and v5 , adding the following additional requirement to the valuations is
sufficient:
v() = 1 and v() = 0 implies v(( )) = 0 .

By contraposition, this is equivalent to the following:

(vCo11 ) v(( )) = 1 = v() = 0 or v() = 1.

The axiom schema corresponding to this clause is clearly the following:

( ) ( )

With respect to v3 , it is enough to consider the following additional requirement

in the valuations:

v( ) = 1 and v() = 0 implies v(( )) = 0 .

118 3 Some Extensions of mbC

By contraposition, this is equivalent to the following:

(vCo12 ) v(( )) = 1 = v( ) = 0 or v() = 1.

The associated axiom schema is the following:

( ) (( ) )

Therefore, the strong propagation of consistency through conjunctions can be charac-

terized by the following axioms, which do not explicitly use the consistency operator:

Theorem 3.8.5 The extension of mbCciw obtained by adding axiom schema (co1 )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schemas:

( ) ( ) (co11 )
( ) (( ) ) (co12 )

Through analogous reasoning, it is easy to characterize the strong consistency

propagation for disjunction and implication:

Theorem 3.8.6 The extension of mbCciw obtained by adding axiom schema (co2 )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schemas:

( ) ( ) (co21 )
( ) (( ) ) (co22 )
( ) ( ) (co23 )
( ) ( ) (co24 )

Theorem 3.8.7 The extension of mbCciw obtained by adding axiom schema (co3 )
is equivalent to the extension of mbCciw obtained by adding the following axiom
schemas:

( ) ( ) (co31 )
( ) (( ) ) (co32 )
( ) ( ) (co33 )
( ) ( ) (co34 )

It was claimed that in [18] it is possible to define a non-trivial congruence for C1+ .
The existence of such congruence would give us a chance to obtain some kind of
algebraizability for this logic (see [19]), contrary to what happens with the weaker
logic C1 . Unfortunately, this is not the case since the congruence relation proposed
in [18] fails to be reflexive, as we shall see in Sect. 4.3 of Chap. 4. In view of such
3.8 A Stronger Consistency Propagation 119

circumstances, the question of establishing a satisfactory algebraic semantics for C1+

in a traditional sense is still a challenging open problem.
Some solutions are possible, though, in an innovative sense. Taking into account
that the possible-translations semantics (see Sect. 6.8 of Chap. 6) is an apt tool for
analyzing and providing semantical meaning and algebraic contents to certain com-
plex logics, and particularly to paraconsistent logics, a new notion of algebraizabil-
ity, called possible-translations algebraizability, was proposed in [20]. This method
obtains algebraic semantics up to translations, applicable to several non-classical
logics and particularly to several paraconsistent logics.
The method has been put in more formal mathematical tools in [21], where the
possible-translations semantics and the concept of possible- translations algebraiz-
ability are versed in categorial terms, showing that the (weak) product of finitely-
algebraizable (or Blok-Pigozzi algebraizable) propositional logics is also finitely-
algebraizable, under certain conditions.
Two related approaches to that of possible-translations semantics, namely Fidel
structures and swap structures, are investigated in Chap. 6.
A different approach to abstract algebraic logic is developed in [22] by means
of a certain behavioral approach centered around the weaker notion of behavioral
equivalence, a method with roots in computer science.
Although a common generalization of all such methods (as well as a deeper
investigation on their scope) is still missing, such new methods show that algebraic
notion of congruence has its own limits, but that such limits do not block the deeply
rooted relationship between logic and algebra.
The next chapter is devoted to the question of algebraizability of some LFIs by
means of standard tools. As mentioned above, alternative semantical characteriza-
tions for LFIs, some of them having an algebraic nature, will be provided in Chap. 6.

References

1. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Inconsis-
tency. In Handbook of Philosophical Logic, ed. Dov M. Gabbay and Franz Guenthner (2nd.
edn.), vol. 14, 193. Springer. doi:10.1007/978-1-4020-6324-4_1.
2. Avron, Arnon. 2005. Non-deterministic matrices and modular semantics of rules. In Logica
Universalis, ed. Jean-Y. Bziau, 149167. Basel: Birkhuser Verlag.
3. Carnielli, Walter A., and Joo Marcos. A taxonomy of C-systems. In [Carnielli, Walter A.,
Marcelo E. Coniglio, and Itala M. L. DOttaviano, eds. 2002. Paraconsistency: The Logical
Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP
2000), Vol. 228 of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New
York], 194.
4. Avron, Arnon, Beata Konikowska, and Anna Zamansky. 2011. Analytic calculi for basic logics
of formal inconsistency. In Logic without Frontiers: Festschrift for Walter Alexandre Carnielli
on the occasion of his 60th birthday, ed. Jean-Yves Beziau and Marcelo Esteban Coniglio, vol.
17 of Tribute Series, 265275. College Publications.
5. Avron, Arnon, Beata Konikowska, and Anna Zamansky. Modular construction of cut-free
sequent calculi for paraconsistent logics. In Proceedings of the 2012 27th annual ACM/IEEE
120 3 Some Extensions of mbC

symposium on logic in computer science, LICS 2012, 8594. IEEE, 2012. doi:10.1109/LICS.
2012.20.
6. Avron, Arnon, Beata Konikowska, and Anna Zamansky. 2013. Cut-free sequent calculi for
C-systems with generalized finite-valued semantics. Journal of Logic and Computation 23(3):
517540. doi:10.1093/logcom/exs039.
7. Avron, Arnon, Beata Konikowska, and Anna Zamansky. 2015. Efficient reasoning with incon-
sistent information using C-systems. Information Sciences 296: 219236. doi:10.1016/j.ins.
2014.11.003.
8. Batens, Diderik. 1980. Paraconsistent extensional propositional logics. Logique et Analyse
9091: 195234.
9. Marcos, Joo. 2008. Possible-translations semantics for some weak classically-based paracon-
sistent logics. Journal of Applied Non-Classical Logics 18(1): 728.
10. Newton C. A. da Costa. 1993. Sistemas formais inconsistentes (Inconsistent formal systems,
in Portuguese). Habilitation thesis, Universidade Federal do Paran, Curitiba, Brazil, 1963.
Republished by Editora UFPR, Curitiba, Brazil, 1993.
11. Loparic, Andra. 1986. A semantical study of some propositional calculi. The Journal of Non-
Classical Logic, 3(1):7395. http://www.cle.unicamp.br/jancl/.
12. Baaz, Matthias. 1986. Kripke-type semantics for da Costas paraconsistent logic C . Notre
Dame Journal of Formal Logic 27(4): 523527.
13. da Costa, Newton C.A. and Elias H. Alves. 1976. Une smantique pour le calcul C1 (in French).
In Comptes Rendus de lAcadmie de Sciences de Paris (A-B), 283:729731.
14. da Costa, Newton C.A., and Elias H. Alves. 1977. A semantical analysis of the calculi Cn .
Notre Dame Journal of Formal Logic 18(4): 621630.
15. Loparic, Andra and Elias H. Alves. The semantics of the systems Cn of da Costa. In [Arruda,
Ayda I., Newton C. A. da Costa, and Antonio M. A. Sette, eds. 1980. Proceedings of the Third
Brazilian Conference on MathematicalLogic, Recife 1979. Sociedade Brasileira de Logica,
Campinas], 161172.
16. Carnielli, Walter A., and Joo Marcos. 1999. Limits for paraconsistent calculi. Notre Dame
Journal of Formal Logic 40(3): 375390.
17. Bziau, Jean-Yves. 1990. Logiques construites suivant les mthodes de da Costa. I. Logiques
paraconsistantes, paracompletes, non-althiques construites suivant la premire mthode de da
Costa (in French). Logique et Analyse (N.S.), 131/132:259272.
18. da Costa, Newton C. A, Jean-Yves Bziau, and Otvio Bueno. 1995. Aspects of paraconsistent
logic. Bulletin of the IGPL, 3(4):597614.
19. Font, Josep Maria, and Ramn Jansana. 2009. A General Algebraic Semantics for Sentential
Logics. Vol. 7 of Lecture Notes in Logic, 2nd edn. Ithaca, NY, USA: Association for Symbolic
Logic.
20. Bueno-Soler, Juliana, and Walter A. Carnielli. 2005. Possible-translations algebraization for
paraconsistent logics. Bulletin of the Section of Logic 34(2): 7792.
21. Bueno-Soler, Juliana, Marcelo E. Coniglio, and Walter A. Carnielli. 2007. Possible-translations
algebraizability. In Handbook of paraconsistency Jean-Yves Bziau, Walter A. Carnielli, and
Dov M. Gabbay, editors, vol. 9 of Studies in logic (Logic & Cognitive Systems), 321340.
College Publications.
22. Caleiro, Carlos, and Ricardo Gonalves. 2009. Behavioral algebraization of da Costas C-
systems. Journal of Applied Non-Classical Logics 19(2): 127148.
Chapter 4
Matrices and Algebraizability

This chapter examines the question of characterizability of mbC by means of finite

matrices, as well as the algebraizability of mbC and its extensions. Firstly, some
negative results, in the style of J. Dugundjis famous theorem for modal logics, will
be obtained for several extensions of mbC. This means that, for a wide variety of
LFIs extending mbC, it is impossible to find a semantical characterization using finite
matrices. Despite this, it is possible to define a vast family of 3-valued extensions of
mbC, most of them being algebraizable in the sense of Blok and Pigozzi (see [1]).
This is an interesting result, taking into account that several extensions of mbC
cannot be algebraizable in this sense, as in the case for da Costas paraconsistent
systems Cn .

4.1 Logical Matrices

According to Anellis [2], the idea of logical matrices goes back to C. Pierce who
in 1902 developed a truth-table device for the binary connectives of his algebra of
logic (there is some polemics, however, around the origins of truth-tables, sometimes
credited to E. Schrder, L. Wittgenstein and/or B. Russell). In a rigorous and general
way the notion of a logical matrix was defined by J. ukasiewicz and A. Tarski in
1930. The definition given in this Section is (up to some inessential details) the same
as that of ukasiewicz and Tarski.

Definition 4.1.1 Let  be a propositional signature. An algebra for  is a pair

A = A, O such that A is a nonempty set (the domain of the algebra) and O is a
mapping which assigns, to each n-ary connective c of , an operation cA : An A
in A. 

When there is no risk of confusion, the same symbol will be used for a connective
c and its interpretation O(c) = cA . Moreover, if  is finite, the mapping O will be
Springer International Publishing Switzerland 2016 121
W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_4
122 4 Matrices and Algebraizability

substituted by the list of connectives of  itself. Thus, an algebra for the signature
0 will be written as A = A, , , , 1, for instance.

Definition 4.1.2 Let  be a propositional signature. A logical matrix (or a matrix

semantics, or simply a matrix) over  is a triple M = A, D, O such that A, O
is an algebra for  and D is a subset of A. Elements of D are called designated or
distinguished. 

By simplicity, and when the context is obvious, a logical matrix will be simply
denoted by M = A, D. A logical matrix M defines a Tarskian logic over the
language L as follows:  |=M iff, for every homomorphism h : L A (where
A is the algebra underlying M), if h[] D then h() D. In particular, is valid
in the logic generated by M iff h() D for every homomorphism h. Given a matrix
M = A, D, O, an homomorphism h : L A is called a valuation over M

4.2 Uncharacterizability by Finite Matrices

A Tarskian propositional logic L with a consequence relation is said to be char-

acterized by a matrix M if  = |=M . If M is a class of matrices over  then L
is characterized by M if = MM |=M . That is,  iff  |=M for every
M M.
+
 L {CPL , CPL, CPL }. Then L is characterized by the log-
Example 4.2.1 Let
ical matrix over 0, 1 defined after Definition 2.4.4, by considering the tables for the
corresponding signature (+ , c or c ), and where 1 is the only designated value.

Wjcicki proved that any Tarskian propositional logic can be characterized by a

class of logical matrices (see, for instance [3]). This means that the class of Tarskian
logics and the class of logics which are defined by families of matrices are the same.
If L is characterized by a single finite matrix M then L is clearly decidable by
using the truth-tables of M. This is why characterizability by a finite matrix is an
important feature of a logic. In 1940, Dugundji showed that no modal system lying
between logics S1 and S5 can be characterized by a single finite logic matrix (see [4]),
from which an alternative semantical characterization should be find for them. The
scope of this theorem was enlarged to other modal systems (see, for instance [5]
and [6]). The aim of this section is to prove some Dugundji-like theorems applied to
LFIs.
Definition 4.2.2 For any formula L and n 0 let n be the formula of L
which is recursively defined as follows: 0 = , and n+1 = (( n ) ( n )). 
m
Definition 4.2.3 For any sequence (i )i0 of formulas in L and m 0 let i=0 i
0
be the formula of L
    which is recursively
defined as follows: i=0 i = 0 , and
m+1 m
i=0 i = i=0 i m+1 , for m 0. 
4.2 Uncharacterizability by Finite Matrices 123

Definition 4.2.4 ((m) formulas) Let be a formula of L . For every m 0 let

(m) be the formula of L given by

m
(m) = i m+1 .
def

i=0

Lemma 4.2.5 Let L be a logic over the signature  containing mbC, and let M be
a matrix semantics over  with n truth-values (where n 2) such that L is sound
for M. If p is a propositional variable and m n 1 then p(m) is valid in M.

Proof Let v be a valuation over M and ai = v( pi ), for 0 i m + 1. Since the

domain M of M has n elements, with n m + 1, then there are i and j, with 0 i <
j m + 1, such that ai = a j . Let i be the minimum index such that 0 i < m + 1
and ai = a j for some j such that i < j m + 1. Let j be the minimum index such
that i < j m + 1 and ai = a j . Then, through the construction of formulas p k
and given the fact that v is a homomorphism, it follows that ai+k = a j+k for every
0 k m + 1 j. In particular,

am+1 {a0 , . . . , am } ()

On the other hand, as L is sound for M and L extends mbC, the valuation v satisfies
the following for every formulas and in L :
(i) v( ) D iff v() D and v() D;
(ii) v( ) D iff v() / D or v() D,
where D is the set of designated values of M. Thus, if am+1 D, then v( p(m)) D
m
by (ii). Otherwise, if am+1 / D, then v i=0 p i
/ D, by () and by (i), and so
v( p(m)) D, by (ii). In any case, it holds that v( p(m)) D for every valuation v,
and thus p(m) is valid in M. 

Definition 4.2.6 Let be a formula which depends exclusively on the proposi-

tional variable p. Then, Ciaeci is the system defined over  by adding the axiom
schemas (ci), (cf), (ce), (ca ), (ca ), (ca ) (recall Definition 3.6.1) and (ci ) to
mbC. 

It is immediate to see that the logic Ciaeci is a strong LFI.

Definition 4.2.7 Fix a natural number n N. The matrix Mn = M, D is such

that M = N { 21 } and D = N. Consider the usual order in N and define x 21 ,
for every x M. Then the operations in Mn are defined as follows:
124 4 Matrices and Algebraizability

0 if x = n and y = n + 1
x n y =
max(x, y) otherwise

x n y = min(x, y)

1

if x N and y = 21

y
2
if x = 21 andy N
x n y =
0
if x = 21 = y


max(x, y) otherwise

2
1
if x = 0
n x = 0 if x = 21


x + 1 otherwise


0 if x {0, 21 }
n x = 1
2
otherwise

Lemma 4.2.8 The logic Ciaeci is sound for each Mn .

Proof We begin by observing the following

Facts: For every x, y M:
(i) x n y D iff x D and y D;
(ii) x n y D iff x D or y D;
(iii) x n y D iff x / D or y D.
The details of this easy proof are left to the reader. Recall now the axioms of mbC from
Definition 2.1.12. From Facts(iii), it follows that axiom (Ax9) is valid in Mn . More-
over, Mn satisfies the deduction meta-theorem w.r.t. n . It is, thus, immediate that
axioms (Ax1) and (Ax2) are valid in Mn . From Facts(i) the validity of axioms (Ax3),
(Ax4) and (Ax5) in Mn is immediate. Additionally, axioms (Ax6), (Ax7) and (Ax8)
are valid in Mn because of Facts(ii). Now, if x = 21 , then n x = 0 D and, cor-
respondingly, x n n x D for every x, showing that axiom (Ax10) is valid in
Mn . Concerning axiom (bc1), if n x D, then x {0, 21 }. Suppose that x = 0;
then, n x = 21 / D and (bc1) becomes validated by Mn . Now, if n n x D,
then n n x = 0 and n x = 21 . From this, x = k + 1 for some k N and hence
n x = k + 2. Thus, (x n n x) D, showing that axiom (ci) is valid in Mn . After
a quick inspection of the table below, it becomes evident that Axioms (cf) and (ce)
are valid in Mn .
4.2 Uncharacterizability by Finite Matrices 125

x n x n n x n n x n x x n n n x
1 1
2 0 2 0 0
1
0 2 0 0 0
k+1 k+2 k+3 k+3 k+3

Now, by observing that n x D iff x {0, 21 }, it follows that axioms (ca ), (ca )
and (ca ) are valid in Mn . See the table below.

x y x n y x n y x n y
0 0 0 0 0
0 21 1
2 0 1
2
1 1
2 0 2 0 0
1 1 1 1
2 2 2 2 0

Axiom (ci ) is also valid in Mn . Firstly, observe that the formula , interpreted
as ny = (y n n y) n n y where y M interprets , always takes the value 21 for
any interpretation of . Accordingly, the table for the strong negation interpreted
in Mn is as follows:

x ny x = x n ny
1
0 2
k+1 1
2
1
2 0

From this, the table below shows that (ci ) is valid in Mn .

x n x x n n x ny x = ny (x n n x) n ny x n ny x n (x n n x)
1 1 1
0 2 2 0 2 0
k+1 k+2 k+2 1
2 0 k + 2
n n+1 0 1
2 0 0
1 1 1
2 0 2 0 2 0

Finally, from Facts(iii), rule MP is sound in Mn , concluding the proof. 

Lemma 4.2.9 No formula of the form p(m) is derivable in Ciaeci , for m 0 and
a propositional variable p.

Proof We will prove that M2m+1 falsifies the formula p(m). Thus, consider a valu-
ation vn over Mn where n = 2m + 1, such that vn ( p) = 1.
Fact: For every m 0 it holds then that vn ( pi ) = 2i + 1, for 0 i m, while
vn ( p m+1 ) = 21 .
126 4 Matrices and Algebraizability

The proof of the Fact is done by induction on m. If m = 0 then v1 ( p 0 ) = v1 ( p) = 1 =

2.0 + 1. On the other hand, v1 ( p 1 ) = 1 (v1 ( p 0 ) 1 1 v1 ( p 0 )) = 1 (1 1 1 1) =
1 (1 1 2) = 1 0 = 21 .
Suppose the result is true for 0 k m. We will prove the case for m + 1.
Thus, let n  = 2(m + 1) + 1 = n + 2, where n = 2m + 1. Observe that, by induc-
tion hypothesis and by the definition of the matrices, vn  ( pi ) = vn ( pi ) = 2i + 1 for
0 i m. In particular, vn  ( p m ) = n and then vn  ( p m+1 ) = n  (n n  (n + 1)) =
n  (n + 1) = n + 2 = 2(m + 1) + 1. Finally, vn  ( p m+2 ) = n  (n  n  (n  + 1)) =
n  0 = 21 , proving the Fact.
    
m
From this, vn ( p(m)) = n i=0 (2i + 1) n 21 = n n 21 = 21 / D. This
shows that p(m) is not valid in Mn . But Ciaeci is sound for Mn , because of
Lemma 4.2.8, and so p(m) cannot be derivable in Ciaeci for every m 0. 

We thus arrive at our first Dugundji-like theorem for LFIs:

Theorem 4.2.10 Let L be a propositional logic defined over the signature . If
L lies between mbC and Ciaeci , then L is not characterizable by a finite matrix
semantics.

Proof Suppose, by contradiction, that L is characterizable by a finite matrix seman-

tics M with, say, m 2 truth-values (the case m = 1 is ruled out, since L cannot be
trivial). By Lemma 4.2.5, the formula p(m) is valid in M for any propositional vari-
able p, and is then derivable in L . Therefore, p(m) is derivable in Ciaeci , which
contradicts Lemma 4.2.9. From this, we conclude that L cannot be characterizable
by a finite matrix semantics. 

The last theorem encompasses most of the systems analyzed thus far taking
into account that Ciaeci extends mbCcici . However, the systems which include
axiom (cl) lie outside the scope of Theorem 4.2.10. As was done in [7], it is possible
to slightly modify the proof above to obtain a similar theorem for several systems
which include axiom (cl). We begin by introducing the strongest logic within the
scope of the second theorem of non-characterizability by finite matrices.
Definition 4.2.11 Let Cilaeci be the system obtained from Ciaeci (see Defini-
tion 4.2.6) by adding axiom schema (cl). 
Clearly, the logic Cilaeci is a strong LFI.

Definition 4.2.12 Let n be a natural number. The matrix Mn = M, D, O   is

defined as the matrix Mn = M, D, O of Definition 4.2.7, but now the conjunction
is defined as follows:

 0 if x N and y = x + 1
x n y =
max(x, y) otherwise


4.2 Uncharacterizability by Finite Matrices 127

Lemma 4.2.13 The logic Cilaeci is sound for each Mn .

Proof Analogous to the proof of Lemma 4.2.8. By taking into account that Facts(i) in
the proof of Lemma 4.2.8 is still valid for the new conjunction, we must still analyze
the other axioms explicitly using the new conjunction, namely (cl), (ci) and (ci ).
The following table is enough to prove that (cl) is valid in Mn :

x n x x n n x n (x n n x) n x n (x n n x) n n x
1 1
0 2 2 0 0 0
k+1 k+2 0 1
2
1
2 0
1 1
2 0 2 0 0 0

With respect to (ci), the following table shows that this axiom is valid in Mn .

x n x x n n x n x n n x n n x n (x n n x)
1 1 1
0 2 2 0 2 0
k+1 k+2 0 1
2 0 0
1 1 1
2 0 2 0 2 0

Finally, the validity of axiom (ci ) in Mn is proved by table below.

x n x x n n x ny x = ny (x n n x) n ny x n ny x n (x n n x)
1 1 1
0 2 2 0 2 0
k+1 k+2 0 1
2 0 0
1 1 1
2 0 2 0 2 0

Definition
m 4.2.14 For any sequence (i )i0 of formulas in L and 0 m 0, let

i=0 i be the formula
  of L which is
recursively defined as follows: i=0 i = 0 ,
m+1 m 
and i=0 i = i=0 i m+1 , for m 0. 

Definition 4.2.15 ( [m] formulas) For 0 i < j let i j be the formula of L given
by ( pi p j ) ( pi p j ). For m 1, let

[m] = (i j pm+1 ).
def

1i< jm

Lemma 4.2.16 For every m 1, the formula [m] is not derivable in Cilaeci .
128 4 Matrices and Algebraizability

Proof Fix m 1 and let v be a valuation over the matrix logic Mm of Defini-
tion 4.2.12 such that v( pi ) = i, for 1 i m, and v( pm+1 ) = 21 . If 1 i < j m
then

v(i j ) = m (i m m j) m (i m m j) = m ( j + 1) m ( j + 1) = j + 2

and v(i j pn+1 ) = ( j + 2) m 21 = 21 accordingly. From this, v( [m] ) = 21 .

This means that [m] cannot be derived in Cilaeci , as Lemma 4.2.13 shows that
this logic is sound for Mm . 

The weakest logic within the scope of the second theorem of non-characterizability
is now introduced by means of finite matrices.
Definition 4.2.17 Let Cl be the system defined over  which is obtained from
mbC by replacing axiom schema (Ax10) with (cl). Equivalently, Cl is obtained
from CPL+ by adding axiom schemas (bc1) and (cl). 
It is easy to see that the logic Cl is a strong LFI.
Lemma 4.2.18 Let and be formulas in L . Then ((( ) ( ))
is a theorem of Cl .

Proof Observe that Cl satisfies the deduction meta-theorem. It is, therefore, enough
to prove that ( ), Cl . Consider the following sequence:

1. ( ) (hyp)
2. (hyp)
3. ( ) (Ax4)
4. ( ) (Ax5)
5. (MP 2,3)
6. (MP 2,4)
7. ( ) (cl)
8. (MP 1,7)
9. ( ( )) (bc1)
10. ( ) (MP 8,9)
11. (MP 5,10)
12. (MP 6,11)

This shows that ( ), Cl , as required. 

Lemma 4.2.19 Let L be a logic over the signature  containing Cl , and let M
be a matrix semantics over  with n 2 truth-values such that L is sound for M.
If m > n, then [m] is valid in M.

Proof Let v be a valuation over M and ai = v( pi ) for 1 i m. Since the domain

M of M has n elements and m > n, there are i and j with 1 i < j m such
that ai = a j and then v((i j pn+1 ) = (((ai ai ) (ai ai )) b with
4.2 Uncharacterizability by Finite Matrices 129

b = v( pn+1 ) (for simplicity, we are using the same symbol for the connectives of
 and their interpretation in M). But ((( pi pi ) ( pi pi )) pn+1 is a
theorem of Cl , by Lemma 4.2.18, and is thus a theorem of L . Consequently,
v((i j pn+1 ) D, where D is the set of designated truth-values of M. Given
that L is sound for M and L extends Cl , the valuation v is such that

v( ) D iff v() D or v() D

for every formula and in L . From here, v( [m] ) D by definition of formula

[m] . Therefore [m] is valid in M. 

Our second Dugundji-like theorem for LFIs is as follows:

Theorem 4.2.20 Let L be a propositional logic defined over the signature . If
L lies between Cl and Cilaeci then L is not characterizable by a finite matrix
semantics.

Proof Suppose, by contradiction, that L is characterizable by a finite matrix seman-

tics M with, say, n 2 truth-values (as in Theorem 4.2.10, the case n = 1 is dis-
carded since L is not trivial). If m > n, then the formula [m] is valid in M, following
Lemma 4.2.19, and is, accordingly, derivable in L . From here, it follows that [m]
is derivable in Cilaeci , which contradicts Lemma 4.2.16. Therefore, L cannot be
characterizable by a finite matrix semantics. 

By observing that Cilaeci extends mbCcici cl, we see that, besides the systems
studied in [7], Theorems 4.2.10 and 4.2.20 cover all the systems displayed in the
Table 3.3 at the end of Sect. 3.3 of Chap. 3.

4.3 The Problem of Algebraizability of LFIs

The aim of this section is to briefly analyze the question of algebraizability of LFIs
in the sense of Blok and Pigozzi (see [1]). Thus, some LFIs extending the ones
previously studied will be analyzed, showing that, in order to obtain an algebraizable
extension of mbC, it is necessary to arrive at a 3-valued logic. Seeing that, several
3-valued LFIs, all of them algebraizable in the sense of Blok and Pigozzi, will be
studied in Sect. 4.4. The question of algebraizability of LFIs will be analyzed again
in Chap. 6, proving (by using swap structures) that mbC and several of its extensions
can be characterized by a single finite non-deterministic matrix, despite not being
characterizable by a single finite matrix (as it was proved in Sect. 4.2).
Algebraic logic, as the study of the relationship between logic and algebraic struc-
tures, begins in the 19th century from the investigations of Boole [8, 9], De Morgan
[10], Jevons [11], Peirce [1214] and Schrder [15]. Their work was based on the
similarities between logical equivalence, on the one hand, and equality, on the other.
The investigations of Boole led to the development of the Boolean algebras, while
130 4 Matrices and Algebraizability

the studies of the other authors cited above were afterwards generalized, originating
the modern theory of relation algebras. An interesting survey on these pioneering
works can be found in [16].
The relationship between logic and algebra from the contemporary perspective
comes back to the ideas of Lindenbaum and Tarski, in which the formulas of a given
logic are interpreted by means of algebras with operations associated to the logical
connectives. In [1], Blok and Pigozzi proposed a generalization of the original alge-
braization techniques in an effort to encompass a wider range of logics. After this,
several generalizations of Blok and Pigozzis method were proposed in the litera-
ture. For a more complete background, we refer interested readers to the excellent
references [1720].
The notion of algebraizability in the sense of Blok and Pigozzi admits several
equivalent characterizations. For our purposes, the following characterization will
suffice:

Definition 4.3.1 (Blok and Pigozzi algebraizability, Theorem 4.7 in [1]) Let  be
a propositional signature, and let L be a standard propositional logic defined over
the language L , with a consequence relation L (recall Definition 2.1.4). Then
L is algebraizable in the sense of Blok and Pigozzi if there exists a nonempty set
( p1 , p2 ) L of formulas depending on variables p1 and p2 , and a nonempty
set E( p1 ) L L of pairs of formulas depending on variable p1 satisfying the
following properties:
(BP1) L ( p1 , p1 ), for every ( p1 , p2 ) ( p1 , p2 );
(BP2) ( p1 , p2 ) L ( p2 , p1 ), for every ( p1 , p2 ) ( p1 , p2 );
(BP3) ( p1 , p2 ), ( p2 , p3 ) L ( p1 , p3 ), for every ( p1 , p2 ) ( p1 , p2 );
(BP4) ( p1 , pn+1 ), . . . , ( pn , p2n ) L (#( p1 , . . . , pn ), #( pn+1 , . . . , p2n )), for
every ( p1 , p2 ) ( p1 , p2 ), every n-ary connective # of  and every n 1;
(BP5) p1 L ( ( p1 ), ( p1 )), for every ( p1 , p2 ) ( p1 , p2 ) and every
 ( p1 ),( p1 ) E( p1 ); 
(BP6) ( ( p1 ), ( p1 )) : ( p1 , p2 ) ( p1 , p2 ),  ( p1 ), ( p1 ) E( p1 ) L p1 .

The sets ( p1 , p2 ) and E( p1 ) are called system of equivalence formulas and

defining equations, respectively. 

Definition 4.3.2 Let  be a propositional signature, and let L L be a

relation defined over the algebra of formulas L . The relation is a congruence in
L if it satisfies the following properties:
(i) for every L (reflexivity).
(ii) implies for every , L (symmetry).
(iii) and implies for every , , L (transitivity).
(iv) Given i and i in L (for 1 i n) such that 1 1 , , n n , then
#(1 , . . . , n ) #(1 , . . . , n ) for every n-ary connective # of  and every
n 1.
4.3 The Problem of Algebraizability of LFIs 131
 
A congruence in L is trivial if either = (, ) : L or =
L L . 

Definition 4.3.3 Let L be a standard logic defined over the language L .

(i) A congruence in L is compatible with a theory  L if it satisfies the
following:
and  L implies that  L .

(ii) A congruence in L is a logical congruence in L if is compatible with every

theory . Equivalently, is a logical congruence in L if, for every and ,

implies that L and L .

The following formula was assumed in [21] as defining a non-trivial congruence

in the algebra of formulas L0 which is compatible with every theory of the logic
C1+ (see Remark 3.8.2):

( p1 , p2 ) = p1 p2 ( p1 p2 ) ( p2 p1 )
def

recalling that, in the language L0 of C1+ , is an abbreviation and not a primi-
tive connective. The relation  is then defined as follows:  iff C + (, ).
Unfortunately, despite having interesting properties,  is not a congruence, as it
is not reflexive. Moreover, it is imediate to see that  iff C + . Thus,
the unique formulas for which the proposed relation is reflexive are the well-
behaved ones, and so this relation is the usual Lindenbaum-Tarski congruence over
the classical fragment inside C1+ . The properties of will be proved below for
Cio (see [7, 22]), a logic defined over  which is somewhat weaker than Cilo, the
version of C1+ over .

Definition 4.3.4 The logic Cio, defined over the signature , is the system obtained
from Cilo (recall Remark 3.8.2) by removing axiom (cl). Equivalently, Cio is the
system obtained from mbC by adding axiom schemas (ci), (cf) and (co1 )(co3 ). 

Clearly, the logic Cio is a strong LFI.

Definition 4.3.5 Let ( p1 , p2 ) = p1 p2 ( p1 p2 ) ( p2 p1 ) where

def

is the consistency operator of . Consider the following relation over the set L of
formulas:
 iff Cio (, ) .


132 4 Matrices and Algebraizability

As usual, (, ) denotes the formula () (or, using a common notation,

[ p1 /, p2 /]), where is a substitution such that ( p1 ) = and ( p2 ) = and
is the unique endomorphism in the algebra L extending .

Proposition 4.3.6 Let and be formulas in L . Then, for every valuation v for
Cio,
v((, )) = 1 iff v() = v() = 0 or v() = v() = 0 .

Proof It is easy to check by building a truth-table in Cio for (, ). 

Proposition 4.3.7 The relation , defined above, satisfies properties (ii)(iv) of

Definition 4.3.2. However,  is not reflexive and so it is not a logical congruence in
Cio. The same holds for Cilo.

Proof Items (ii) and (iii) are obvious. In order to prove (iv), we only need to prove
the following:

(1) if  , then  ; and

(2) if  , then  .

This is a consequence of the fact that the binary connectives of Cio are classical, that
is, they are truth-functional with respect to their valuation semantics. This feature
guarantees the preservation of equivalence through those connectives.
In order to prove (1), let , L such that  , and let v be a valuation
for Cio. By soundness of Cio w.r.t. valuations, v((, )) = 1 and, thus, either
v() = v() = 0 or v() = v() = 0 by Proposition 4.3.6. If v() = v() = 0,
then v() = v() = 0 by clause (valCf ) and v((, )) = 1 by Proposi-
tion 4.3.6. On the other hand, if v() = v() = 0, then v((, )) = 1, again
using Proposition 4.3.6. Then (, ) is valid and  by completeness of
Cio. This proves (1).
Now, assume that  , and let v be a valuation for Cio. As proved above,
either v() = v() = 0 or v() = v() = 0. Therefore, by (vCi) it follows that
v() = v() = 0. Hence v((, )) = 1 by Proposition 4.3.6. This shows
that (, ) is valid, thus  by completeness of Cio.
Finally, in order to see that  is not reflexive, take a propositional variable p and
a valuation v for Cio such that v( p) = v( p) = 1. Then v(( p, p)) = 0 by Propo-
sition 4.3.6, and then p  p, by completeness of Cio w.r.t. valuations. Therefore, 
is not reflexive and consequently not a congruence for Cio. The same proof holds
for Cilo. 

It is possible, however, to define a non-trivial logical congruence in an extension

of Cio, as we shall see below.

Definition 4.3.8 (Retroprogagation of consistency, [22]) Let L be an extension of

mbC defined over the signature . The strong retropropagation of consistency of
each binary connective of  in L is given by the following axiom schemas:
4.3 The Problem of Algebraizability of LFIs 133

( ) ( ) (cr1 )
( ) ( ) (cr2 )
( ) ( ) (cr3 )

The axioms above are characterized by a valuation semantics satisfying the fol-
lowing clauses:
(vCr1 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCr2 ) v(( )) = 1 = v() = 1 or v() = 1.
(vCr3 ) v(( )) = 1 = v() = 1 or v() = 1.

Definition 4.3.9 ([22]) The logic Ciore, defined over the signature , is the sys-
tem obtained from Cio (recall Definition 4.3.4) by adding axiom schemas (ce) and
(cr1 )(cr3 ). 

At this point of the book, the proof of the following results is quite easy:
Theorem 4.3.10 The logic Ciore is sound and complete for Cio-valuations which
satisfy additionally clauses (vCe) and (vCr1 )(vCr3 ).

Corollary 4.3.11 The logic Ciore is a (strong) LFI.

Additionally:
Proposition 4.3.12 Let v be a Cio-valuation. Then v satisfies clauses (vCr1 )-(vCr3 )
if and only if it also satisfies the following clauses:
(vCr11 ) v( ) = 1 and v( ) = 1 = v(( )) = 1.
(vCr12 ) v( ) = 1 and v( ) = 1 = v(( )) = 1.
(vCr13 ) v( ) = 1 and v( ) = 1 = v(( )) = 1.

Proof Let v be a Cio-valuation. The only situation in which v() = v() = 0 is

when v() = v() = v() = v() = 1. But then, v(#) = 1 for # {, , }.
In this way, retropropagation of consistency holds for the binary connectives
iff v((#)) = 1 for # {, , }. Indeed, if v((#)) = 0 for some #
{, , }, then v((#)) = 1 while v( ) = 0, violating the retropopaga-
tion of consistency for #, as the following table shows:

# (#) (#)
1 1 1 1 0 0 1 1 0
0 1 =

The converse is obvious. 

134 4 Matrices and Algebraizability

Corollary 4.3.13 The logic Ciore is characterized as the extension of Cio obtained
by adding axiom schema (ce) and the following axiom schemas:

( ) ( ) ( ) (cr11 )
( ) ( ) ( ) (cr12 )
( ) ( ) ( ) (cr13 )

As observed for Cilo in Remark 3.8.2, the logic Ciore satisfies all the requirements
of Definition 3.8.1. The next step is to define a non-trivial congruence for Ciore.

Definition 4.3.14 Let ( p1 , p2 ) be the following formula of L :

( p1 , p2 ) = ( p1 p2 ) ( p2 p1 ) ( p1 p2 ) ( p2 p1 ) .
def

Proposition 4.3.15 Let and be formulas in L . Then, for every valuation v for
Ciore,
v((, )) = 1 iff v() = v() and v() = v() .

Proof It is immediate by definition of (, ) and the semantics of and . 

 
Proposition 4.3.16 Let ( p1 , p2 ) = ( p1 , p2 ) . Then ( p1 , p2 ) satisfies condi-
tions (BP1)(BP4) of Definition 4.3.1.

Proof It is quite easy to see that conditions (BP1)(BP3) of Definition 4.3.1 are
satisfied by ( p1 , p2 ). We prove below that condition (BP4) is also satisfied.
Concerning and , let v be a Ciore-valuation such that v(( p1 , p2 )) = 1. There-
fore, v( p1 ) = v( p2 ) and v( p1 ) = v( p2 ), by Proposition 4.3.15.
Case 1 (). As v( p1 ) = v( p2 ) = 1 then v(( p1 , p2 )) = 1, by
Proposition 4.3.15. By completeness of Ciore, ( p1 , p2 ) Ciore ( p1 , p2 ).
Case 2 (). We have the following cases:
(1.1) v( p1 ) = v( p2 ) = 0. Then v( p1 ) = v( p2 ) = 1 and so v( p1 ) = v( p2 ) = 1.
From here, v( p1 ) = v( p2 ) = 0 by propagation of consistency. Therefore,
v(( p1 , p2 )) = 1.
(1.2) v( p1 ) = v( p2 ) = 1. There are two subcases:
(1.2.1) v( p1 ) = 1. Then v( p1 ) = v( p2 ) = 0 and hence v( p2 ) = 1. From here,
v( p1 ) = v( p2 ) = 1. Thus, v( p1 ) = 0 iff v( p1 ) = 1, and v( p2 ) = 0 iff
v( p2 ) = 1. Since v( p1 ) = v( p2 ) = 1 then v( p1 ) = v( p2 ) = 1, by (vCe).
This shows that v( p1 ) = v( p2 ) = 0 and hence v(( p1 , p2 )) = 1.
(1.2.2) v( p1 ) = 0. Then v( p1 ) = v( p2 ) = 1 and v( p2 ) = 0. Additionally,
v( p1 ) = v( p2 ) = 1 by propagation of consistency. Thus, v(( p1 ,
p2 )) = 1.
In all the cases above, it follows that v(( p1 , p2 )) = 1. Thus, ( p1 , p2 )
( p1 , p2 ) by completeness of Ciore.
4.3 The Problem of Algebraizability of LFIs 135

Case 2 (# {, , }). By using the axioms and rules of CPL+ it follows that
   
( p1 , p3 ), ( p2 , p4 ) Ciore ( p1 # p2 ) ( p3 # p4 ) ( p3 # p4 ) ( p1 # p2 ) .
 
Now, let  = p1 p3 , p2 p4 , ( p1 # p2 ) . Then , p1 Ciore p3 .
However, p3 Ciore ( p3 # p4 ) by strong propagation of consistency, and then
, p1 Ciore ( p3 # p4 ). Analogously, it is proven that , p2 Ciore ( p3 # p4 ), so
, p1 p2 Ciore ( p3 # p4 ). Given that ( p1 # p2 ) Ciore p1 p2 by retroprop-
agation of consistency, then  Ciore ( p3 # p4 ). By the deduction meta-theorem, it
follows that p1 p3 , p2 p4 Ciore ( p1 # p2 ) ( p3 # p4 ). Mutatis mutan-
dis, it is proven that p3 p1 , p4 p2 Ciore ( p3 # p4 ) ( p1 # p2 ). By the
properties of , it follows that
   
( p1 , p3 ), ( p2 , p4 ) Ciore ( p1 # p2 ) ( p3 # p4 ) ( p3 # p4 ) ( p1 # p2 ) .

From this clause, it is obtained that

( p1 , p3 ), ( p2 , p4 ) Ciore (( p1 # p2 ), ( p3 # p4 )) .


Corollary 4.3.17 Consider the following relation over the set L of formulas:

 iff Ciore (, ) .

Then  is a congruence in the algebra of formulas L , that is, it satisfies the con-
ditions of Definition 4.3.2. Moreover,  is a logical congruence in Ciore (recall
Definition 4.3.3).
Proof Immediate from the previous proposition and the corresponding
definitions. 
The last result shows that Ciore, unlike C1 , admits a non-trivial congruence,
suggesting the possibility of defining some kind of algebraization within the setting
of Blok and Pigozzi (see Definition 4.3.1) or in the more general framework of Font
and Jansana (see [19]). In fact, this is the case:
Theorem 4.3.18 The logic Ciore is algebraizable in the sense of Blok and Pigozzi
(see Definition 4.3.1) where the system of equivalence formulas  is ( p1 , p2 ) =
{( p1 , p2 )} and the system of defining equations is E( p1 ) =  p1 , p1 p1  .
Proof The system ( p1 , p2 ) satisfies conditions (BP1)-(BP4), by Proposition 4.3.16.
Now, observe that Ciore ( p1 ( p1 p1 )) (( p1 p1 ) p1 ) by strong
propagation and retropropagation of consistency. That being so, ( p1 , p1 p1 )
is equivalent in Ciore to ( p1 ( p1 p1 )) (( p1 p1 ) p1 ). But the lat-
ter is equivalent in Ciore to p1 . This shows that conditions (BP5)(BP6) are also
satisfied. 
136 4 Matrices and Algebraizability

This positive result should not be so surprising: in the next section (see Theo-
rem 4.4.29 in Sect. 4.4.6) it will be proved that Ciore is strong enough to be char-
acterized by a 3-valued matrix, as mentioned in [22]. Thus, Ciore lies outside the
scope of the Dugundji-like results stated in Sect. 4.2.

4.4 Some 3-Valued LFIs

In this section, some 3-valued LFIs will be analyzed, obtaining in some case examples
of systems which are algebraizable in the sense of Blok and Pigozzi. The systems
will be presented in chronological order of appearance.

4.4.1 Halldns Logic of Nonsense (1949)

To the best of our knowledge, the first paraconsistent logic presented by means of
3-valued logic matrices was put forth by Halldn, in his monograph The Logic of
Nonsense (see [23]) in 1949.
To be rigorous, Halldn was not primarily interested in defining a paraconsistent
logic (as Jaskowski was at the same time, see [24]), but more with dealing with
vagueness or, more specifically, with meaningless propositions from a formal point
of view. Thus, for Halldn paraconsistency was, so to speak, a by-product of his
investigations, not a primary goal. In Halldns terms a proposition is meaningful
when it is either true or false, and is meaningless otherwise. According to him, This
proposition is false, The man with a hundred hairs is bald and The man with a
hundred hairs is not bald are meaningless. This is in agreement with his proposal for
a 3-valued matrix logic with 1 (true), 0 (false) and 21 (meaningless) as truth-values.
The basic operators are conjunction and negation , which coincide with classical
operators over the classical truth-values 0 and 1. Accordingly, if every component of
a complex proposition is classical so is the truth-value of that sentence. However, if
any component is meaningless so is the complex proposition. Since he is interested in
having tautologies (in fact, he includes all the classical tautologies over the signature
{, }), he is forced to assume that the meaningless truth-value 21 is designated;
otherwise, no sentence would be valid, given that 21 propagates to any complex
formula. Finally, in order to improve the expressive power of the logic, Halldn also
includes an unary operator to represent meaningfulness. Formally:

Definition 4.4.1 (Halldns nonsense logic [23]) Let M H = M, D be the 3-

valued matrix over  H = {, , } with domain M = {1, 21 , 0} and set of desig-
nated values D = {1, 21 }, where the truth-tables associated to each connective are as
follows (by abuse of notation, the symbol of the operator interpreting a connective
of  H will coincide with the symbol of that connective):
4.4 Some 3-Valued LFIs 137

The logic associated with this matrix will be called H3. 

Let H3 be the consequence relation of H3. Thus:  H3 iff, for every  H -
homomorphism h : L H M, if h[] D, then h() D. As stipulated at the
end of Sect. 4.1, such homomorphisms are called valuations over H3.
Since the algebra L H is freely generated by the set V ar of propositional variables,
any mapping h 0 : V ar M will generate a unique valuation h for H3 extending h 0 .
As a consequence of the definitions above, the logic H3 is paraconsistent. More-
over, it is a strong LFI (recall Definition 2.1.9):
Theorem 4.4.2 Let ( p) = { p} for a propositional variable p. Then the logic H3
is a strong LFI.

Proof Let p and q be two distinct propositional variables and let h be a valuation for
H3 such that h( p) = 21 and h(q) = 0. Then h[{ p, p}] D but h(q) / D, showing
that p, p H3 q. Thus, clause (i.a) of Definition 2.1.9 is satisfied. Now consider-
ing a valuation h  such that h  ( p) = 1 and h  (q) = 0, we show that clause (i.b) of
Definition 2.1.9 is satisfied. If h  is another valuation such that h  ( p) = h  (q) = 0
then we see that clause (i.c) of Definition 2.1.9 is also satisfied. Finally, it is clear
that for no valuation h and formula it is the case that h() D, h() D and
h() D simultaneously. Therefore clause (ii) of Definition 2.1.9 is satisfied. 

It is an interesting fact that the idea of a consistency operator was already present
in such an early approach. Despite the presence of , the expressive power of matrix
M H is still weak: it is well-known that the operators of M H cannot express the
unary operator such that (1) = 1 and (x) = 0 otherwise (see, for instance,
Theorem 7.3.8 in [25]). Moreover, the Halldn conjunction  does not satisfy all
the desirable properties: in general,  H3 and  H3 . This can be
proved by taking and as two distinct propositional variables and a valuation h
such that h() = 0 and h() = 21 (analogously for the second statement). However,
H3 is strong enough to recover classical logic CPL (defined over the signature
  = {, }) by means of the following DAT:
Proposition 4.4.3 (DAT between CPL and H3, [25]) Let t : L  L H be the
mapping which replaces with and with . Then the following holds: for every
finite  {} L  ,

 CPL iff p1 , . . . , pn , t[] H3 t (),

where { p1 , . . . , pn } is the set of propositional variables occurring in  but not in .

138 4 Matrices and Algebraizability

Proof See Theorem 7.3.22 in [25]. 

With respect to the propagation and retropropagation of consistency, the following

properties can be easily proved (see Proposition 7.3.19 in [25]).

Proposition 4.4.4 The following holds for every and in H3:

1. H3 , but H3 .
2. H3 , and H3 .
3. , H3 (  ); (  ) H3 and (  ) H3 .

4.4.2 Segerbergs Logic of Nonsense (1965)

With the aim of overcoming some of the limitations of H3 (for instance, its inability
to define the operator , as noted above), K. Segerberg proposed an interesting variant
of it in 1965, using the operators of the closely related 3-valued nonsense logic B3
of A. Bochvar (introduced in 1938, see [26]) and changing the set of designated
values from {1} to D = {1, 21 }. This move changes the features of the consequence
relation, and so instead of being paracomplete (as in the case of the original Bochvars
logic B3), the resulting logic is paraconsistent (and, in fact, an LFI).1 In formal terms:

Definition 4.4.5 (Segerbergs nonsense logic [27]) Let M S = M, D be the 3-

valued matrix over  S = {, , } with domain M = {1, 21 , 0} and set of designated
values D = {1, 21 }, such that the truth-tables associated with each connective are as
follows:

The logic associated with this matrix will be called S3. 

It is easy to see that M S is obtained from M H by replacing the operator

with . Observe that, for every x M, x = (x  x) and so H3 is a proper
fragment of S3 (since, as observed above, cannot be defined in M H ).
Corollary 4.4.6 The logic S3 is a strong LFI with consistency operator defined
as above.

1 Despite this, and as in the case of Halldn, it should be clear that paraconsistency was not the main

motivation for Segerbergs nonsense logic.

4.4 Some 3-Valued LFIs 139

Additionally, a strong negation can be defined in M S as x = x, as well

def

as a so-called stabilizer operator x = x for every x M. Furthermore, a

def

disjunction and two deductive implication operators can be defined accordingly:

x  y = (x  y), x y = x  y and x  y =  (x y). The truth-
def def def

tables of these operators are as follows:

Observe that x = x  x and x = x. By using the stabilizer operator

, it is possible to define a conjunction  and a disjunction  which are bet-
ter behaved than the original  and  (see Proposition 4.4.7 below). From here,
an equivalence operator can be obtained as follows: x  y =  x  y;
def

x  y = x  y; and x y = (x y)  (y x) = (x  y)  (y  x).

def def

The truth-tables for these operators are displayed below.

Proposition 4.4.7 Let h be a valuation for S3. Then:

(i) h( p1 p2 ) D iff h( p1 ) = 0 or h( p2 ) D (idem for );
(ii) h( p1  p2 ) D iff h( p1 ) D and h( p2 ) D;
(iii) h( p1  p2 ) D iff h( p1 ) D or h( p2 ) D;
(iv) h( p1 p2 ) D iff either h( p1 ) D and h( p2 ) D, or h( p1 ) = h( p2 ) = 0.

Proof Immediate fom the truth-tables. 

The next step is to show that S3 is algebraizable in the sense of Blok and Pigozzi
(recall Definition 4.3.1). The systems S ( p1 , p2 ) and E S ( p1 ) for S3 will be quite
similar to those found for Ciore.
140 4 Matrices and Algebraizability

Definition 4.4.8 Let ( p1 , p2 ) be the following formula of LS :

( p1 , p2 ) = ( p1 p2 )  ( p1 p2 ) .
def


Proposition 4.4.9 Let and be formulas in LS . Then for every valuation h for
S3 it holds that:
1. h((, )) D iff we have:
1.1 either h() D and h() D, or h() = h() = 0; and
1.2 h() = h().
Thus, h((, )) D iff h() = h().
2. h( ) D.
3. h() = h(( )).
4. h((, )) D iff h() D.
Proof 1. h((, )) D iff, in accordance with Proposition 4.4.7(ii), h( )
D and h( ) D. By item (iv) of the same proposition, h( ) D is
equivalent to 1.1, while h( ) D iff either h() D and h() D, or
h() = h() = 0. By the definition of , h( ) D is equivalent to 1.2.
Now, suppose that h((, )) D. If h() = 1 and h() = 21 , then h() = 1 and
h() = 0, violating clause 1.2. Analogously, it is impossible to have h() = 21 and
h() = 1. This shows that h() = h(). The converse is obvious, in light of 1.1
and 1.2.
2. Trivial.
3. If h() = 1 then h() {0, 1} and so = 1; hence, h(( )) = 1. If
h() = 0 then h() = 21 and so = 21 ; therefore, h(( )) = 0.
4. By item 1, h((, )) D implies that h() = h( ). So, by item 2,
h() D. Conversely, suppose that h() D. By item 2, h( ) D and, by
item 3, h() = h(( )). As argued in the proof of item 1, h() = h( )
and then h((, )) D. 
Theorem 4.4.10 The logic S3 is algebraizable in the sense of Blok and Pigozzi
(see Definition 4.3.1) with a system of equivalence formulas givenby S ( p1 , p2 ) = 
{( p1 , p2 )}, and a system of defining equations given by E S ( p1 ) =  p1 , p1 p1  .
Proof It is easy to prove that the system S ( p1 , p2 ) satisfies conditions (BP1)
(BP4) by item 1 of Proposition 4.4.9. By item 4 of the same proposition, conditions
(BP5)(BP6) follow easily. 

4.4.3 da Costa and DOttavianos Logic J3 (1970)

In his pioneering paper A propositional calculus for inconsistent deductive systems

from 1948, Jaskowski establishes the problem and criterion of paraconsistency using
three conditions:
4.4 Some 3-Valued LFIs 141

Accordingly, the problem of the logic of inconsistent systems is formulated here in the
following manner: the task is to find a system of the sentential calculus which: (1) when
applied to the inconsistent systems would not always entail their overfilling, (2) would be rich
enough to enable practical inference, (3) would have an intuitive justification. Obviously,
these conditions do not univocally determine the solution, since they may be satisfied in
varying degrees, the satisfaction of condition 3 being rather difficult to appraise objectively.
(Jaskowski [24])

In the same paper, Jaskowski proposes the so called discussive (or discursive)
logic as a possible solution to the problem posed by his teacher J. ukasiewicz.
In 1970, DOttaviano and da Costa proposed a 3-valued modal logic called J3
as an alternative solution to Jaskowskis problem (see [28]). They also presented a
first-order version of this logic. Afterwards, in her Ph.D. thesis [29] and in a series
of papers DOttaviano develops several results concerning the model theory of first-
order paraconsistent logic J3 (see [3032]). Together with Dunns paper [33], this
constitutes a pioneering work on 3-valued paraconsistent first-order model theory,
which was subsequently rediscovered in the context of inconsistent databases and
the theory of quasi-truth (see Sect. 7.9 of Chap. 7 and Sect. 9.3 of Chap. 9). The
basic language of J3 is formed by a disjunction, a paraconsistent negation and a
possibility operator. Formally:
Definition 4.4.11 (DOttaviano and da Costas logic J3, [28]) Let M J = M, D
be the 3-valued matrix over  S = {, , } with domain M = {1, 21 , 0} and set of
designated values D = {1, 21 }, with the following truth-tables associated with each
connective:

The logic associated with this matrix is called J3. 

Observe that the possibility operator coincides with the stabilizer operator 
defined in Segerbergs logic.
Remark 4.4.12 (The ubiquity of J3) The logic J3 has independently been re-
introduced by several authors with different signatures and different motivations.
With the aim of dealing with inconsistent databases, Carnielli, Marcos and de Amo
introduced in 2000 a 3-valued LFI called LFI1, which they have proven to be func-
tionally equivalent to J3 (see [34, 35]). That is, the primitive operators of both
matrices are interdefinable, where the signature of LFI1 includes an inconsistency
operator instead of . This logic will be briefly analyzed in Sect. 4.4.7. In turn, in
2004, Batens and De Clercq introduced, by means of a Hilbert calculus, a first-order
paraconsistent logic called CLuNs (see [36]). The propositional fragment of this
logic coincides with v , a logic introduced in 1960 by Schtte in the context of
Proof Theory (see [37]). The properties of the logic v were studied by Batens in
142 4 Matrices and Algebraizability

1980 (see [38]). Interesting enough, the logic v , that is, the propositional frag-
ment of CLuNs, is sound and complete with respect to the matrix M J of J3.2 In
[39, 40], the notion of quasi-truth introduced in 1986 by Mikenberg et al. (see [41])
was generalized by Coniglio and Silvestrini. This generalization is axiomatized by a
first-order paraconsistent logic called LPT1, whose propositional fragment is char-
acterized by a 3-valued matrix logic called MPT. As proven in [40], the logic MPT
coincides (up to language) with J3, thus constituting a new variant of J3 which again
arises independently in a different context and from a different motivation. Conse-
quently, LPT1 coincides (up to language) with the first-order version of J3 studied
by DOttaviano and also with the semantics for first-order LFI1 proposed in [34]
for inconsistent databases.3 However, the semantics of triples proposed in [40] for
LPT1 is of a different nature than those of the other two approaches. This point will
be revisited in Sect. 4.4.7, in Chap. 7 and in Sect. 9.3 of Chap. 9. In 2015, yet another
formulation of this logic was independently found by Lwe and Tarafder in [42],
called PS3 . This logic was employed in models of a certain paraconsistent set theory
(see Sect. 8.5 in Chap. 8).

Besides the ubiquity of J3 in the universe of paraconsistency, an interesting con-

nection between the matrix M J of J3 and that of ukasiewiczs 3-valued logic 3
can be established.

Definition 4.4.13 (ukasiewiczs 3-valued logic 3, [43]) Recall the matrix M L =

M, D L  over  L = { L , } with domain M = {1, 21 , 0} and set of designated
values D L = {1} which defines the well-known ukasiewiczs 3-valued logic 3,
where the truth-tables associated with these connectives are as follows:

It is well known (see, for instance, [29]) that the operators of J3 can be defined from
those of 3 as follows: x y = (x L y) L y and x = x L x. But the con-
def

verse is also true: from the operators of J3 it is possible to define the implication L of
3 as follows: x L y = (x y) ( y x), where x y = (x y)
def def

(see [29]). Thus, 3 and J3 are functionally equivalent (we will return to the close
relationship between J3 and 3 in Remark 4.4.16). Observe that the truth-table for
is as follows:

2 Itis worth noting that the logic v was exclusively presented in [37] by means of a Hilbert style
calculus, and not as a 3-valued matrix logic.
3 The definition of quantifiers in LPT1 differs from [39] to [40], the latter coinciding with that of [29,

34] (see [40]).

4.4 Some 3-Valued LFIs 143

The strong negation , the consistency operator , Bochvars operator and a new
unary operator , given by following truth-tables

are definable by the matrix M L of 3 (and, therefore, by the matrix M J of J3) as

follows: x = x, x = x x, x = x x and x = x x,
def def def def

where , and are defined as above. Then:

Proposition 4.4.14 For every x, y V , x  y = (x y) (x y). Then

def

every truth-funcion definable in M S can be defined by M L (and, accordingly, by

M J ). From here, Bochvars logic B3 can be seen as a sublogic of 3, while H3 and
S3 are sublogics of J3 after considering suitable signatures. The converse is not true:
there exist truth-tables which are definable in M J but cannot be defined in M S .

Proof The first part is immediate from the observations above. In order to see that
M J is strictly stronger than M S , see [44]. 

It is easy to prove that J3 is a strong LFI. An implication and an equiva-

lence operator can be defined naturally in J3 as follows: x J y = x y and
def

x J y = (x J y) (y J x). The truth-tables for these operators are dis-

def

played below.

Observe that J is a deductive implication: , J3 iff  J3 J .

Moreover, by considering the natural order 0 21 1 in M, then corresponds to
the infimum (and so is the supremum) and J is the residuum of : z x y
iff z x J y, for every x, y, z M.
It is easy to see that items (i)(iv) of Proposition 4.4.7 hold for J , , and J
in the place of , ,  and , respectively. From here, by defining
144 4 Matrices and Algebraizability

J ( p1 , p2 ) = ( p1 J p2 ) ( p1 J p2 )
def

the content of Proposition 4.4.9 can be adapted for the respective operators of J3.
With this information, the algebraizability of J3 in the sense of Blok and Pigozzi
follows easily:
Theorem 4.4.15 The logic J3 is algebraizable in the sense of Blok and Pigozzi
(see Definition 4.3.1) with a system of equivalence formulas given by
 J ( p1 , p2 ) = { J( p1 , p2 )} and a system of defining equations given by E J ( p1 ) =

 p1 , p1 J p1  .
Remark 4.4.16 Consider again the signature  L = { L , } of 3 and assume that
J3 is presented in that signature (given the fact that the matrices M J and M L are
functionally equivalent). As it was proved in Example 4.1.2 of [45], the mapping
t : LL LL given by t () = (where = L ) is a conservative
def

translation from J3 into 3 that is:

 J3 iff t[] 3 t () .

Moreover, for every LL : 3 and 3 , on the one hand, and

J3 and J3 on the other. From here, it is proved in [45] that, despite
being different, both logics 3 and J3, when defined over the same signature  L , are
algebraizable (in the sense of Blok and Pigozzi) with the same equivalent algebraic
semantics, namely the variety of 3-valued Wajsberg algebras (a reduction of the 3-
element MV-algebra). This constitutes an alternative algebraization of J3 to the one
put forth in Theorem 4.4.15, which is defined in another language.

4.4.4 Settes Logic P1 (1973)

In 1963, together with his famous hierarchy of systems Cn (for n 1) N. da Costa

proposed requisites that these systems (or, in general, a calculus for dealing with
non-trivial inconsistent theories) should satisfy:
I) In these calculi the principle of contradiction, (A&A), must not be a valid schema;
II) From two contradictory formulas, A and A, it will not in general be possible to deduce
an arbitrary formula B; III) it must be simple to extend Cn , 1 n , to corresponding
predicate calculi (with or without equality) of first-order; IV) Cn , 1 n , must contain
the most part of the schemata and rules of C0 [the classical propositional calculus],4 which
do not interfere with the first conditions.
(da Costa, [46])

In 1973, in [47], A. Sette introduced a 3-valued paraconsistent logic called P1,

which aimed to be the simplest possible way of satisfying da Costas aforementioned
requirements. Indeed, principles I and II are clearly satisfied. Concerning IV, the logic

4 Our clarification.
4.4 Some 3-Valued LFIs 145

P1 is maximal with respect to propositional classical logic in the following sense:

by adding to P1 as an axiom schema any classical tautology that it is not derivable
in P1, the resulting logic collapses to classical logic. This feature certainly is not
satisfied by da Costas calculi Cn . It is also worth noting that Settes logic extends
C1 , so all the calculi Cn , for n 1. With respect to III, a natural extension of P1 to
first-order languages will be proposed in Chap. 7.
Definition 4.4.17 (Settes logic P1, [47]) Let M P = M, D be the 3-valued matrix
over  P = {, P } with domain M = {1, 21 , 0} and set of designated values D =
{1, 21 }, where the truth-tables associated with each connective are defined as follows:

The logic P1 is the one associated with the matrix M P . 

A remarkable feature of P1 is that paraconsistency only happens on the atomic
level: any complex formula is classical in the sense that it can only take the truth-
values 0 or 1 (this is obvious from the truth-tables).
Observe that the primitive implication of P1 is the same as the one defined
in Sect. 4.4.2 (using the symbol ) with the matrix M S of Segerbergs logic S3,
while P x = x for every x. This shows that P1 is a sublogic of S3 (and also of
J3 in virtue of Proposition 4.4.14), provided that suitable signatures are considered.
The inclusion is strict: M P is unable to define an operator (different to the identity
mapping) returning 21 as an output; for instance, the negation of S3 cannot be
defined with matrix M P .
The strong negation as well as the conjunction  and the disjunction  defined
in Sect. 4.4.2 for S3 were also accounted by Sette, being defined in matrix M P as
follows: x = P ( P x x), x  y = P ( P x y) and x  y = x y.
def def def

From here, the equivalence operator can be defined as above. Additionally, the
consistency operator can be defined as x = x P P x in matrix M P . Thus,
def

the logic P1 could be seen as defined over the signature  (and, in fact, this will be
the case, see Definition 4.4.19 below). Clearly, it is a strong LFI.
Given that P1 is able to define all those operators, it follows (as in the case of
Segerbergs logic S3, see Theorem 4.4.10) that it is algebraizable in the sense of Blok
and Pigozzi:
Theorem 4.4.18 The logic P1 is algebraizable in the sense of Blok and Pigozzi with
a system of equivalence formulas P ( p1 , p2 ) = {( p1 , p2 )} (where the formula
 now expressedin the signature  P ) and a system of
introduced in Definition 4.4.8 is
defining equations E P ( p1 ) =  p1 , p1 p1  .
The algebraizability of P1 in the sense of Blok and Pigozzi was firstly established
by Lewin et al. [48] in 1990. They propose the following systems for algebraization:
146 4 Matrices and Algebraizability

P ( p1 , p2 ) = { p1 p2 , p2 p1 , P p1 P p2 , P p2 P p1 }, and
 
E P ( p1 ) = ( p1 p1 ) p1 , p1 p1  .

In [7] (p. 77, comment after Theorem 127), the logic P1 was axiomatized as an
LFI defined over the signature  as follows:
Definition 4.4.19 (P1 as an LFI) Let P1 be the Hilbert calculus over  obtained
from Ci (that is, mbC plus axioms (ci) and (cf)) by adding the following axiom
schemas:

(cw)
( ) (cv1 )
( ) (cv2 )
( ) (cv3 )


Observe that P1 , since P1 extends mbCci. Valuations for P1 will be defined
as usual, but this time some particularities of P1 will be taken into consideration.
In view of these circumstances, the resulting semantics will be much simpler than
expected.

Definition 4.4.20 A function v : L {0, 1} is a valuation for P1 , or a P1 -

valuation, if it satisfies clauses (vAnd), (vOr) and (vImp) from Definition 2.2.1,
plus the following clauses:
(vNeg)a v( p) = 0 = v( p) = 1, if p is atomic
(vNeg)c v() = 0 v() = 1, if is complex
(vCon)a v( p) = 1 v( p) = 0 or v( p) = 0, if p is atomic
(vCon)c v() = 1, if is complex. 

Observe that any P1 -valuation v satisfies clause (vCf ) from Sect. 3.5. In fact,
if p V ar then v( p) = 1 implies v( p) = 0, by (vNeg)c , and so v( p) = 1, by
(vNeg)a . On the other hand, if is complex then v() = 1 implies v() = 0, by
(vNeg)c , and so v() = 1, by (vNeg)c . Clearly, P1 -valuations satisfy clause (vCon)
from Definition 2.2.1. Additionally, clause (vCi) (see Definition 3.1.7) is satisfied by
P1 -valuations. In fact, if is complex then v() = 0 and so (vCi) holds trivially.
Otherwise, if is atomic and either v() = 0 or v() = 0 then v() = 1, by
(vCon)a , and so v() = 0, by (vNeg)c .
By extending the proof of soundness and completeness of mbC w.r.t. valuations,
it is immediate to prove the following:
Theorem 4.4.21 (Soundness and Completeness of P1 w.r.t. valuations) For every
 {} L :
 P1  P1 .
4.4 Some 3-Valued LFIs 147

By considering the 3-valued matrix logic P1 as defined over the signature 

(recalling that conjunction and disjunction are interpreted as the operators  and 
defined in Sect. 4.4.2), the proof of the following result is easy.

Lemma 4.4.22 Let v be a P1 -valuation. Let h be the valuation for P1 such that, for
every atomic formula p,

1 iff v( p) = 1, and v( p) = 0
h( p) = 21 iff v( p) = 1, and v( p) = 1

0 iff v( p) = 0.
 
Then, for every L , v() = 1 iff h() D, where D = 1, 21 .

Proof Given a P1 -valuation v, let h be the valuation for P1 defined as indicated.
The result will be proven by induction on the complexity of .
If is atomic, the result follows by definition of h.
Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH). If is a non-atomic formula then, by definition of the operators in
P1, h() D iff h() = 1.
Case 1: = . Then:
() Assume that v() = 1. If is atomic, suppose that v() = 1. Then h() = 21 ,
by definition of h, and so h() = P h() = 1, by definition of P . Otherwise, if
v() = 0, then h() = 0, by definition of h, and so h() = P h() = 1. Now, if
is not atomic then v() = 0, by Definition 4.4.20, and so, by (IH), h() = 0. Thus,
h() = 1.
() If h() = 1 then, by the truth-table of P , h() {0, 21 }. If h() = 0 then, by
(IH), v() = 0 and so v() = v() = 1. Otherwise, if h() = 21 then is atomic
and so, by definition of h, v() = v() = 1.
Case 2: = . Then:
() Assume that v() = 1. Suppose that is atomic. If v() = 1 then v() = 0
and so h() = 1, by definition of h. Then h() = 1. Otherwise, if v() = 0 then
v() = 1 and so h() = 0, by definition of h. Then h() = 1. In case is not
atomic then h() {0, 1}, by the truth-tables of P1, and so h() = 1.
() If h() = 1 then h() {0, 1}, by the truth-table of . If is complex then
v() = 1. Suppose now that is atomic. If h() = 0 then v() = 0, by (IH), whence
v() = 1 and so v() = 1. Otherwise, if h() = 1 then v() = 1, by (IH). If
v() = 1 then, by (IH) (recalling that the complexity of is strictly lower than the
complexity of ), h() = 1, a contradiction (since h() = 1). Then v() = 0
and so v() = 1.
Case 3: = . Then: v() = 1 iff v() = 1 and v( ) = 1 iff (IH) both h(),
h( ) D iff h() D.
Case 4: = . Then: v() = 1 iff either v() = 1 or v( ) = 1 iff (IH) either
h() D or h( ) D iff h() D.
148 4 Matrices and Algebraizability

Case 5: = . Then: v() = 1 iff either v() = 0 or v( ) = 1 iff (IH) either

h() = 0 or h( ) D iff h() D. 

Theorem 4.4.23 The logic P1 coincides with P1, that is, P1 is sound and complete
w.r.t. the matrix logic P1: for every  {} L ,  P1 iff  P1 .

Proof The proof is divided in two parts: soundness and completeness.

(Soundness) It is an easy task to check that every axiom of P1 is valid in P1. Clearly,
if h() D and h( ) D, then h() D, for every valuation h for P1. From
this, it is proved by induction on the length of a derivation in P1 : if  P1 , then
 P1 .
(Completeness) Suppose that  P1 , and let v be a P1 -valuation such that
v[] {1}. By the previous lemma, there exists a valuation h for P1 such that, for
every L , v() = 1 iff h() D. From this, h[] D and then, by hypothesis,
h() D. Hence v() = 1. This means that  P1 and so, by Completeness of
P1 w.r.t. valuations,  P1 . 

Finally, a new proof of the maximality of P1 w.r.t. CPL will be given below, based
on a very general proof of maximality for propositional logics presented in [49].
Firstly, observe that the truth-tables of P1 extend the ones of CPL. That is, the
operators and P , when restricted to {0, 1}, coincide with the classical implication
and the classical negation, respectively. Moreover, by the definition of the operators
in P1, if is a complex formula and h is a homomorphism (that is, a valuation) for
P1 then h() {0, 1}. Accordingly, it is easy to prove the following DAT for P1:
Proposition 4.4.24 Let  {} be a set of formulas over  P whose propositional
variables belong to { p1 , . . . , pn }. Then

 CPL p1 , . . . , pn ,  P1 .

Taking into account that CPL can be presented in signature  P , the proof of maxi-
mality of P1 w.r.t. CPL can now be proved:
Theorem 4.4.25 The logic P1 is maximal w.r.t. CPL defined over the signature
 P . That is: if is a formula over  P such that CPL but P1 then the logic
P1+ obtained from P1 by adding as a theorem (together with all its instances by
substitutions) coincides with CPL.

Proof Let ( p) = ( p p) and ( p) = P ( p p), for any propositional vari-

able p. Then h( ( p)) = 1 and h(( p)) = 0, for every P1-valuation h.
Let ( p1 , . . . , pk ) be a tautology of CPL which is not valid in P1. Then there
exists a P1-valuation h 0 such that h 0 () = 0. Since the operators of P1 extend the
ones of CPL, there is some 1 j k such that h 0 ( p j ) = 21 . Now, consider the
substitution over  P such that, for every propositional variable p,
4.4 Some 3-Valued LFIs 149

( p1 ) if h 0 ( p) = 1,
( p) = ( p1 ) if h 0 ( p) = 0,

p1 if h 0 ( p) = 21 ,

and let ( p1 ) = ().5 It is immediate that, if p is a propositional variable and h

def

is a P1-valuation, then h( ( p)) = 1 if h( p) {0, 1}. It follows from the fact that
( p), being an instance of , is a tautology of CPL. On the other hand, if h( p) = 21
then h( ( p)) = h 0 (), by the very construction of the formula (it can be easily
proved by induction on the complexity of ). In this way, the following holds, for
every P1-valuation h:

1 if h( p) {0, 1},
h( ( p)) =
0 if h( p) = 21

for every propositional variable p. That is, the truth-table of ( p) coincides with
the one of the consistency operator ( p). Now, let  {} be a set of formulas
over  P whose propositional variables belong to { p1 , . . . , pn } such that  CPL .
Then p1 , . . . , pn ,  P1 , by Proposition 4.4.24. By the observation above, this
means that ( p1 ), . . . , ( pn ),  P1 . But the later implies that  P1+ , by
definition of P1+ and the fact that every formula ( p) is an instance of . Since P1+
is contained in CPL, it follows that P1+ coincides with CPL and so P1 is maximal
w.r.t. CPL. 
In Chap. 7, a first-order version of P1 will be defined, by extending the system P1 .

4.4.5 Asenjo-Priests Logic LP (19661979)

In 1966, Asenjo proposed a propositional logic to deal with antinomies. Asenjos

basic idea, at the semantical level, is that a proposition
 can receive
 either one or two
truth-values: it can be true, false or both, that is, true, false . In the latter case, the
proposition is an antinomy. Based on this idea, the classical 2-valued truth-tables were
extended by taking this possibility into account. So, for instance, if a proposition is
true and another proposition is antinomic, the compound proposition is also
antinomic (by the laws of the material implication extended to this setting). The
resulting truth-tables for the connectives of implication, conjunction, disjunction,
equivalence and negation coincide with the corresponding ones proposed by Kleene
in [50] for his strong logic K3 defined over the truth-values 1, 21 and 0. In K3,
1 and 21 are designated; however, tautologies only can get the value 1. It is worth
noting that (Kleene and Asenjos) 3-valued tables for conjunction, disjunction and
negation concide with the ones proposed by ukasiewicz in [43] for his well-known
logic 3 (recall Definition 4.4.13 and the comments below it). However, 1 is the only

that, if is a substitution for variables, then denote its unique extension to an endomor-
5 Recall

phism over the algebra of formulas.

150 4 Matrices and Algebraizability

designated value in 3. Apparently, 1 (seen as true) and 21 (seen as antinomic)

are the designated values in Asenjos approach, but it is not completely clear in view
of the following paragraph (see [51], p. 104),6 preceding the definition of an axiom
system for the logic of antinomies:
We move now from semantics to syntax. What we require is a system of axioms for a non-
trivially inconsistent propositional calculus. It should not be possible to prove any proposition
from A and A, which means that the rule of deduction to the absurd should not be deducible.
Of course, the principle of contradiction should not be deducible either.

On the one hand, 21 should be taken as a designated value in order to admit non-
trivial contradictions, given that, according to Asenjos definitions, the negation of
an antinomy is an antinomy (formally, the negation of 21 is 21 ). But, on the other hand,
the schema ( ) (formalizing the principle of contradiction in that language)
is valid in Asenjos logic if 21 is taken as a designated value. This would contradict
Asenjos latter requirement. Very possibly his semantical approach would be some-
thing similar to Kleenes (in which tautologies only can get the truth-value 1, despite
1
2
being also a designated value), but this is mere speculation.
The idea of considering a 3-valued logic extending the classical 2-valued matrices
with a third value (representing both true and false) in order to deal with paradoxes
and antinomies was also pondered by Priest in 1979, when he introduced in [52]
the 3-valued logic LP (the Logic of Paradox). This logic, which can be seen as a
formalization of Asenjos logic from the point of view of logical matrices, is probably
the best known many-valued paraconsistent logic. As suggested by its name, LP was
designed for handling logical paradoxes. In Priests perspective, similar to Asenjos,
the truth-value 1 represents trueness, the truth-value 0 represents falseness, while
the third truth-value 21 is both true and false, that is, paradoxical. It is worth noting
that Kleenes interpretation for the non-classical third-value 21 is different from the
paradoxical given by Priest and the antinomical given by Asenjo: it represents an
undefined or undetermined informational state.
Up to minor differences in notation, the original presentation of LP was as follows:

Definition 4.4.26 (Asenjo-Priests logic LP[52]) Let M L P = M, D be the 3-

valued matrix over  L P = {, } with domain M = {1, 21 , 0} and set of designated
values D = {1, 21 }, where the truth-tables associated with each connective are defined
as follows:

The logic associated with the matrix M L P is called LP. 

6 This somewhat unclear aspect of Asenjos logic was already observed by Priest in [52], p. 228.
4.4 Some 3-Valued LFIs 151

Priest proposed also several derived connectives defined from M L P : a disjunc-

tion = ( ), an implication L P = ( ) and a bi-
def def

implication L P = ( L P ) ( L P ).
def

Concerning the relationship of LP with Asenjos approach, Priest himself men-

tioned in [53], p. 157 the following:
The logic [[of antinomies of Asenjo]] is a very simple and natural one, and has been redis-
covered a number of times since. For example, it and its properties were spelled out in more
detail in Priest [1979],7 where it is termed L P (the Logic of Paradox), a name by which it
is now standarly known.

The first-order version of LP, called LPQ, was also presented in [52]. The seman-
tics is given by (usual) first-order structures together with mappings which assign
any truth-value of {1, 21 , 0} to the atomic sentences. Complex sentences of the form
and are evaluated by using the truth-tables of LP. Finally, a sentence
= x gets the value 1 if [x/a] gets the value 1 for any a in the domain of
def

the given first-order structure (here, a is a new constant associated to a); takes the
value 0 if [x/a] takes the value 0 for some a; and gets the value 21 otherwise, that
is, if the truth-value of [x/a] takes the value 1
is designated for every a, and [x/b] 2
for some b. This semantics can be alternatively expressed by associating pairs of sets
to any predicate, see [54], pp. 7678 (an equivalent semantical approachin terms
of triples of setswill be given in Sects. 7.9 and 7.10 of Chap. 7 for some first-order
LFIs). As proved in [40], the logic LPQ is a (proper) fragment of the first-order
version of J3. Clearly, LP is a (proper) fragment of J3.
The logic LP can be viewed as a simple way to formalize, under the point of
view of many-valued logics, the philosophical perspective underlying Priest and
Sylvans dialetheism (see [53, 55]). The main thesis of dialetheism is that there are
true contradictions, that is, that some sentences can be both true and false, at the
same time and in the same way. A brief discussion about dialetheism can be found
in Chap. 1 of this book.

4.4.6 Ciore and Other Related Systems

Recall that in Sect. 4.3 the logic Ciore was proposed as a suitable extension of
Cio which is algebraizable in the sense of Blok and Pigozzi (see Theorem 4.3.18).
However, as was mentioned in [22], the logic Ciore is so strong that it can be
characterized by a 3-valued matrix logic as introduced in [34]. This claim will be
proved in Theorem 4.4.29 below.
Definition 4.4.27 ([34]) Let Me = M, D be the 3-valued matrix over the signa-
ture  with domain M = {1, 21 , 0} and set of designated values D = {1, 21 } such that
the truth-tables associated to each connective are the following:

7 The author refers to [52].

152 4 Matrices and Algebraizability

The logic associated with this matrix is called LFI2. 

Let LFI2 be the consequence relation of LFI2. That is:  LFI2 iff for every
-homomorphism h : L D, if h[] D then h() D. Such homomorphisms
are the valuations for LFI2.
In order to prove that Ciore coincides with LFI2 (or, equivalently, that Ciore
is sound and complete w.r.t. the matrix Me , as mentioned in [22]), the following
technical result will be useful.
Lemma 4.4.28 Let v be a Ciore-valuation. Then thereexists a valuation h for LFI2
such that, for every L , v() = 1 iff h() 1, 21 .

Proof Let v be a valuation for Ciore. Consider a mapping h 0 : V ar D such that,

for every atomic formula p,

1 iff v( p) = 1, and v( p) = 0
h 0 ( p) = 21 iff v( p) = 1, and v( p) = 1

0 iff v( p) = 0.
Furthermore, let h be the unique valuation for LFI2 extending h 0 such that h( p) =
h 0 ( p) for every p V ar . By induction on the complexity of a formula in L , we
shall prove the following:

1 iff v() = 1, and v() = 0 (a)
h() = 21 iff v() = 1, and v() = 1 (b)

0 iff v() = 0. (c)

If is atomic, the result follows by definition of h and h 0 .

Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH).
Case 1: = . Then:
(a)() If h() = h() = h() = 1, then h() = 0 by the truth-table of .
By (IH), v() = 0 and so v() = v() = 1. Using (vCf ), v() = v() =
v() = 0.
4.4 Some 3-Valued LFIs 153

(a)() If v() = 1 and v() = 0, then v() = v() = 0 by (vCe). By (IH),

h() = 0 and then h() = h() = h() = 1.
(b)() If h() = h() = 21 , then h() = 21 by definition of the truth-table of
. By (IH), v() = 1 and v() = 1, so v() = v() = 1. By (vCe), v() =
v() = v() = 1.
(b)() If v() = v() = 1, then v() = v() = 1 by (vCf ). Using (IH), h() =
1
2
, so h() = h() = h() = 21 .
(c)() If h() = h() = 0, then h() = 1. By (IH), v() = 1 and v() =
v() = 0.
(c)() If v() = v() = 0, then v() = 1. Using (IH), h() = 1, so h() =
h() = h() = 0.
Case 2: = . Then:
(a)() If h() = h() = h() = 1, then h() {0, 1} by the truth-table of .
(a.1) If h() = 0 then, by (IH), v() = 0 and so v() = 1. From here, v() =
v() = 1 by (vCi) and v() = v() = 0.
(a.2) If h() = 1, then v() = 1 and v() = 0 by (IH). Using (vCi), v() =
v() = 1 and v() = v() = 0.
(a)() If v() = v() = 1 and v() = 0, then v() = 0 or v() = 0 by (vCon).
(a.3) If v() = 0, then, by (IH), h() = 0 and then h() = h() = h() = 1.
(a.4) If v() = 0, then v() = 1. By (IH), h() = 1 and then h() = h() =
h() = 1.
(b) By the definition of the truth-table of , h() = h() = h() = 21 . On the
other hand, by (vCon) and (vCi), it is impossible to have v() = v() = 1 and
v() = v() = 1 at the same time.
(c)() If h() = h() = 0, then h() = 21 by the truth-table of . By (IH),
v() = v() = 1 and so v() = v() = 0.
(c)() If v() = v() = 0, then v() = v() = 1 by (vCi). Using (IH), h() = 21
and then h() = h() = h() = 0.
Case 3: = . Then:
 1  If h() = h(
(a)()   ) = h() h( ) = 1, then, by the truth-table of , h()
1, 2 , and h( ) 1, 21 , and either h() = 1 or h( ) = 1. By (IH), v() = v( ) =
1 and either v() = 0 or v( ) = 0. From (vAnd) and (vCi), v() = v( ) = 1
and either v() = 1 or v( ) = 1. By (vCo1 ), v(( )) = 1 and so v() =
v(( )) = 0 by (vCon).
(a)() Suppose that v() = v( ) = 1 and v() = v(( )) = 0.
 having v() =
Using (vCr 11 ) (see Proposition 4.3.12) and (vAnd), besides
  v() = 1,
either v() = 0 or v( ) = 0. By (IH), h() 1, 21 and h( ) 1, 21 , such
that either h() = 1 or h( ) = 1. Therefore, by the truth-table of , h() =
h() h( ) = 1.
(b)() If h() = h() h( ) = 21 , then h() = h( ) = 21 by definition of the
truth-table of . By (IH), v() = v() = 1 and v( ) = v( ) = 1. By (vAnd)
and (vCr 11 ), v() = v( ) = 1 and v() = v(( )) = 1.
(b)() If v() = v( ) = 1 and v() = v(( )) = 1, then v() =
v( ) = 1 by (vAnd) and so v( ) = 1 by (vCo11 ) (see Sect. 3.8 of Chap. 3). Then,
154 4 Matrices and Algebraizability

by (vAnd) and (vCo12 ), v() = 1. Using (IH), h() = h( ) = 21 and hence h() =
h() h( ) = 21 .
(c)() If h() = h() h( ) = 0, then h() = 0 or h( ) = 0. By (IH), v() = 0
or v( ) = 0 and then v() = 0 by (vAnd).
(c)() If v() = v( ) = 0, then v() = 0 or v( ) = 0 by (vAnd). Using (IH),
h() = 0 or h( ) = 0 and so h() = h() h( ) = 0.
The proof for the other cases, namely when is or , is entirely
analogous. Details are left to the reader. 

It is worth noting that all the clauses for Ciore-valuations are used in the proof
above. The adequacy of Ciore w.r.t. the proposed 3-valued matrix semantics follows,
then, easily:
Theorem 4.4.29 The logic Ciore coincides with LFI2; that is, Ciore is sound and
complete w.r.t. matrix Me : for every  {} L ,  Ciore iff  LFI2 .

Proof The proof is divided in two parts.

(Soundness) It is easy to see that every axiom of Ciore is a tautology of matrix Me .
Additionally, if h() D and h( ) D, then h() D for every valuation h
over Me . From here, as usual, it is easy to prove (by induction on the length of a
derivation in Ciore) that if  Ciore , then  LFI2 .
(Completeness) Suppose that  LFI2 and let v be a Ciore-valuation such that
v[] {1}. By Lemma 4.4.28, there exists a valuation h for LFI2 such that for every
L , v() = 1 iff h() D. That being so, h[] D and then, by hypothesis,
h() D. Thus, v() = 1. This shows that  Ciore and then, by Theorem 4.3.10,
 Ciore . 

As a consequence of the last result, it is now easy to see that, if we extend Ciore by
adding axiom schema (cl) to get an extension of C1+ presented over the signature ,
the resulting system collapses with CPL , the inessential extension of CPL obtained
by adding the axiom schema (see Definition 2.4.4).

Corollary 4.4.30 Let Cilore be the logic obtained from Ciore by adding axiom (cl).
Then Cilore coincides with CPL .

Proof Observe the truth-table for the instance = ( p1 p1 ) p1 of (cl)

def

in Me :

p1 p1 p1 p1 ( p1 p1 ) p1 ( p1 p1 ) p1
1 0 0 1 1 1
1 1 1 1
2 2 2 2 0 0
0 1 0 1 1 1
4.4 Some 3-Valued LFIs 155

In order to validate the formula , the valuations cannot assign the value 21 to p1 . This
 
means that Cilore is characterized
  by Me by using valuations such that h( p) 0, 1
for every p V ar . Since 0, 1 is a subalgebra of M which coincides with the
algebra underlying the matrix semantics of CPL , the logic Cilore concides with
CPL . 

Indeed, the last result is solely a particular case of a more general fact: as occurs
with P1 (recall Theorem 4.4.25), the logic Ciore is maximal with respect to CPL .
This means that, if a formula is a CPL -tautology, but some instance of (seen
as an axiom schema) is not valid in Ciore, then the system obtained from Ciore by
adding as an axiom schema coincides with CPL .
In an unpublished manuscript [56] (see more details in [7]), J. Marcos proposes
a generalization of the 3-valued maximal paraconsistent logics P1 and J3. The gen-
eralization consists of considering any matrix M = M, D over  with domain
M = {1, 21 , 0} and set of designated values D = {1, 21 } such that the operators ,
and (as well as the defined connective ) satisfy the clauses of Proposition 4.4.7.
The consistency operator is defined as in the previous logics, while the negation
operator is defined either as in P1 or as in the other logics examined above. Then, by
taking a formula ( p1 , p2 ), as in Definition 4.4.8 (where the connectives involved in
its definition are the primitive ones of the given matrix logic), a result analogous to
Proposition 4.4.9 can be obtained. From here, it follows easily that the given matrix
logic is algebraizable in the sense of Blok and Pigozzi, by proving a result analogous
to Theorem 4.4.10.
There are potentially 8,192 of such logics. However, the question of determining
how many distinct (that is, inequivalent) logics can be constructed in this manner
remains open.
Another interesting feature of these logics is that all of them are maximal relative
to CPL : the proof is similar to that of P1 (see Theorem 4.4.25). Two interesting
facts can be also mentioned: the operators of Settes matrix M P are definable in any
other of such matrices. In turn, the matrix M J of J3 (considered as LFI1 in this
context) is able to define all the operators of the other matrices.
Finally, we will discuss an improvement to the maximimality result of these logics
(w.r.t. CPL ) which can be obtained from a general result stated in [57].
According to the definition adopted here, a logic L1 is said to be maximal with
respect to another logic L2 when: (i) both logics are defined over the same language;
(ii) the consequence relation of L1 is contained in that of L2 ; and (iii) if is a
formula such that is valid in L2 but not in L1 , then the extension of L1 obtained
by adding as an axiom schema coincides with L2 .
A stronger notion of maximality was proposed in [58]: instead of considering in
Item (iii) an axiom schema , an inference rule schema of the form {1 , . . . , n }, 
is assumed. The first notion of maximality is called weak maximality, while the latter
is called strong maximality. Obviously strong maximality implies weak maximality,
but the converse is generally not true (see [58]).
156 4 Matrices and Algebraizability

In [57], sufficient conditions were obtained to guarantee the equivalence between

both notions of maximality within the framework of Hilbert calculi (see Theo-
rem 4.4.34 below). Firstly, observe that an inference rule schema {1 , . . . , n }, 
generalizes the idea of axiom schema: every instance of the rule by substitutions
can be considered in a derivation. In order to prove Theorem 4.4.34, some technical
lemmas are necessary.
Lemma 4.4.31 Let L be a standard propositional logic defined by a Hilbert cal-
culus over a signature  and let Lr be the logic obtained from L by adding the
inference rule schema r = {1 , . . . , n }, . Assume that L has (possibly derived)
connectives (binary) and (unary) such that the following holds for every set of
formulas  {, } in L :
(i) , ( ) L ;
(ii) L ( );
(iii) L ( );
(iv) If , Lr and , Lr then  Lr .
Let p be a propositional variable not occurring in {1 , . . . , n , }. Then

( p 1 ), . . . , ( p n ) Lr ( p ) .

Proof For j = 1, . . . , n, it holds that ( p j ), p Lr j by (i) and by the definition

of Lr . From this, ( p 1 ), . . . , ( p n ), p Lr j for every 1 j n. Since
1 , . . . , n Lr then ( p 1 ), . . . , ( p n ), p Lr . From here,
( p 1 ), . . . , ( p n ), p Lr ( p ), as Lr ( p ) and (i). On the
other hand, p Lr ( p ) by (iii) and so ( p 1 ), . . . , ( p n ), p Lr
( p ). By (iv), it follows that ( p 1 ), . . . , ( p n ) Lr ( p ) as required.

Corollary 4.4.32 Consider L and Lr as in Lemma 4.4.31. Assume that L satisfies
the deduction meta-theorem, DMT: , L iff  L ( ) for every 
{, } L . Then Lr also satisfies DMT with respect to .
Proof Assume that , Lr . By induction on the length of the derivation 1 . . . k
of from  {} in Lr , it will be shown that  Lr ( ). Since L satis-
fies DMT with respect to , the only case to be analyzed is when is obtained
by using the rule r = {1 , . . . , n }, . Thus, there is some substitution , such
that = () and { (1 ), . . . , (n )} {1 , . . . , k1 }. By induction hypothesis,
 Lr ( ( j )) for every 1 j n. Given that p does not occur in r , we can
assume that ( p) = and so, by Lemma 4.4.31 and the structurality of Lr , (
(1 )), . . . , ( (n )) Lr ( ()). Hence,  Lr ( ()); that is,
 Lr ( ). The converse follows from condition (i) of Lemma 4.4.31. 
Corollary 4.4.33 Consider L and Lr as in Corollary 4.4.32. Let L r be the logic
obtained from L by adding the axiom schema 1 (2 ( (n ) . . .).
Then the logics Lr and L r coincide; that is, for every  {} L ,  Lr iff
 L r .
4.4 Some 3-Valued LFIs 157

Proof It is clear that 1 , . . . , n Lr and then, by DMT, Lr 1 (2 (

(n ) . . .). This means that L r is a sublogic of Lr . On the other hand, by
condition (i) of Lemma 4.4.31, it holds that 1 , . . . , n L r . From this, Lr is a
sublogic of L r , completing the proof. 

Theorem 4.4.34 Let L be a logic given by a Hilbert calculus such that L is weakly
maximal with respect to another logic L . Assume that L satisfies the conditions
of Corollary 4.4.32 for every rule r = {1 , . . . , n },  and that L is finitary and
satisfies DMT with respect to . Then L is strongly maximal with respect to L .

Proof Let  {} be a set of formulas such that  L  but  L . By finitari-

ness of L , there exists a finite set {1 , . . . , n }  such that 1 , . . . , n L 
and 1 , . . . , n L . Since L satisfies DMT with respect to it follows
that L  1 (2 ( (n ) . . .). Since L satisfies condition (i) of
Lemma 4.4.31, it also follows that L 1 (2 ( (n ) . . .). As L
is weakly maximal with respect to L , it follows that the logic L r obtained from
L by adding the axiom schema 1 (2 ( (n ) . . .) coincides with
L . On the other hand, L r coincides with Lr , i.e., the logic obtained from L by
adding the inference rule schema {1 , . . . , n }, , by Corollary 4.4.33. This means
that the logic Lr is L and so L is strongly maximal with respect to L . 

Proposition 4.4.35 The logic P1 is strongly maximal with respect to CPL in a

signature solely containing and .

Proof Recall from Definition 4.4.17 that P1 can be defined over a signature just
containing an implication and a negation , such that P1 is weakly maximal
w.r.t. CPL (see Theorem 4.4.25). Additionally, CPL is finitary and satisfies DMT
with respect to . As previously shown, it is possible to define a classical nega-
tion = ( ) and a disjunction = ( ) in P1. Then
def def

P1 ( ) for every . Additionally, , P1r and , P1r implies that

, ( ) P1r for every , , , where P1r is any extension of P1 by an infer-
ence rule r . Thus, P1 satisfies the conditions (i)(iv) of Lemma 4.4.31 for every
extension P1r . On the other hand, P1 satisfies DMT with respect to . Thus, P1
and every extension P1r of it satisfies the hypothesis of Corollary 4.4.32. Thus, by
Theorem 4.4.34, it follows that P1 is strongly maximal with respect to CPL. 

Remark 4.4.36 The strong maximality of P1 with respect to CPL was already stated
in [58] using general techniques of non-deterministic matrices.

Corollary 4.4.37 The logics S3 and J3 are strongly maximal with respect to CPL .
Moreover, all the 8,192 three-valued matrix logics of [56] are strongly maximal with
respect to CPL .

Proof The proof is similar to that of P1 given in Theorem 4.4.25. 

158 4 Matrices and Algebraizability

4.4.7 LFI1, MPT and J3

As mentioned in Remark 4.4.12, the logic J3 was independently studied by several

authors at several times, and with different motivations. From the point of view of
LFIs, it is important to notice that this logic was re-introduced by Carnielli et al. [34]
under the name of LFI1 (see also [35]), as a logical tool for analyzing inconsistent
databases. The signature assumed by the authors was  = {, , , , }, that is,
with the inconsistency operator as a primitive; this makes sense, since the logic LFI1
is intended to deal with inconsistent databases, and so the inconsistency operator
arguably can be seen as more relevant, in this context, than the consistency operator
. In [7] the logic LFI1 was axiomatized in the signature  = {, , , , }, but
most of the axioms describe properties of (now considered as the defined connective
= ) instead of considering properties of itself (that axiomatization is
def

reproduced below, in Definition 4.4.40). For this reason, that axiomatization is a mere
translation (in the signature ) of the axiomatization of LFI1 originally presented in
terms of the inconsistency operator . In this section a new axiomatization of LFI1
in terms of will be given, such that the properties of will be more salient (see
Definition 4.4.41 below).
It was also observed in Remark 4.4.12 that the logic J3 was independently re-
introduced, yet one more time, by Coniglio and Silvestrini in [39, 40] under the name
MPT, so constituting the propositional basis of a first-order logic called LPT1. The
latter is the logical counterpart of a Tarskian-style semantics of triples for first-order
languages called partial structures (or pragmatic structures), which generalize the
theory of Quasi-Truth introduced in [41] by Mikenberg, da Costa and Chuaqui (see
an additional discussion on Quasi-truth in Sect. 9.3 of Chap. 9). A brief presentation
of MPT and the relationship with both J3 and LFI1 will be given below. On the
other hand, in Chap. 7 the Tarskian-style semantics of pragmatic structures intro-
duced in [40] for logic LPT1 will be adapted to the first-order extension of a new
axiomatization of LFI1 to be presented in Definition 4.4.41 below. This will consti-
tute an important link between the two-valued semantics for LFIs presented in the
first part of Chap. 7 (which was taken from [59]), and the semantics of triples for
3-valued LFIs proposed in [39, 40]. The same kind of connection will be obtained
for the first-order version of Settes logic P1 to be presented also in Chap. 7, by using
its axiomatization as an LFI given in Definition 4.4.19.
Firstly, the original presentation of the logic LFI1 will be recalled now:
Definition 4.4.38 (Propositional logic LFI1[34, 35]) Let MLFI1 = M, D be the
3-valued matrix over  = {, , , } with domain M = {1, 21 , 0} and set of des-
ignated values D = {1, 21 } such that the implication , the conjunction , the dis-
junction and the negation are defined as in J3 (the implication is J , see
Sect. 4.4.3), and with (x) = (x), where is the consistency operator definable
in J3. That is, the truth-tables of the operators in MLFI1 are the following:
4.4 Some 3-Valued LFIs 159

The logic LFI1 is the one associated to the matrix MLFI1 . 

The valuations over LFI1 are the homomorphisms h : L M. Clearly, the
consistency operator (that is, Halldns meaningfulness operator) can be defined in
MLFI1 as (x) = (x). It coincides with the operator given in Definition 4.4.27.
def

On the other hand, the operator of J3 can be defined as (x) = x (x), and so
def

the matrices of J3 and LFI1 are functionally equivalent being, by its turn, functionally
equivalent to that of ukasiewicz 3-valued logic 3. Observe that (x) = (x
def

(x)) using the primitive operators of MLFI1 .

For convenience, from now on the version LFI1 of logic LFI1 defined over
signature  will be regarded.

Definition 4.4.39 Let LFI1 be the matrix logic defined over signature  obtained
from the 3-valued matrix of logic LFI1 (see Definition 4.4.38) by taking instead
of as a primitive operator. The 3-valued matrix for LFI1 over  will be called
MLFI1 .

Clearly, LFI1 and LFI1 coincide, up to language. The axiomatization of LFI1

proposed in [7] is the following:

Definition 4.4.40 (An axiomatization of LFI1 [7]) Let LFI1 be the Hilbert calcu-
lus over  obtained from mbCci (recall Definition 3.1.7) by adding the axioms (ce),
(cf) (Recall Sect. 3.5)) plus the following (where and are abbreviations
of and ( ) ( ), respectively):

( ) (( ) ( )) (cj1 )
( ) (( ) ( )) (cj2 )
( ) ( ) (cj3 )

A new axiomatization of LFI1 will be offered now, also in the signature  of

mbC, in which the negation plays a more important role than in LFI1 .
160 4 Matrices and Algebraizability

Definition 4.4.41 Let LFI1 be the Hilbert calculus over  obtained from mbCci
(recall Definition 3.1.7) by adding the axioms (ce), (cf) (Recall Sect. 3.5)) plus the
following:

( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg 1 )
( ) ( ) (neg 2 )

Axioms (neg1 ), (neg2 ), (neg1 ), (neg2 ), (neg 1 ) and (neg 2 ) are also
included in the usual Hilbert calculus for Nelsons paraconsistent logic N4, to be ana-
lyzed in Chap. 5 (see Definition 5.1.24); recall from Sect. 3.6 that (neg1 ), (neg1 )
and (neg 1 ) were also considered for characterizing consistency propagation. They
express the De Morgans laws, in this case with respect to the paraconsistent negation
, as well as a useful reduction law for negated implications.

Definition 4.4.42 We say that an mbCci-valuation v (recall Definition 3.1.7) is an

LFI1 -valuation if it also satisfies (vCf ) and (vCe) (see Sect. 3.5), plus the following
clauses:
(vDM ) v(( )) = 1 v() = 1 or v() = 1.
(vDM ) v(( )) = 1 v() = v() = 1.
(vCIp ) v(( )) = 1 v() = v() = 1. 

Let LFI1 be the consequence relation w.r.t. LFI1 -valuations. The proof of the
following result is now straightforward:
Theorem 4.4.43 (Soundness and Completeness of LFI1 w.r.t. valuations) For
every  {} L :

 LFI1  LFI1 .

Now, the logic LFI1 will be proved to be equivalent to the logic LFI1 (with its
associated semantics of LFI1 -valuations). The technique will be analogous to that
used for P1 and Ciore.
Lemma 4.4.44 Let v be an LFI1 -valuation. Then there  exists
 a valuation h for
LFI1 such that, for every L , v() = 1 iff h() 1, 21 .

Proof Let v be a valuation for LFI1 . Consider the unique valuation h for LFI1
such that, for every atomic formula p,
4.4 Some 3-Valued LFIs 161

1 iff v( p) = 1, and v( p) = 0
h( p) = 21 iff v( p) = 1, and v( p) = 1

0 iff v( p) = 0.

By induction on the complexity of a formula in L , it will be proved the following:

1 iff v() = 1, and v() = 0 (a)
h() = 21 iff v() = 1, and v() = 1 (b)

0 iff v() = 0. (c)

If is atomic, the result follows by the very definition of h.

Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH).
Case 1: = # where # {, }. Given that the truth-tables for and in LFI1
coincide with the respective tables in Ciore, the proof of this case is the same as the
corresponding one in the proof of Lemma 4.4.28, since the only clauses used for v
in that proof are (vCf ), (vCe), (vCon) and (vCi).
Case 2: = . Then:
(a) h() = h( ) = h() h( ) = 1 iff, by the truth-table of , h() = h( ) =
1, iff, by (IH), v() = v( ) = 1 and v() = v( ) = 0, iff, by (vAnd) and (vDM ),
v( ) = 1 and v(( )) = 0.  
(b) h() = h() h( ) = 21 iff h(), h( ) 1, 21 and either h() = 21 or h( ) =
1
2
, by definition of the truth-table of , iff, by (IH), v() = v( ) = 1, and either
v() = v() = 1 or v( ) = v( ) = 1, iff either v() = v() = v( ) = 1 or
v() = v( ) = v( ) = 1, iff v() = v( ) = 1, and either v() = 1 or v( ) =
1 (exercise to the reader), iff, by (vAnd) and (vDM ), v() = v( ) = 1, and
v() = v(( )) = 1.
(c) h() = h() h( ) = 0 iff either h() = 0 or h( ) = 0, iff, by (IH), either
v() = 0 or v( ) = 0, iff v() = 0, by (vAnd).
Case 3: = . Then:
(a) h() = h( ) = h() h( ) = 1 iff, by the truth-table of , either h() = 1
or h( ) = 1, iff, by (IH), either v() = 1 and v() = 0, or v( ) = 1 and v( ) =
0, iff, by (vNeg), either v() = 1 or v( ) = 1, and either v() = 0 or v( ) = 0
(exercise to the reader), iff v( ) = 1 (by (vOr) and v(( )) = 0 (by
(vDM )).
(b) h() = h() h( ) = 21 iff, by the truth-table of , either h() = h( ) = 21 ,
or h() = 21 and h( ) = 0, or h() = 0 and h( ) = 21 , iff, by (IH), either v() =
v() = v( ) = v( ) = 1, or v() = v() = 1 and v( ) = 0, or v() = 0 and
v( ) = v( ) = 1, iff, by (vNeg), either v() = 1 or v( ) = 1, and v() =
v( ) = 1 (exercise to the reader), iff v( ) = 1 (by (vOr)) and v(( )) =
1 (by (vDM )).
(c) h() = h() h( ) = 0 iff h() = h( ) = 0 iff, by (IH), v() = v( ) = 0 iff
v() = v() = v( ) = 0, by (vOr).
162 4 Matrices and Algebraizability

Case 4: = . Then:
(a) h() = h() h( ) = 1 iff, by the truth-table of , either h() = 0 or
h( ) = 1, iff, by (IH), either v() = 0, or v( ) = 1 and v( ) = 0, iff either
v() = 0 or v( ) = 1, and either v() = 0 or v( ) = 0, iff v( ) = 1 (by
(vImp)), and v(( )) = 0 (by (vCip )).  
(b) h() = h() h( ) = 21 iff, by the truth-table of , h() 1, 21 and
h( ) = 21 , iff, by (IH), v() = v( ) = v( ) = 1, iff either v() = 0 or v( ) = 1,
and v() = v( ) = 1 (exercise to the reader), iff v( ) = 1 (by (vImp)), and
v(( )) = 1 (by (vCip )).  
(c) h() = h() h( ) = 0 iff h() 1, 21 and h( ) = 0 iff, by (IH), v() = 1
and v( ) = 0 iff v( ) = 0. 

From the previous Lemma, the equivalence between LFI1 and the logic generated
by the Hilbert calculus LFI1 follows easily. That is, the calculus LFI1 is sound
and complete w.r.t. the 3-valued matrix semantics of the logic LFI1 (considered over
the signature ):
Theorem 4.4.45 The logic LFI1 coincides with LFI1 , that is, LFI1 is sound
and complete w.r.t. the matrix of LFI1 : for every  {} L ,  LFI1 iff
 LFI1 .

Proof The proof can be divided in two parts.


(Soundness) It is immediate
 to see that every axiom
 of LFI1 is a tautology
  of LFI1 .
Besides this, if h() 1, 21 and h( ) 1, 21 , then h() 1, 21 , for every
valuation h over LFI1 . Thus, by induction on the length of a derivation in LFI1 , it
is easy to prove that, if  LFI1 , then  LFI1 .
(Completeness) Suppose that  LFI1 and let v be an LFI1 -valuation such that

v[] {1}. Using Lemma 4.4.44, there  1 such that, for
 1 is a valuation h for LFI1
every L , v() = 1 iff h() 1, 2 . From this, h[] 1, 2 and then, by
 
hypothesis, h() 1, 21 . Therefore v() = 1. This shows that  LFI1 and then,
by Theorem 4.4.43,  LFI1 . 

Corollary 4.4.46 The following properties of the consistency operator hold in

LFI1 :
(1) LFI1 ;
(2) LFI1 ;
(3) ( ) ( );
(4) ( )  ( ); 
(5) ( ) ( ) ( ) ;
(6) ( );
(7) ( ).

Proof Items-(1)(4) follow from the results stated in Sect. 3.6 of Chap. 3. The proof
of the other items is immediate from the truth-tables of LFI1 and the completeness
theorem for LFI1 w.r.t. LFI1 . 
4.4 Some 3-Valued LFIs 163

Another presentation for LFI1 (or, equivalently, for J3) will be now briefly
described. It is the 3-valued logic MPT, introduced in [39, 40], as being the proposi-
tional basis of a first-order paraconsistent logic called LPT1. The latter is the logical
counterpart of a Tarskian-style semantics of triples for first-order languages called
pragmatic structures which, by its turn, constitutes a generalization of the theory
of Quasi-Truth introduced in [41] (see also Sect. 9.3 of Chap. 9). As it was proved
in [40], the logic LPT1 coincides with the first-order extension of J3, as well as
with the first-order extension of LFI1 presented in [34]. The first-order extensions
of these systems and other LFIs will be studied in detail in Chap. 7.

Definition 4.4.47 (Propositional logic of Pragmatic Truth MPT, [39, 40]) Let
M P T = M, D be the 3-valued matrix over  P T = {, , } with domain M =
{1, 21 , 0} and set of designated values D = {1, 21 } such that the implication is
defined as in P1 (see Definition 4.4.17), the conjunction is defined as in J3 (see
Sect. 4.4.3) and the negation is also defined as in J3 (see Definition 4.4.11):

The logic MPT is the one associated to the matrix M P T . 

The following connectives of LFI1 can be defined in M P T as follows:

= ( )
def

= ( )
def

=
def

= ( )
def

=
def

= .
def

From this, it is clear that MPT and LFI1 are functionally equivalent. Since
the signature of MPT is simpler than , the axiomatization of MPT is a bit
simpler than that of LFI1 given in Definition 4.4.41. It is worth noting that, in
terms of the operators of LFI1, the implication of MPT can be defined as
= ( ), where = ( ).
def def

Definition 4.4.48 (The calculus LPT for MPT, [39, 40]) Let  P T be the signature
{, , }, and consider the abbreviations above for and in  P T . The Hilbert
calculus LPT over  P T is obtained from mbCci (recall Definition 3.1.7) by adding
164 4 Matrices and Algebraizability

the axioms (ce), (cf) (Recall Sect. 3.5)),8 plus the following:

( ) ( ) (ca )
( ) (cv3 )
   
( ) ( ) ( ) (Ip )

The proof of the following result, which uses the same techniques presented in
this chapter, can be found in [39, 40]:
Theorem 4.4.49 (Soundness and Completeness of LPT w.r.t. MPT) The logic LPT
coincides with MPT, that is, LPT is sound and complete w.r.t. the matrix of MPT:
for every  {} L P T ,  LPT iff  MPT .

Remark 4.4.50 It should be observed that, despite not being part of the signature of
MPT, some basic axioms for the disjunction = ( ) were included
def

in LPT, namely (Ax6), (Ax7) and (Ax8). They are necessary in order to give sense
to the axioms which essentially involve disjunction, namely (Ax9) and (Ax10). Of
course, after proving Theorem 4.4.49, all the properties of the disjunction described
by formulas in L P T must be provable in LPT, provided that is viewed as an
abbreviation.

Finally, an interesting alternative presentation of MPT (and the corresponding

one for its counterpart LPT) will be examined. These logics will be defined over
the signature 1 = {, , , , } (recall Definition 2.1.3), where the 3-valued
operators are the corresponding ones (primitive or definable) in LFI1.

Definition 4.4.51 (Modified Propositional logic of Pragmatic Truth MPT0) Let

M P T 0 = M, D be the 3-valued matrix over 1 = {, , , , } with domain
M = {1, 21 , 0} and set of designated values D = {1, 21 } such that , , and are
defined as in LFI1 (see Definition 4.4.38), while the strong negation coincides
with the one definable in LFI1:

should be clear that, in the axioms involving disjunction and consistency , these operators
8 It

are not the primitive ones from , but the corresponding abbreviations in  P T . Moreover, the
implication symbol of  must be replaced by the corresponding symbol of  P T .
4.4 Some 3-Valued LFIs 165

The logic associated to the matrix M P T 0 will be called MPT0. 

The following connectives of LFI1 can be defined in M P T 0 as follows:

= ( )
def

=
def

and then it follows immediately that MPT0 and LFI1 are functionally equivalent
(whence MPT0, MPT, J3 and 3 are functionally equivalent too).
A sound and complete Hilbert calculus for MPT0, called LPT0, will now be
defined. An interesting feature of LPT0 is that it is an expansion of a Hilbert calculus
for CPL over the signature c = {, , , }, and no properties of the derived
connectives or are described in it.
Definition 4.4.52 (The calculus LPT0 for MPT0) Let 1 be the signature {, , ,
, }. The Hilbert calculus LPT0 over 1 is defined by taking axiom schemas
(Ax1)-(Ax10) from mbC (see Definition 2.1.12) plus the following:
Axiom schemas:

(TND)
 
(exp)
(dneg)
(cf)
(ce)
( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg 1 )
( ) ( ) (neg 2 )
166 4 Matrices and Algebraizability

Inference rule:

(MP)

Observe that axioms (Ax1)(Ax9), (TND) and (exp), plus (MP) constitute an
adequate Hilbert calculus for CPL in the signature c = {, , , } (recall Def-
initions 2.1.3 and 2.4.3). Axiom (dneg) was already considered in the definition of
logic C1s (recall Definition 3.6.11).

Definition 4.4.53 A function v : L1 {0, 1} is a valuation for LPT0, or an LPT0-

valuation, if it satisfies the following clauses:
(vAnd) v( ) = 1 v() = 1 and v() = 1
(vOr) v( ) = 1 v() = 1 or v() = 1
(vImp) v( ) = 1 v() = 0 or v() = 1
(vSNeg) v() = 0 v() = 1
(vNeg) v() = 0 = v() = 1
(vNegE1) v() = 1 = v() = 1
(vNegE2) v() = v()
(vDM ) v(( )) = 1 v() = 1 or v() = 1.
(vDM ) v(( )) = 1 v() = v() = 1.
(vCIp ) v(( )) = 1 v() = v() = 1. 

Let LPT0 be the consequence relation w.r.t. LPT0-valuations. The proof of the
following result is straightforward, by using the usual techniques:
Theorem 4.4.54 (Soundness and Completeness of LPT0 w.r.t. valuations) For
every  {} L1 :

 LPT0  LPT0 .

The logic MPT0 can be now proved to be equivalent to the logic LPT0. The
technique will be analogous to that used for P1, Ciore and LFI1.
Lemma 4.4.55 Let v be an LPT0-valuation. Then there exists  a valuation h for
MPT0 such that, for every L1 , v() = 1 iff h() 1, 21 .

Proof Given a valuation v for LPT0, consider the unique valuation h for MPT0 such
that, for every atomic formula p,

1 iff v( p) = 1, and v( p) = 0
h( p) = 21 iff v( p) = 1, and v( p) = 1

0 iff v( p) = 0.

By induction on the complexity of a formula in L1 , it will be proved that

4.4 Some 3-Valued LFIs 167

1 iff v() = 1, and v() = 0 (a)
h() = 21 iff v() = 1, and v() = 1 (b)

0 iff v() = 0. (c)

If is atomic, the result follows by definition of h.

Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH).
The cases when = , = , = and = , are proved as in
the proof of Lemma 4.4.44. So, it remains to prove the following case:
Case 0: = . Then:
(a) h() = h() = 1 iff h() = 0, iff, by (IH), v() = 0, iff, by (vSNeg) and
(vNegE1), v() = 1 and v() = 0.
(b) h() = 21 , and v() = v(). Thus, the equivalence follows easily.
 
(c) h() = h() = 0 iff h() 1, 21 , iff, by (IH), v() = 1, iff, by (vSNeg),
v() = 0.
This concludes the proof. 

From the previous Lemma, the equivalence between MPT0 and the logic gen-
erated by the Hilbert calculus LPT0 follows easily. That is, the calculus LPT0 is
sound and complete w.r.t. the 3-valued matrix semantics of the logic MPT0:
Theorem 4.4.56 The logic LPT0 coincides with MPT0, that is, LPT0 is sound
and complete w.r.t. the matrix of MPT0: for every  {} L1 ,  LPT0 iff
 MPT0 .

Proof The proof is obvious from the previous results, and is left as an exercise to
the reader. 

A first-order extension of LFI1 will be presented in Chap. 7, together with a

first-order version of all the extensions of mbC introduced in the previous chapters.
Moreover, besides defining, for each of such systems, an adequate semantics of
Tarskian (classical) structures equipped with a suitable paraconsistent valuation over
{0, 1}, an interesting Tarkian-style semantics of triples (adapted from [40]) will also
be presented for the quantified version QLFI1 of LFI1 . As a consequence of our
previous discussion, this semantics of triples is also adequate for first-order J3 and
for LPT1, the quantified version of MPT. The same kind of triplet semantics will
be given for the quantified version of P1.
This chapter investigated a huge number of 3-valued LFIs, in contrast with the
LFIs studied in Chap. 3 which cannot be characterized by finite matrices, as proved
in Sect. 4.2. Among such 3-valued logics, the question of maximality has been also
individualized.
168 4 Matrices and Algebraizability

References

1. Blok, Willem J., and Don Pigozzi. 1989. Algebraizable Logics, vol. 77(396) of Memoirs of the
American Mathematical Society. American Mathematical Society, Providence, RI, USA.
2. Anellis, Irving. 2004. The genesis of the truth-table device. Russell: The Journal of Bertrand
Russell Studies 24: 5570.
3. Wjcicki, Ryszard. 1984. Lectures on propositional calculi. Ossolineum, Wroclaw, Poland.
http://www.ifispan.waw.pl/studialogica/wojcicki/papers.html.
4. Dugundji, James. 1940. Note on a property of matrices for Lewis and Langfords calculi of
propositions. The Journal of Symbolic Logic 5(4): 150151.
5. Chagrov, Alexander V., and Michael Zakharyaschev. 1997. Modal logic, vol. 35. Oxford Logic
Guides Oxford: Oxford University Press.
6. Coniglio, Marcelo E., and Newton M. Peron. 2014. Dugundjis theorem revisited. Logica
Universalis 8(34): 407422. doi:10.1007/s11787-014-0106-4.
7. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Incon-
sistency. In Handbook of Philosophical Logic (2nd. edition), eds. Dov M. Gabbay and Franz
Guenthner, vol. 14, 193. Springer. doi:10.1007/978-1-4020-6324-4_1.
8. Boole, George. 1847. The Mathematical Analysis of Logic, Being an Essay Towards a Calculus
of Deductive Reasoning. Cambridge: Macmillan, Barclay, & Macmillan. Reprinted by Basil
Blackwell, Oxford.
9. Boole, George. 1854. An Investigation of the Laws of Thought on Which are Founded the
Mathematical Theories of Logic and Probabilities. London: Macmillan. Reprinted by Dover,
1958.
10. De Morgan, August. 1847. Formal Logic: Or, the Calculus of Inference, Necessary and Prob-
able. London: Taylor and Walton. Reprinted by The Open Court Company, London, 1926.
11. Jevons, William S. 1864. Pure logic: Or, the logic of quality apart from quantity. London: E.
Stanford. https://archive.org/details/purelogicorlogi00jevogoog.
12. Peirce, Charles S. 1870. Description of a notation for the logic of relatives, resulting from an
amplification of the conceptions of Booles calculus of logic. Memoirs of the American Academy
9: 317378. Reprinted in vol. III of [Hartshorne, Charles, and Paul Weiss, eds. Collected Papers
of Charles Sanders Peirce, Vols. 16. Cambridge: Harvard University Press, 19311935].
13. Peirce, Charles S. 1880. On the algebra of logic. Chapter I: Syllogistic. Chapter II: The logic
of non-relative terms. Chapter III: The logic of relatives. American Journal of Mathematics 3:
1557. Reprinted in vol. III of [Hartshorne, Charles, and Paul Weiss, eds. Collected Papers of
Charles Sanders Peirce, Vols. 16. Cambridge: Harvard University Press, 19311935].
14. Peirce, Charles S. 1885. On the algebra of logic; a contribution to the philosophy of notation.
American Journal of Mathematics 7(2): 180202. Reprinted in vol. III of [Hartshorne, Charles,
and Paul Weiss, eds. Collected Papers of Charles Sanders Peirce, Vols. 16. Cambridge: Har-
vard University Press, 19311935].
15. Schrder, Ernst. 18901910. Vorlesungen ber die Algebra der Logik, Vols. IIII (in German).
Leipzig: B.G. Teubner. Reprints: Chelsea, 1966; Thoemmes Press, 2000.
16. Burris, Stanley. 2013. The algebra of logic tradition. In The Stanford Encyclopedia of Philos-
ophy, ed. Edward N. Zalta. Summer 2013 edition.
17. Czelakowski, Janusz. 2001. Protoalgebraic Logics. Trends in Logic Series, vol. 10. Dordrecht:
Kluwer Academic Publishers.
18. Font, Josep Maria, Ramn Jansana, and Don Pigozzi. 2003. A survey of abstract algebraic
logic. Studia Logica 74(12): 1397.
19. Font, Josep Maria, and Ramn Jansana. 2009. A General Algebraic Semantics for Sentential
Logics. Vol. 7 of Lecture Notes in Logic, 2nd ed. Ithaca, NY, USA: Association for Symbolic
Logic.
20. Font, Josep Maria. 2016. Abstract algebraic logic: An introductory textbook. Vol. 60 of Math-
ematical Logic and Foundations Series. London: College Publications.
21. da Costa, Newton C.A., Jean-Yves Bziau, and Otvio Bueno. 1995. Aspects of paraconsistent
logic. Bulletin of the IGPL 3(4): 597614.
References 169

22. Carnielli, Walter A., and Joo Marcos. A taxonomy of C-systems. In [Carnielli, Walter A.,
Marcelo E. Coniglio, and Itala M. L. DOttaviano, eds. 2002. Paraconsistency: The Logical
Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP
2000), Vol. 228 of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New
York], pp. 194.
23. Halldn, Sren. 1949. The logic of Nonsense. Uppsala: Uppsala Universitets rsskrift.
24. Jaskowski, Stanisaw. 1948. Rachunek zdan dla systemw dedukcyjnych sprzecznych (in Pol-
ish). Studia Societatis Scientiarun TorunesisSectio A I(5): 5777. Translated to English as
A propositional calculus for inconsistent deductive systems. Logic and Logical Philosophy
7: 3556, 1999. Proceedings of the Stanisaw Jaskowskis Memorial Symposium, held in Torun,
Poland, July 1998.
25. Corbaln, Mara I. 2012. Conectivos de Restaurao Local (Local Restoration Connectives, in
Portuguese). Masters thesis, IFCH, State University of Campinas. http://www.bibliotecadigital.
unicamp.br/document/?code=000863780&opt=4&lg=Den_US.
26. Bochvar, Dmitri A. 1938. Ob odnom trechzna c nom iscislenii i ego primenenii k analizu
paradoksov klassiceskogo funkcionalnogo iscislenija (in Russian). Matmaticeskij Sbornik,
46(2): 287308. Translated to English by M. Bergmann 1981 On a Three-valued Logical Cal-
culus and Its Application to the Analysis of the Paradoxes of the Classical Extended Functional
Calculus. History and Philosophy of Logic 2: 87112.
27. Segerberg, Krister. 1965. A contribution to nonsense-logics. Theoria 31(3): 199217.
28. DOttaviano, Itala M.L., Newton C.A. da Costa. 1970. Sur un problme de Jaskowski (in
French). Comptes Rendus de lAcadmie de Sciences de Paris (A-B) 270: 13491353.
29. DOttaviano, Itala M.L. 1982. Sobre uma Teoria de Modelos Trivalente (On a three-valued
model theory, in Portuguese). Ph.D. thesis, IMECC, State University of Campinas, Brazil.
30. DOttaviano, Itala M.L. 1985. The completeness and compactness of a three-valued first-order
logic. Revista Colombiana de Matemticas XIX(12): 7794.
31. DOttaviano, Itala M.L. 1985. The model extension theorems for J3-theories. In Methods in
Mathematical Logic. Proceedings of the 6th Latin American Symposium on Mathematical Logic
held in Caracas, Venezuela, August 16, 1983, Lecture Notes in Mathematics, ed. Carlos A.
Di Prisco, vol. 1130, 157173. Berlin: Springer.
32. DOttaviano, Itala M.L. 1987. Definability and quantifier elimination for J3-theories. Studia
Logica 46(1): 3754.
33. Michael Dunn, J. 1979. A theorem in 3-valued model theory with connections to number theory,
type theory, and relevant logic. Studia Logica 38(2): 149169.
34. Carnielli, Walter A., Joo Marcos, and Sandra de Amo. 2000. Formal inconsistency and evo-
lutionary databases. Logic and Logical Philosophy 8: 115152.
35. de Amo, Sandra, Walter A. Carnielli, and Joo Marcos. 2002. A logical framework for inte-
grating inconsistent information in multiple databases. In Foundations of Information and
Knowledge Systems. Proceedings of the Second International Symposium, FoIKS 2002, held
in Salzau Castle, Germany, February 20-23, 2002, Lecture Notes in Computer Science, eds.
Thomas Eiter and Klaus-Dieter Schewe, vol. 2284, 6784. Berlin: Springer.
36. Batens, Diderik, and Kristof De Clercq. 2004. A rich paraconsistent extension of full positive
logic. Logique et Analyse 185188: 227257.
37. Schtte, Kurt. 1960. Beweistheorie (in German). Berlin: Springer.
38. Batens, Diderik. 1980. Paraconsistent extensional propositional logics. Logique et Analyse
9091: 195234.
39. Silvestrini, Luiz H. 2011. Uma Nova Abordagem Para A Noo De Quase-Verdade (A new
approach to the Notion of Quasi-Truth, in Portuguese). Ph.D. thesis, IFCH, State University
of Campinas, Brazil. http://www.bibliotecadigital.unicamp.br/document/?code=000788964&
opt=4&lg=en_US.
40. Coniglio, Marcelo E., and Luiz H. Silvestrini. 2014. An alternative approach for quasi-truth.
Logic Journal of the IGPL 22(2): 387410. doi:10.1093/ljigpal/jzt026.
41. Mikenberg, Irene, Newton C.A. da Costa, and Rolando Chuaqui. 1986. Pragmatic truth and
approximation to truth. The Journal of Symbolic Logic 51(1): 201221.
170 4 Matrices and Algebraizability

42. Lwe, Benedikt, and Sourav Tarafder. 2015. Generalized algebra-valued models of set theory.
The Review of Symbolic Logic 8(1): 192205. doi:10.1017/S175502031400046X.
43. ukasiewicz, Jan. 1920. O logice trjwartosciowej (in Polish). Ruch Filozoficzny 5: 170171.
Translated to English as On three-valued logic. In Jan ukasiewicz Selected Works, ed.
Ludwik Borkowski. North Holland, 8788, 1990.
44. Tomova, Natalya. 2012. A lattice of implicative extensions of regular Kleenes logics. Reports
on Mathematical Logic 47: 173182.
45. Blok, Willem J., and Don Pigozzi. 2001. Abstract algebraic logic and the deduction theorem.
Preprint. http://www.math.iastate.edu/dpigozzi/papers/aaldedth.pdf.
46. da Costa, Newton C.A. 1974. On the theory of inconsistent formal systems (Lecture delivered
at the first Latin-American Colloquium on Mathematical Logic, held at Santiago, Chile, July
1970). Notre Dame Journal of Formal Logic 15(4): 497510.
47. Sette, Antonio M.A. 1973. On the propositional calculus P1 . Mathematica Japonicae 18(13):
173180.
48. Lewin, Renato A., Irene Mikenberg, and Maria G. Schwarze. 1990. Algebraization of paracon-
sistent logic P1 . The Journal of Non-Classical Logic 7(1/2): 7988. http://www.cle.unicamp.
br/jancl/.
49. Coniglio, Marcelo E., Francesc Esteva, and Llus Godo. 2016. Maximal logics in the lattice of
degree preserving logics of n . To appear.
50. Kleene, Stephen C. 1952. Introduction to Metamathematics. Amsterdam: North-Holland.
51. Asenjo, Florencio G. 1966. A calculus for antinomies. Notre Dame Journal of Formal Logic
16(1): 103105.
52. Priest, Graham. 1979. The logic of paradox. Journal of Philosophical Logic 8(1): 219241.
53. Priest, Graham. 2007. Paraconsistency and Dialetheism. In Handbook of the History of Logic,
(The Many Valued and Nonmonotonic Turn in Logic), eds. Dov M. Gabbay and John H. Woods,
vol. 8, 129204. North Holland. doi:10.1016/S1874-5857(07)80006-9.
54. Priest, Graham. 2006. In contradiction: A study of the transconsistent, 2nd ed. Oxford Univer-
sity Press.
55. Priest, Graham, and Francesco Berto. 2013. Dialetheism. In The Stanford Encyclopedia of
Philosophy, ed. Edward N. Zalta, Summer 2013 edition.
56. Marcos, Joo. 2000. 8K solutions and semi-solutions to a problem of da Costa. Unpublished
draft.
57. Ribeiro, Mrcio M., and Marcelo E. Coniglio. 2012. Contracting logics. In Logic, Language,
Information and Computation. Proceedings of WoLLIC 2012, Buenos Aires, Argentina, Sep-
tember 3-6, 2012, Lecture Notes in Computer Science, eds. Luke Ong and Ruy de Queiroz,
vol. 7456, 268281. Springer. doi:10.1007/978-3-642-32621-9_20.
58. Arieli, Ofer, Arnon Avron, and Anna Zamansky. 2010. Maximally paraconsistent three-valued
logics. In Proceedings of the 12th International Conference on the Principles of Knowledge
Representation and Reasoning (KR 2010), Toronto, Ontario, Canada, May 9-13, 2010, eds.
Fangzhen Lin, Ulrike Sattler, and Miroslaw Truszczynski, 310318. AAAI Press.
59. Carnielli, Walter A., Marcelo E. Coniglio, Rodrigo Podiacki, and Tarcsio Rodrigues. 2014. On
the way to a wider model theory: Completeness theorems for first-order logics of formal incon-
sistency. The Review of Symbolic Logic 7(3): 548578. doi:10.1017/S1755020314000148.
Chapter 5
LFIs Based on Other Logics

In the previous chapters, all the LFIs studied were based on positive classical logic,
CPL+ . The basic system considered was mbC, and several extensions of it were
proposed. Moreover, we have arrived at several 3-valued logics, most of them alge-
braizable in the sense of Blok and Pigozzi.
In this chapter, we will analyze LFIs defined over other logical bases: positive
intuitionistic logic, the four-valued Belnaps logic and some families of fuzzy logics,
as well as positive modal logics.

5.1 LFIs Based on Positive Intuitionistic Logic

In this section, the question of defining LFIs based on positive intuitionistic logic
instead of CPL+ will be discussed. We begin by analyzing some well-known para-
consistent logics based on IPL+ , namely Johanssons minimal logic and Nelsons
logic. Finally, a weaker version of mbC called imbC will we defined, obtained from
the former by changing the positive base CPL+ by IPL+ .
It is important to recall the signatures  = {, , , , }, + = {, , } and
0 = {, , , } from Definition 2.1.3 of Chap. 2. To begin, some subsystems of
intuitionistic logic will be briefly described.

5.1.1 Basic Features of Positive Intuitionistic Logic

Recall the calculus IPL+ for positive intuitionistic logic, defined over the signature
+ , introduced in Definition 3.7.2. Some basic properties of IPL+ will be analyzed
in this section.

Springer International Publishing Switzerland 2016 171

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_5
172 5 LFIs based on Other Logics

Proposition 5.1.1 The calculus IPL+ satisfies the deduction meta-theorem, DMT:

, IPL+ iff  IPL+ .

Proof Analogous to that of Proposition 2.1.14 item (i). 

Another immediate and useful property of IPL+ is the following:

Proposition 5.1.2 The formula is derivable in the calculus IPL+ for every
formula .

Proof As proven in [1], is a theorem of any structural calculus contain-

ing (Ax1) and (Ax2) as theorems and MP as the only inference rule. 

Definition 5.1.3 (Kripke semantics for IPL+ ) A Kripke model for IPL+ is a
triple K = W, , v , where W is a nonempty set (of possible worlds or states),
is a reflexive and transitive relation (that is, a preorder) between states and
v : W L+ {0, 1} is a mapping which satisfies the following clauses (where
sometimes w
w will be written instead of w w
):
(v1) v(w, p) = 1 and w w
implies v(w
, p) = 1 (for p V ar );
(v2) v(w, ) = 1 iff v(w, ) = 1 and v(w, ) = 1;
(v3) v(w, ) = 1 iff v(w, ) = 1 or v(w, ) = 1;
(v4) v(w, ) = 1 iff, for every w
w, v(w
, ) = 0 or v(w
, ) = 1. 

Definition 5.1.4 (Satisfiability in IPL+ w.r.t. Kripke models) A formula L+ is

satisfied by a Kripke model K for IPL+ and by a world w W , denoted by K, w  ,
if v(w, ) = 1. A formula is valid in IPL+ , denoted by IPL+ , if it is satisfied
by every Kripke model K for IPL+ at every w W . 

Definition 5.1.5 (Semantical consequence relation for IPL+ ) Let  {} L+

be a set of formulas. We say that is a semantical consequence of  in IPL+ , denoted
by  IPL+ , if the following holds: for every Kripke model K for IPL+ and for
every state w W ,

K, w  whenever K, w  for every .



The following results, despite being well-known, will be proven here for the
convenience of the reader.

Proposition 5.1.6 Let be a formula in L+ and let K be a Kripke model for IPL+ .
Then, K, w  and w w
implies K, w
 .
5.1 LFIs Based on Positive Intuitionistic Logic 173

Proof Let , K and w such that K, w  , and fix w

w. By induction on the
complexity of , it will be proven that K, w
 . If is a propositional variable
then the result holds by clause (v1) of Definition 5.1.3. Assume that the result holds
for every formula with complexity less than or equal to n, and suppose that has
complexity n + 1.
Case 1: is . Then, by (v2), K, w  implies that K, w  and K, w  .
By induction hypothesis, K, w
 and K, w
 , and so K, w
 by (v2).
Case 2: is . The proof is analogous, but now using (v3).
Case 3: is . Let w

w
. Then, w

w by transitivity, so either K, w


or K, w

 , by (v4). This shows that K, w

 , by (v4). 

Theorem 5.1.7 (Soundness of IPL+ w.r.t. Kripke models) Let  {} L+ be a

set of formulas. Then:

 IPL+ =  IPL+ .

Proof Given a Kripke model K for IPL+ and a state w W , we firstly observe that
if K, w  and K, w  then, for every w
w it holds that: K, w

(by Proposition 5.1.6), and K, w
 whenever K, w
 (by clause (v4) of
Definition 5.1.3). In this way, K, w
 for every w
w; in particular, K, w 
(by reflexivity of ). This shows that satisfiability is preserved by MP. It remains
to prove that every instance of every axiom schema of IPL+ is satisfied by every
Kripke model K for IPL+ at every state w W . To exemplify the technique used for
the proof, we will prove that every instance of axiom schema (Ax2) of the form
      

is satisfied by every Kripke model K for IPL+ at every state w W . Thus, let K be
a Kripke model for IPL+ and let w W . Let w
w such that

v(w
, ( )) = 1. (a)

We want to prove that

v(w
, ( ) ( )) = 1. (b)

By (v4), it is enough to consider w

w
such that

v(w

, ) = 1 (c)

and then prove that

v(w

, ) = 1. (d)
174 5 LFIs based on Other Logics

In order to prove (d) under such assumptions (namely, (a) and (c)), let w

such
that v(w

, ) = 1. By the transitivity of , we infer w

w
and so, by (a),

v(w

, ) = 1. (e)

Since w

, it follows by (c) that v(w

, ) = 1. By reflexivity of and by (e),

it follows that v(w

, ) = 1. This shows that (d) holds, whence K, w

 as
required. 
In order to prove the completeness theorem, some notions are required.
Recall from Definition 2.2.4 that a set of formulas  L+ is a closed theory of
IPL+ if the following holds:  iff  IPL+ . A closed theory  of IPL+ is prime
if it satisfies the following:  iff either  or ;  is non-trivial if
 = L+ . We construct now the canonical model of IPL+ as follows:
+
Definition 5.1.8 Let WcIPL be the set of non-trivial and prime closed theories of
IPL+ . The canonical model for IPL+ is the triple
+  + +
MIPL
c = WcIPL , , vcIPL
+
such that is the binary relation defined on WcIPL given as follows:  
iff
+ +
 
. On the other hand, vcIPL : WcIPL L+ {0, 1} is a mapping such that
+
vcIPL (, ) = 1 iff .

+ +
Proposition 5.1.9 MIPL
is a Kripke model for IPL such that, for every non-trivial
c
+
and prime closed theory  and every formula : MIPL
c ,   iff .
+
Proof Clearly WcIPL is a nonempty set such that is a preorder defined within it.
+
By the very definition, the mapping vcIPL satisfies clause (v1) of Definition 5.1.3.
Clause (v2) holds by (Ax3), (Ax4) and (Ax5). In turn, (v3) is satisfied because the the-
+ +
ories are prime. Concerning (v4), let  WcIPL such that vcIPL (, ) = 1.
+
This means that . Consider 
such that  
and vcIPL (
, ) =
1; then, 
(as  
) and 
, so 
by MP. Therefore,
+
vcIPL (
, ) = 1. Conversely, suppose that  is such that, for every 
,
+ +
vcIPL (
, ) = 1 provided that vcIPL (
, ) = 1. Suppose, by contradiction, that
+
vcIPL (, ) = 0. This means that /  and, by Theorem 2.2.6, there is
+
a 
WcIPL such that 
is maximal non-trivial with respect to in IPL+
and  
; that is, 
.1 Given that 
, IPL+ (by axiom (Ax1) and
MP), it follows that / 
. On the other hand, if
/ 
, then 
, IPL+
and hence 
, IPL+ , by MP. But then, by the deduction meta-theorem DMT
(which holds in IPL+ ), it follows that 
IPL+ , which is a contradiction.
+
Therefore, 
. This shows that there exists 
 such that vcIPL (
, ) = 1

+
1 By adapting the proof of Theorem 2.2.6, it is easy to see that 
is prime, hence 
WcIPL .
5.1 LFIs Based on Positive Intuitionistic Logic 175

+
but vcIPL (
, ) = 0, which contradicts our original assumption about . Therefore,
+ + +
vcIPL (, ) = 1. This shows that vcIPL also satisfies (v4) and MIPL
c is, hence,
+ +
a Kripke model for IPL . Clearly, Mc ,   iff  by the very definition
IPL
+
of vcIPL . 
Theorem 5.1.10 (Completeness of IPL+ w.r.t. Kripke models) Let  {} L+
be a set of formulas. Then:

 IPL+ =  IPL+ .
+
Proof Suppose that  IPL+ and consider the canonical model MIPL c . Let  
maximal relative to in IPL+ , which always exists by Theorem 2.2.6. Clearly, 
+ +
is prime and so  WcIPL is such that, by definition, vcIPL (, ) = 1 for every
+ +
, while vcIPL (, ) = 0. By Proposition 5.1.9, MIPL
c is a Kripke model for
IPL+ and, thus,  IPL+ by Definition 5.1.5. 

5.1.2 Johanssons Minimal Logic

Questioning the non-constructive character of the explosion law, in 1937 (see [2]),
Johansson proposed a subsystem of propositional intuitionistic logic which rejects
that principle. The resulting system, known as Johanssons minimal logic, is such that
its implication-negation fragment coincides with the propositional fragment of the
variant of first-order intuitionistic logic proposed in 1925 by Kolmogorov (see [3]).
When defined over the signature 0 , Johanssons minimal logic can be obtained
from IPL+ by adding one axiom for negation:
Definition 5.1.11 (Johanssons minimal logic Min) The calculus Min defined over
the signature 0 is obtained from IPL+ by adding the following axiom schema:

( ) ( ) (cp)

The following important meta-theorem is immediadely obtained:
Proposition 5.1.12 The calculus Min satisfies the deduction meta-theorem, DMT:

, Min iff  Min .

Remark 5.1.13
(1) The system obtained from Min by adding either the explosion law (
) or the axiom schema ( ) is the intuitionistic propositional logic IPL
defined over the signature 0 (see, for instance, [4]).
(2) It is worth noting that the logic Min is paraconsistent w.r.t. the negation . How-
ever, it is useless as a paraconsistent logic since, as it is well known, any negated
176 5 LFIs based on Other Logics

formula can be derived from a contradiction. Under these circumstances, contradic-

tions trivialize negations, as the next theorem states.
Theorem 5.1.14 Let and be formulas in L0 . Then: , Min and so the
formula ( ) is derivable in Min.
Proof From Min ( ) (obtained by axiom (Ax1)) it follows, by MP,
that Min . However, Min by axiom (cp) and MP.
Therefore, Min by item (iii) of Definition 2.1.4 of Tarkian logics.
Again by MP, , Min . Therefore, using DMT it follows that Min
( ) as required. 
Remark 5.1.15 A logic with such characteristics (namely, the ability to derive from
any contradiction any instance by substitutions of a given schema in the case of Min,
any instance of the schema p1 ) is called partially explosive (see [5, 6]). Otherwise,
a paraconsistent logic is called boldly paraconsistent. By obvious reasons, boldly
paraconsistent systems are more interesting than partially explosive ones, seen as
paraconsistent logics. In Sect. 5.3, a 4-valued modal LFI based on Belnap and Dunns
logic BD will be studied, proving that it is boldly paraconsistent.
The system Min can be semantically characterized by Kripke models augmented
with an upper set contained in the set of possible worlds. In what follows, the def-
initions and proofs of soundness and completeness of Min w.r.t. the (augmented)
Kripke models found in Humberstones vast monograph [7] will be adapted to our
framework.
Definition 5.1.16 (Kripke semantics for Min) A Kripke model for Min is a quadru-
ple K = W, Q, v where W is a nonempty set (of possible worlds, or states); Q is
a subset of W (of contradictory states) such that, if w Q and w w
then w
Q;
is a reflexive and transitive relation between states; and v : W L0 {0, 1} is a
mapping satisfying, besides the clauses (v1)(v4) of Definition 5.1.3, the following:
(v5) v(w, ) = 1 iff, for every w w
, v(w
, ) = 0 or w
Q. 
The notions of satisfiability and validity in Min w.r.t. Kripke models are analo-
gous to those found in Definition 5.1.4 for IPL+ . The same applies to the semantical
consequence relation Min for Min, adapted from the one used for IPL+ in Defini-
tion 5.1.5.
Proposition 5.1.17 Let be a formula in L0 and let K be a Kripke model for Min.
Then, K, w  and w w
implies K, w
 .
Proof The proof is analogous to that for IPL+ in Proposition 5.1.6; that is, by
induction on the complexity of . The only new case to be considered is when
is , given that both semantics coincide in the positive fragment. Then, suppose
that v(w, ) = 1 and let w
w. In order to prove that v(w
, ) = 1, consider
w

w
such that v(w

, ) = 1. Since w

w (by the transitivity of ) then w

Q
by clause (v5) of Definition 5.1.16. Then, v(w
, ) = 1 by (v5). 
5.1 LFIs Based on Positive Intuitionistic Logic 177

Theorem 5.1.18 (Soundness of Min w.r.t. Kripke models) Let  {} L0 be a

set of formulas. Then:

 Min =  Min .

Proof It is an extension of the proof for IPL+ in Theorem 5.1.7 given the fact
that, as long as the positive fragment is concerned, both semantics coincide. So,
all that remains is to prove the validity of axiom (cp) w.r.t. the new models. Let
be an instance ( ) ( ) of axiom (cp). Consider a Kripke model
K for Min, w W , and w
w such that K, w
 . In order to prove that
K, w
 , let w

w
such that K, w

 . Now we must demonstrate that

K, w

 . Thus, consider w

such that K, w

 . Since w

w
then
K, w

 , by hypothesis. Since K, w

 (by Proposition 5.1.17), w

Q
by (v5). Therefore, K, w

 , showing that K, w  . From here, axiom (cp) is

proven to be valid. 

In order to obtain the completeness theorem, some properties of Min must be

established.

Proposition 5.1.19 The following formulas are theorems of Min:

(1) ;
(2) ( ) ( );
(3) (( ) );
(4) ( ( ));
(5) ( ( ));
(6) ( );
(7) ( ) .

Proof (1) By Proposition 5.1.2, is derivable in Min. On the other hand,

( ) ( ) is also a theorem of Min by (cp). The result fol-
lows by MP.
(2) By (cp), Min ( ) ( ). Using item (1) and DMT, it is easy
to prove that Min . From here, the result follows.
(3) Immediate, by using DMT.
(4) Immediate, by items (3), (4) and DMT.
(5) Immediate, by item (4), taking as .
(6) Immediate, by items (5), (1) and DMT.
(7) Immediate, by item (6), (cp), MP and DMT.

178 5 LFIs based on Other Logics

As a matter of fact, from the previous result, it is possible to characterize Min in

an alternative way:

Corollary 5.1.20 The calculus obtained from IPL+ by adding the following axiom
schemas:

( ) ( ) (cp
)
(ce)

coincides with Min.

Proof Let Min

be the calculus obtained by adding axiom schemas (cp
) and (ce) to
IPL+ . Observe that Min
satisfies DMT. By items (1) and (2) of Proposition 5.1.19,
(cp
) and (ce) are derivable in Min and, thusly, Min
is contained in Min. Con-
versely, by (cp
) the formula ( ) ( ) is derivable in Min
.
By (ce) and DMT, Min
, and again using DMT, it subsequently follows
that ( ) ( ) is derivable in Min
. This shows that Min
coincides
with Min. 

Definition 5.1.21 The canonical model for Min is the quadruple

 
MMin
c = WcMin , Q Min
c , , vc
Min

such that:

(i) WcMin is the set of non-trivial and prime closed theories of Min;
(ii) Q Min
c is the set of non-trivial and prime closed contradictory theories of Min,
that is:  Q Min
c iff  WcMin and {, }  for some formula ;
(iii) is the following binary relation defined on WcMin :  
iff  
;
(iv) vcMin : WcMin L0 {0, 1} is the mapping such that

vcMin (, ) = 1 iff .



Proposition 5.1.22 MMin

c is a Kripke model for Min such that, for every non-trivial
closed theory  and every formula : MMinc ,   iff .

Proof Taking into account the fact that the semantics of the positive fragment of
Min coincides with the one for IPL+ , the proof is an extension of that of Proposi-
tion 5.1.9. It follows, consequently, that the mapping vcMin satisfies clauses (v1)(v4)
of Definition 5.1.3 (observe that, in order to prove the satisfaction of (v4), Theo-
rem 2.2.6 is now used for logic Min instead of logic IPL+ ). Concerning clause (v5),
suppose that vcMin (, ) = 1 and let 
 such that vcMin (
, ) = 1. By Def-
inition 5.1.21, {, } 
and so 
Q Min c . Conversely, let  Wc
Min
such
5.1 LFIs Based on Positive Intuitionistic Logic 179

that vcMin (, ) = 0. Then,  Min . Let 0 =  {}. If 0 Min ,

def

then  Min by DMT, and so (by item (7) of Proposition 5.1.19) it fol-
lows that  Min , a contradiction. Therefore, 0 Min . By Theorem 2.2.6,
there exists a theory 
which is maximal non-trivial with respect to in Min
such that 0 
. Since 
is deductively closed and non-trivial, 
WcMin and,
by definition, vcMin (
, ) = 1. Now, suppose that 
Q Min

c . Then,  Min

and  Min for some formula . As , Min (by Theorem 5.1.14),


Min , a contradiction. Therefore, 
/ Q Min
c . This shows that there is some

  in which vc ( , ) = 1 and 
Min
/ Q Min Min
c . Thus, vc satisfies clause (v5).
By the very definition of vc , it is clear that Mc ,   iff .
Min Min

Theorem 5.1.23 (Completeness of Min w.r.t. Kripke models) Let  {} L0
be a set of formulas. Then:

 Min =  Min .

Proof Analogous to the proof of Theorem 5.1.10, but now using the canonical model
MMin
c . 
From Remarks 5.1.13 and 5.1.15, it is clear that the logic Min is not of much inter-
est from a paraconstistentist point of view. That being so, it is not a good candidate
to play the role of a minimal paraconsistent logic with IPL+ as a positive basis.

5.1.3 Nelsons Paraconsistent Logic N4

In his famous paper [8] published in 1949, Nelson observed a non-constructive

feature of both Johanssons logic as well as intuitionistic logic: the truth of a negated
sentence cannot be stated in a direct way, but by a reduction of the negated sentence
to an absurdity. The failure of the contructiveness of these logics is convincingly
exemplified by the following: when a sentence of the form ( ) is provable,
it does not mean that either or is provable, as one could expect within this
framework. Accordingly, Nelson then proposes a system called N which is formed by
first-order intuitionistic logic plus a so-called strong negation,2 denoted by , which
satisfies the properties of a De Morgan negation as well as the following desirable
meta-property:
 ( ) implies  or  .

Later on, in 1959, he proposes one of the first systems of paraconsistent logic, called
S (see [9]), which differs from N in the fact that it is based on positive first-order
intuitionistic logic. A variant of S called N was proposed in 1984 by Almukdad and
Nelson in [10], which became the standard presentation of Nelsons paraconsistent

2 Recall that the expression strong negation is reserved, in this book, for a negation with a Boolean

character.
180 5 LFIs based on Other Logics

logic. In 2003, Odintsov [11] proved that Nelsons paraconsistent logic, rebaptized
as N4, is complete with respect to a class of algebras called N4-lattices. Moreover,
it is algebraizable in the sense of Blok and Pigozzi w.r.t. the variety of N4-lattices.
Additionally, a Kripke semantics for the strong negation of N4 was developed by
Thomason [12], while Odintsov proposed in [13] a useful semantics in terms of
twist-structures, a general framework which was independently proposed by Fidel
in [14] and by Vakarelov in [15]. Modal extensions and generalized twist structures
for N4 as well as related systems were recently introduced by Rivieccio and Ono
(see [1618]).
The rest of this section closely follows the presentation of the system N4 offered
by Odintsov in [19], including his adaptation, for N4, of the semantics proposed
by Fidel in [20] for Nelsons system N , as well as the semantics in terms of twist
structures. A variant of twist structures for LFIs will be introduced in Chap. 6.

Definition 5.1.24 (Nelsons paraconsistent logic N4) The calculus N4 defined over
the signature 0 is obtained from IPL+ by adding the following axiom schemas:

(ce)
(cf)
( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg1 )
( ) ( ) (neg2 )
( ) ( ) (neg 1 )
( ) ( ) (neg 2 )


Remark 5.1.25 Observe that axioms (neg1 ), (neg2 ), (neg1 ), (neg2 ), (neg 1 )
and (neg 2 ) are also included in the Hilbert calculus LFI1 proposed for the
3-valued logic LFI1 in Definition 4.4.41, as well as in the Hilbert calculus LPT0
for the 3-valued logic MPT0 given in Definition 4.4.52 (see Chap. 4). Additionally,
axioms (neg1 ), (neg1 ) and (neg 1 ) were examined in Sect. 3.6 in the study of
propagation of consistency. By adding the explosion law ( ) to N4, N
is obtained, representing Nelsons original system from 1949. On the other hand, by
adding a constant to the signature and the axiom schemas

(bot1 )
(bot2 )

to N4, the system N4 is obtained, which is defined over the signature 2 =

{, , , , }, being a conservative extension of N4 (see [19]). The
5.1 LFIs Based on Positive Intuitionistic Logic 181

intuitionistic negation can be defined in N and N4 as = ( p1 p1 )

def

and = , respectively. It cannot be defined in N4 since this logic cannot

def

define a bottom formula.

As was proven in previous systems, the following important meta-theorem can

immediately be obtained:

Proposition 5.1.26 The calculus N4 satisfies the deduction meta-theorem, DMT:

, N4 iff  N4 .

Recall from Chap. 2, end of Sect. 2.3, that a propositional (Tarskian) logic L
is self-extensional if it satisfies weak replacement: if 1 1 , , n n , then
(1 , . . . , n ) (1 , . . . , n ) for every formula ( p1 , . . . , pn ), where
stands for  L and  L . The logic N4 is not self-extensional. However, as
in the case of J3 and other related systems (recall Sect. 4.4 of Chap. 4), a stronger
notion of provable equivalence can be defined, in order to obtain a congruence and
so satisfy a suitable form of weak replacement:

Proposition 5.1.27 Let s be the following relation between the formulas of L0 :
s iff and . Then s is a logical congruence in N4 (recall
Definitions 4.3.2 and 4.3.3, Chap. 4). Consequently, N4 satisfies the weak replace-
ment property w.r.t. s .

Recall that an implicative lattice is an algebra  A = A, , , , 1 where

A, , , 1 is a lattice with top element 1 such that {c A : a c b} exists 
for every a, b A (here is the partial order associated with the lattice3 and X
denotes the supremum of the set X A w.r.t. , provided that it exists). Additionally,

is a binary operator for implication such that a b = {c A : a c b}
def

for every a, b A.
The following semantics was proposed in 1979 by M. Fidel for the system N
in [20], and was adapted by Odintsov for system N4. It is worth noting that this kind
of semantics was originally proposed by Fidel in 1977 for da Costas systems Cn
in [21], proving for the first time the decidability of such systems.

Definition 5.1.28 (Fidel structures) A Fidel structure for N4, or simply an F-

structure for N4, is a pair  
E = A, {Na }aA

such that A = A, , , , 1 is an implicative lattice and {Na }aA is a family of

nonempty subsets of A such that, for every a, b, c, d A, the following holds:

3 That is: a b iff a = a b iff b = a b.

182 5 LFIs based on Other Logics

(1) if c Na , then a Nc ;
(2) if c Na and d Nb , then c d Nab ;
(3) if c Na and d Nb , then c d Nab ;
(4) if d Nb , then a d Nab .

An F-structure is saturated if Na = A for every a A. 

The intuitive reading of c Na is that c is a possible negation a of a.

 
Definition 5.1.29 A valuation over an F-structure E = A, {Na }aA for N4 is a
mapping v : L0 A which satisfies the following properties:
(1) v( p) Nv( p) , for every p V ar ;
(2) v(#) = v()#v() for # {, , };
(3) v(( )) = v() v();
(4) v(( )) = v() v();
(5) v(( )) = v() v();
(6) v() = v(). 

It should be clear that a valuation is completely determined by its values over the
set V ar { p : p V ar }. By induction on the complexity of , it is easy to prove
the following:

Proposition 5.1.30 Let v be a valuation over an F-structure E. Then v() Nv()

for every formula .

The last result justifies our intuition about the sets Na . The semantical consequence
relation asociated with F-structures is defined as expected:
 
Definition 5.1.31 Let  {} L0 and let E = A, {Na }aA be a Fidel structure.
The semantical consequence relation  |=EF holds iff, for every valuation v over
E, v() = 1 whenever v( ) = 1 for every . Accordingly, we say that is a
semantical consequence of  (w.r.t. Fidel structures), denoted by  |=F , iff  |=EF
for every F-structure E. 

By the very definition, the proof of the following result is immediate:

Proposition 5.1.32 The semantical consequence relation |=F is structural, that is,
given a set of formulas  {} and a substitution over the signature 0 :  |=F
implies [] |=F ().
 
Definition 5.1.33  An F-structure
 E = A, {Na }aA is a substructure of another F-
structure E
= A
, {Na
}aA
provided that A is an implicative sublattice of A
(that
is, a sublattice in which the implication of A
coincides with that of A when restricted
to the domain A A) and Na Na
for every a A. 
5.1 LFIs Based on Positive Intuitionistic Logic 183

Obviously the semantical consequence relation |=EF is contained in |=EF whenever

E is a substructure of E
, as every valuation over the former is also a valuation over
the latter. Additionally, every F-structure E = A, {Na }aA  is a substructure of a
saturated one. Thus, we can consider E
= A, {Na
}aA where Na
= A for every
a A.
In order to prove completeness of N4 w.r.t. Fidel structures, some additional results
are required.
Given a non-trivial theory  in N4 (that is, a theory such that  N4 for some
formula ) let  be the following relation between the formulas in L0 :

 iff  N4 and  N4 .

Clearly,  is an equivalence relation. Moreover, it is a congruence w.r.t. the con-

nectives of the signature + of IPL+ . Let / be the equivalence class of L0
and let L0 / = {/ : L0 } be the set of all the equivalence classes. By the
observation above, the operations

/ # / = (#)/ for # {, , }
def

are well-defined. Therefore

def  
A = L0 / , , , , ( p1 p1 )/

is an implicative lattice. From here, an F-structure is naturally obtained by taking

N/ = ()/ : /
def

for every / L0 / . This structure, called the Lindenbaum F-structure for
N4over , is coherent with the intuitive reading for the sets Na given above. It is not
hard to prove the following:

Proposition 5.1.34 For every non-trivial theory , the pair

 
E = A , {Na }aL0 /

is an F-structure.

We thus arrive at the desired result:

Theorem 5.1.35 (Soundness and completeness of N4 w.r.t. Fidel structures) Let

 {} be a set of formulas such that  is non-trivial in N4. The following conditions
are equivalent:
(1)  N4 ;
(2)  |=F ;
(3)  |=EF .
184 5 LFIs based on Other Logics

Proof (1) (2): This is the Soundness theorem, which can be easily stated as usual
by proving that all the schema axioms of N4 are semantically valid and that the given
semantics preserves trueness through MP.
(2) (3): It is obvious by Definition 5.1.31.
(3) (1): Consider the mapping v : L0 L0 / given by v() = / . Then, v
is a valuation over E by the very definition of A , such that v() = ( p1 p1 )/
iff  N4 , for every formula . Then, v() = ( p1 p1 )/ for every , and,
by hypothesis, it follows that v() = ( p1 p1 )/ . That is,  N4 . 

Finally, another interesting semantics for N4 can be given. It consists of the

so-called twist structures introduced independently in 1977 by Fidel in [14] and
Vakarelov in [15]. This semantics has shown to be extremely versatile, being devel-
oped afterwards for several non-classical logics (see, for instance, [17, 18, 2225]).
In Chap. 6 we will introduce Fidel structures for several LFIs starting form mbC as
well as a semantics of twist-like structures naturally associated to them, called swap
structures.
The following semantics of twist structures for N4, which we will briefly recall
here, is due to Odintsov (see [13, 19]).

Definition 5.1.36 Consider an implicative lattice A = A, , , , 1 .

(1) The full twist structure for N4over A is the algebra over 0

A = A A, , , ,

such that the operations are defined as follows: for every (a, b), (c, d) A A,
(a, b) (c, d) = (a c, b d);
def
(i)
(a, b) (c, d) = (a c, b d);
def
(ii)
(a, b) (c, d) = (a c, a d);
def
(iii)
(a, b) = (b, a).
def
(iv)
(2) A twist structure for N4over A is a subalgebra B of A , such that 1 (|B|) =
{a A : (a, b) |B| for some b A} = A.
(3) Given a twist structure B, a valuation over B is a homomorphism v : L0 B. If
 {} is a set of formulas, the semantical consequence relation  |=B  holds iff,
for every valuation v over B, 1 (v()) = 1 whenever 1 (v( )) = 1 for every .4
We say that is a semantical consequence of  (w.r.t. twist structures), denoted by
 |= , if  |=B for every twist structure B. 

It is worth noting that the reduct to signature {, } of any twist structure B is a

lattice, where the partial order is defined from that of A as follows:

(a, b) (c, d) a c and d b.

4 As in item (2), 1 denotes the canonical projection on the first coordinate.

5.1 LFIs Based on Positive Intuitionistic Logic 185

For every non-trivial theory , recall the F-structure E defined above.

A Lindenbaum twist structure for N4is defined from this as follows:


E = (/ , ()/ ) : L0 and / , , , , .

Then, the following theorem is obtained:

Theorem 5.1.37 (Soundness and completeness of N4 w.r.t. twist structures) Let
 {} be a set of formulas such that  is non-trivial in N4. The following conditions
are equivalent:
(1)  N4 ;
(2)  |= ;
E 
(3)  |= .
An interesting and obvious remark concerning twist structure semantics (in gen-
eral), is that they are a particular case of matrix semantics over a class of logical
matrices (recall Sect. 4.1 of Chap. 4).
In order to see this, let us consider the case for N4. Given an implicative lattice
A and a twist structure B for N4 over A with domain B, let

D B = x B : 1 (x) = 1 .

Consider now the logical matrix MN4 (B) = B, D B . It is clear that the valuations
def

over the matrix MN4 (B) coincide with the valuations over B in the sense of Defini-
tion 5.1.36(3). Moreover, given a set of formulas  {} over the signature 0 ,

 |=MN4 (B) if and only if  |=B

 .

From this, we obtain the following:

Theorem 5.1.38 (Matrix characterization of twist structures semantics) The conse-
quence relation |= associated to twist structures semantics for N4 coincides with
the matrix semantics generated by the class of logical matrices

KN4 = MN4 (B) : B is a twist structure for N4 over A, for some A .

The logic N4 can be now recast as an LFI. In order to define a consistency operator
in N4, we must observe that the very definition of LFIs (by means of a consistency
connective ) requires the existence of finite trivial theories (namely, theories of the
form {, , }). Because of this, the logic to be regarded as an LFI is N4 , the
conservative extension of N4 obtained by adding a bottom (recall Remark 5.1.25).
The consistency operator proposed is given by = ( ), where is the
def

intuitionistic negation defined in N4 as = .

def

The twist structures associated with N4 are obtained from the ones for N4 with
slight modifications:
186 5 LFIs based on Other Logics

Definition 5.1.39 Let A = A, , , , 0, 1 be a Heyting algebra, that is, an

implicative lattice with bottom element 0. The full twist structure over A for N4 is
the following algebra over 2

A
= A A, , , , ,

such that its reduct to 0 is the full twist structure A  defined over the reduct
A = A, , , , 1 of A (recall Definition 5.1.36), and = (0, 1). The twist
def

structures over A for N4 are defined as in Definition 5.1.36. A valuation over a
5

twist structure B for N4 is a homomorphism v : L2 B (and so v() = ). The

B,
semantical consequence relations |= and |= B
 are defined analogously to |= and
|= of Definition 5.1.36, respectively. 6


By using the same partial order used in the case of N4, the reduct to {, } of any
twist structure for N4 is a bounded lattice, with bottom element = (0, 1) and top
element = (1, 0).
For every non-trivial theory , the Lindenbaum twist structure for N4 at  is

the expansion E, of E to 2 such that = (/ , ()/ ).
def

Theorem 5.1.40 (Soundness and completeness of N4 w.r.t. twist structures) Let

 {} be a set of formulas of L2 such that  is non-trivial in N4 . The following
conditions are equivalent:
(1)  N4 ;
(2)  |=
 ;
B, 
(3)  |= for B = E, .

Proof (1) (2): It is the Soundness theorem, which it is easy to prove as usual.
The only new axioms to be checked are (bot1 ) and (bot1 ), which are both obviously
valid.
(2) (3): It is obvious.
(3) (1): Consider the Lindenbaum twist structure B = E, 
for N4 at .
Let v : L2 |B| such that v() = (/ , ()/ ) for every formula . It is
routine to check that v is a homomorphism of Heyting algebras over 2 . Addi-
tionally, 1 (v()) = ( p1 p1 )/ iff  N4 . It thus follows that 1 (v( )) =
( p1 p1 )/ for every . By hypothesis, 1 (v()) = ( p1 p1 )/ and so
 N4 . 

Theorem 5.1.41 Let ( p) = { p}, where p = ( p p) for a propositional

def

variable p and = for every . Then, the logic N4 is a strong LFI

def

(w.r.t. and ( p)), in accordance with Definition 2.1.9 of Chap. 2.

5 Notethat and belong to the domain of any twist structure and are the bottom and top
elements of A A, respectively.
6 A matrix semantics can be alternatively defined, by extending Theorem 5.1.38 to this context.
5.1 LFIs Based on Positive Intuitionistic Logic 187

Proof Let p and q be two different propositional variables. Let A be any Heyting

algebra where 0 = 1 and consider the full twist structure A for N4 over A. By

considering a valuation v1 over A , such that v1 ( p) = (1, 1) and v1 (q) = (0, 1),
it follows that p, p |=  q. By soundness, p, p N4 q and clause (i.a) of
Definition 2.1.9 is consequently satisfied. Considering now a valuation v2 over the
same algebra, such that v2 ( p) = (1, 0) and v2 (q) = (0, 1) it follows that v2 ( p) =
(1, 0), showing that p, p |= q. By soundness, p, p N4 q and, then, clause (i.b)
of Definition 2.1.9 is satisfied. By considering now a valuation v3 , such that
v3 ( p) = v3 (q) = (0, 1), it follows as above that p, p N4 q and clause (i.c) of
Definition 2.1.9 is thusly satisfied. Finally, if v is any valuation over A such that
1 (v()) = 1 (v()) = 1 then v() = v() = (1, 1) and so v() = (1, 1)
(0, 1) = (0, 1). Therefore, 1 (v()) = 1, showing that the set {, , } is unsat-
isfiable over a non-trivial twist structure for N4 . That is, clause (ii) of Definition 2.1.9
is satisfied. This shows that N4 is an LFI w.r.t. and . 
The proof of the last result shows that the operator
= ( ) is not
def

a consistency operator for N4 : the set { p, p,

p} is satisfiable over any twist-
structure by taking v( p) = (1, 1). Thus clause (iii) of Definition 2.1.7 is violated.
An interesting question concerning N4 is how to define a primitive consistency
operator satisfying the conditions of Definition 2.1.7. This problem was studied for
the case of extensions of the fuzzy logic called Monoidal t-norm based logic (MTL) as
well as their associated algebraic structures, i.e. suitable extensions of MTL-algebras.
A brief account of this approach will be given in Sect. 5.2.

5.1.4 An Intuitionistic Version of mbC

As it was shown in the previous section, Nelsons logic N4 is a paraconsistent logic

based on IPL+ which is attained by adding a constructive negation . By extending
that with a bottom formula, a conservative extension N4 is obtained, which is in fact
an LFI whose positive basis is IPL+ . It is arguable whether N4 or N4 are minimal
paraconsistent logics (or LFIs) based on IPL+ since has several (constructive)
features. Defining an LFI based on IPL+ , by adding a negation with minimal
features to be a constructive negation remains as an open problem. A good candidate
would be an extension of IPL+ to signature  by adding the schema axioms

( ) ( ) (cp
)
  
(bc1)

Observe that, besides (bc1), the unique property required for the negation is
axiom (cp
) defining contraposition. That is, the property (ce) which together with
the latter characterizes Johanssons minimal logic Min (see Corollary 5.1.20), was
eliminated to avoid the uninteresting (from the point of view of paraconsistency)
features of the logic Min.
188 5 LFIs based on Other Logics

In this section, we will adopt a different perspective. Instead of extending IPL+

with a constructive negation, a negation exclusively satisfying tertium non datur
(plus, obviously, the gentle principle of explosion) will be added to IPL+ . The
resulting logic imbC, which is a weaker version of mbC, can be obtained from
by removing axiom (Ax9). In other words, this new logic is based on IPL+ instead
of CPL+ , and this is why it could be regarded as an intuitionistic version of mbC.
Observe that, while it is possible to define a bottom in imbC, cannot be
equivalent to , because the latter characterizes an explosive negation.
Definition 5.1.42 (Intuitionistic imbC) The calculus imbC defined over the signa-
ture  is obtained from IPL+ by adding the following axiom schemas:

(Ax10)
  
(bc1)

Equivalently, imbC is obtained from mbC by removing axiom schema (Ax9).

Since imbC is weaker than mbC, but contains axiom schema (bc1), it follows
that imbC is a strong LFI (recall Definition 2.1.9).
Definition 5.1.43 (Kripke semantics for imbC) A Kripke model for imbC is a triple
K = W, , v where W is a nonempty set (of possible worlds, or states), is a
reflexive and transitive relation between states, and v : W L {0, 1} is a map-
ping satisfying, besides (v1)(v4) of Definition 5.1.3, the following clauses:
(v5) v(w, ) = 1 and w w
implies v(w
, ) = 1;
(v6) v(w, ) = 1 and w w
implies v(w
, ) = 1;
(v7) v(w, ) = 0 implies v(w, ) = 1;
(v8) v(w, ) = 1 implies v(w, ) = 0 or v(w, ) = 0.

The notions of satisfiability and validity in imbC w.r.t. Kripke models are anal-
ogous to those in Definition 5.1.4 for IPL+ . The same applies to the semantical
consequence relation imbC for imbC, adapted from the one used for IPL+ in Defi-
nition 5.1.5.
Proposition 5.1.44 Let be a formula in L and let K be a Kripke model for imbC.
Then, K, w  and w w
implies K, w
 .
Proof The proof is analogous to that of Proposition 5.1.6 for IPL+ , that is, by
induction on the complexity of . The only new cases to be considered are when is
or . However, in these cases, the result follows immediately from clauses (v5)
and (v6) of Definition 5.1.43, respectively. 
Theorem 5.1.45 (Soundness of imbC w.r.t. Kripke models) Let  {} L be
a set of formulas. Then:

 imbC =  imbC .
5.1 LFIs Based on Positive Intuitionistic Logic 189

Proof It is an extension of the proof of Theorem 5.1.7, now taking into account the
new axioms and the new corresponding semantical clauses. Thus, let us prove that
the new axioms are valid w.r.t. the new models. Let K be a Kripke model for imbC
and let w W and w
w.
Axiom (Ax10): Let be an instance of axiom (Ax10). It is clear that
K, w  by (v7).
Axiom (bc1): Let be an instance ( ( )) of axiom (bc1) and
assume that v(w
, ) = 1. In order to prove that v(w
, ( )) = 1, let
w

w
and suppose that v(w

, ) = 1. Since we want to prove that

v(w

, ) = 1, let w

; by transitivity, w

w
. Under these assump-
tions, v(w

, ) = 1 by (v6). Then, v(w

, ) = 0 or v(w

, ) = 0 by clause (v8).
However, v(w

, ) = 1 and so v(w

, ) = 1, by Proposition 5.1.44. From this,

v(w

, ) = 0, and so either v(w

, ) = 0 or v(w

, ) = 1. Thus, v(w

,
) = 1 by (v4), concluding the proof. 

Definition 5.1.46 The canonical model for imbC is the triple

 
MimbC
c = WcimbC , , vcimbC

such that:
(i) WcimbC is the set of non-trivial and prime closed theories of imbC;
(ii) is the following binary relation defined on WcimbC :  
iff  
;
(iii) vcimbC : WcimbC L {0, 1} is the mapping such that

vcimbC (, ) = 1 iff .



Proposition 5.1.47 MimbCc is a Kripke model for imbC such that, for every non-
trivial closed theory  and every formula , MimbC
c ,   iff .

Proof The proof is an extension of that of Proposition 5.1.9, and so it holds that the
mapping vcimbC satisfies clauses (v1)(v4) of Definition 5.1.3 (observe that, in order
to prove the satisfaction of (v4), Theorem 2.2.6 is now used for logic imbC instead
of logic IPL+ ). Concerning clauses (v5) and (v6), the proof is immediate by the very
definitions. Clause (v7) follows easily by axiom (Ax10). Finally, suppose by con-
tradiction that there is some  WcimbC such that vcimbC (, ) = vcimbC (, ) =
vcimbC (, ) = 1. Then, by the validity of axiom (bc1), vcimbC (, ) = 1 for every
formula . That is,  for every formula and  becomes trivial, a contradiction.
Therefore, vcimbC (, ) = 1 implies that vcimbC (, ) = 0 or vcimbC (, ) = 0,
showing that vcimbC satisfies clause (v8). Therefore MimbC
c is a Kripke model for
imbC. Clearly, MimbCc ,   iff  by the very definition of vcimbC . 
190 5 LFIs based on Other Logics

Theorem 5.1.48 (Completeness of imbC w.r.t. Kripke models) Let  {} L

be a set of formulas. Then:

 imbC =  imbC .

Proof Identical to the proof of Theorem 5.1.10 but now using the canonical model
MimbC
c . 

Remark 5.1.49 It is worth noting that the logic imbC is paraconsistent w.r.t. the
negation . Moreover, it is a (strong) LFI with respect to and . This can be easily
checked using Kripke models.

Finally, it is easy to see that it is possible to encode CPL within imbC, as happens
with mbC, by means of a Derivability Adjustment Theorem (or DAT). First, observe
the following:

Proposition 5.1.50 In imbC, the following holds:

(1) imbC p ( p q), where p and q are different propositional variables.
(2) imbC ( ) for every and .

Proof (1) Consider the following Kripke model for imbC: K = W, , v such that
W = {w, w
} where w w
but w
 w; v(w, p) = v(w, q) = 0, v(w
, p) = 1 and
v(w
, q) = 0. Then, v(w, p q) = v(w, p) = 0 and so v(w, p ( p q)) = 0.
Therefore, imbC p ( p q) and then imbC p ( p q) by soundness.
(2) Clearly, , imbC ( ). On the other hand, , , imbC and so
, imbC , by DMT. From here, , imbC ( ). Therefore,
, imbC ( ). The result follows by axiom (Ax10). 

From here, the following DAT can be easily derived, by adapting the proof of
Theorem 2.4.7:

Theorem 5.1.51 Let t

: Lc L0 be the mapping which replaces with . Then
the following holds: for every  {} Lc ,

 CPL iff , t
[] imbC t
() for some  L0 ,

where  = { : }.

The logic imbC opens interesting possibilities for the study of LFIs: in particular,
all the developments for the logic mbC and its extensions described in the previous
chapters can be studied in the context of the logic imbC, taking into account that
axiom (Ax9) is not derivable, in this case. However, several results concerning mbC
are surely no longer valid for imbC, and stronger assumptions should be required.
5.2 LFIs Based on Fuzzy Logics 191

5.2 LFIs Based on Fuzzy Logics

The famous book [26] by Hjek constitutes the first monograph appearing in the
literature which is integrally dedicated to a broad study of a new subject called
Mathematical Fuzzy Logic (MFL), an approach to vagueness from a point of view
based on many-valued (including infinitely-valued) logics. In this book, the so-called
Basic fuzzy logic BL is introduced as being a residuated many-valued logic, which
has semantics on the real unit interval induced by all the continuous t-norms and
their residua. This logic constitutes a generalization of three prominent fuzzy logics,
namely ukasiewicz, Gdel-Dummet and Product logics.
In turn, a so-called Monoidal t-norm based logic MTL was introduced by Esteva
and Godo in [27] as a generalization of BL, which captures the semantics induced by
left continuous t-norms and their residua. Based on extensions of this logic (the most
general residuated fuzzy logic whose semantics is based on t-norms), several new
LFIs were developed in [28]. In that paper, from which most of the definitions and
results of this section were taken, a study of several consistency and inconsistency
operators (primitive or not) defined over extensions of MTL-algebras was devel-
oped. This constitutes the starting point for a foundational study of paraconsistent
fuzzy logic, providing interesting mathematical models for the novel notion of fuzzy
(in)consistency operators, generalizing a first approach proposed in [29]. We refer
only readers interested in the subject of MFL to the excellent reference [30].

5.2.1 Preliminaries on MFL

We begin by recalling some basic notions of MFL (see, for instance, [26] and [30]).

Definition 5.2.1 Let [0, 1] be the usual real interval [0, 1] = x R : 0 x 1 .

A t-norm or triangular norm is a mapping : [0, 1] [0, 1] [0, 1] satisfying the
following properties for every x, x
, y, y
, z [0, 1]:
(1) x y = y x (commutativity);
(2) x (y z) = (x y) z (associativity);
(3) if x x
and y y
, then x y x
y
(monotonicity);
(4) x 1 = x (unity).

A t-norm is continuous if it is a continuous map where the usual topologies are

considered in  [0, 1] and [0, 1] [0, 1]. A t-norm is left-continuous if, for every
x, a [0, 1], ya (x y) = x a and ya (y x) = a x. This is equivalent to
require that is left continuous in either argument w.r.t. the usual topologies in [0, 1]
and [0, 1] [0, 1]. 
192 5 LFIs based on Other Logics

Clearly, if is a t-norm then x 0 = 0 for every x [0, 1]. In fact, x 1 and

0 0, so x 0 1 0 = 0 by (3), (1) and (4).
Any continuous t-norm is left-continuous, but the converse is not true: there are
examples of left-continuous t-norms which are not continuous. The first of such
examples was given by Fodor in [31].
Continuity and left-continuity are remarkable properties of t-norms, since each
of them allows us to define an implication in a natural way:

Theorem 5.2.2 Let be a continuous (a left-continuous, respectively) t-norm. Then

there is a unique binary operator : [0, 1] [0, 1] [0, 1] such that, for every
x, y, z [0, 1],
z x y iff z x y.

The operator is called the residuum of the t-norm .

Clearly, x y = max z [0, 1] : x z y . Additionally, x y = 1 iff

x y; 0 y = 1 and 1 y = y.
We arrive at the following definition:

Definition 5.2.3 A continuous t-norm algebra (a left-continuous t-norm algebra,

respectively) is an structure [0, 1], min, max, , ,1 where is a continuous (a
left-continuous, respectively) t-norm and is its residuum. Let BL (MTL, respec-
tively) be the variety generated by continuous (by left-continuous, respectively) t-
norm algebras. 

Recall that a commutative monoid is an structure A, &, 1 such that & is an
associative and commutative binary operation and x & 1 = x for every x A. The
structures of Definition 5.2.3 can be generalized as follows:

Definition 5.2.4 (i) A (commutative) residuated lattice is an algebra

A, , , &, , 1

such that:
(1) A, &, 1 is a commutative monoid;
(2) A, , , 1 is a lattice with top element 1;
(3) is a binary operator (called residuum of &) satisfying the so-called residu-
ation property: for every x, y, z A,

z & x y iff z x y.

(ii) A residuated lattice is divisible if it satisfies the following: for every x, y A,

x y = x & (x y).
5.2 LFIs Based on Fuzzy Logics 193

(iii) A residuated lattice is prelinear if it satisfies the following: for every x, y A,

(x y) (y x) = 1.

(iv) A (commutative) bounded residuated lattice is an algebra

A, , , &, , 0, 1

such that A, , , &, , 1 is a residuated lattice and 0 is a bottom element of the
lattice A, , , 1 .
(v) A BL-algebra is a bounded divisible prelinear residuated lattice.
(vi) An MTL-algebra is a bounded prelinear residuated lattice. 

The next fundamental result relates the BL-algebras with the variety BL and the
MTL-algebras with the variety MTL.

Theorem 5.2.5 (1) The class of BL-algebras coincides with the variety BL.
(2) The class of MTL-algebras coincides with the variety MTL.

Hjek introduced a propositional logic (by means of a Hilbert calculus) called

BL, proving that this logic is strongly sound and complete with respect to the (truth-
preserving) semantics in the variety BL, provided that the set of premises are finite
(see [26]).
Generalizing this result, Esteva and Godo introduced a Hilbert calculus called
MTL in [27] such that the resulting logic is algebraizable in the sense of Blok
and Pigozzi (recall Definition 4.3.1), where its equivalent algebraic semantics is the
variety MTL of MTL-algebras. The definition of the calculus MTL is as follows:

Definition 5.2.6 (Monoidal t-norm based logic MTL) Consider the signature & =
{, &, , }. The logic MTL is defined over the signature & by means of the
following Hilbert calculus:

Axiom schemas:

( ) (( ) ( )) (A1)
( & ) (A2)
( & ) ( & ) (A3)
( ) (A4)
( ) ( ) (A5)
( & ( )) ( ) (A6)
( ( )) (( & ) ) (A7a)
(( & ) ) ( ( )) (A7b)
(( ) ) ((( ) ) ) (A8)
(A9)
194 5 LFIs based on Other Logics

Inference rule:

(MP)



The following connectives can be defined from & : stands for ;

stands for (( ) ) (( ) ); and  stands for .

Remark 5.2.7 As expected, the logic BL is obtained from MTL by adding the fol-
lowing axiom schema:
( ) ( & ( )) (Div)

which is the converse of axiom (A6).

Definition 5.2.8 Let A = A, , , &, , 0, 1 be an MTL-algebra. A valuation

over A is a mapping v : L& A such that, for every formula and :
(val1 ) v() = 0;
(val2 ) v(#) = v()#v(), for # {, &, }. 

It is easy to prove that v() = 1 and that v( ) = v() v() for every , .

Definition 5.2.9 (1) Let A be an MTL-algebra. The semantical consequence relation

|=A A
MTL is defined as follows: given a set of formulas  {} L& ,  |=MTL iff,
for every valuation v over A, if v( ) = 1 for every  then v() = 1.
(2) Given a set of formulas  {} L& , we say that is a consequence of  in
MTL w.r.t. the truth-preserving semantics, denoted by  |=MTL , if  |=A
MTL for
every MTL-algebra A. 

As a consequence of the algebraizability of MTL w.r.t. MTL, the following holds:

Theorem 5.2.10 (Soundness and completeness of MTL w.r.t. MTL-algebras) Let

 {} L& be a set of formulas. Then:

 MTL  |=MTL .

From the last result, the following holds:

Corollary 5.2.11 Let  {} L& be a set of formulas. Then:

 |=MTL 0 |=MTL for some finite 0 .

Proof It is a direct consequence of Theorem 5.2.10 and the finitariness of the logic
MTL. 
5.2 LFIs Based on Fuzzy Logics 195

Theorem 5.2.10, together with Item (2) of Theorem 5.2.5, states that MTL is
the logic of left-continuous t-norms, in the same way BL is the logic of continuous
t-norms.
The algebraizability of MTL w.r.t. MTL can be generalized to the so-called core
fuzzy logics, which are axiomatic extensions of MTL satisfying the congruence prop-
erty.
In what follows, will stand for ( ) ( ).
Definition 5.2.12 Let &
be a signature extending & . A core fuzzy logic is any
logic L defined by means of a Hilbert calculus over &
which is obtained from
MTL by adding extra schema axioms so that the following property holds:

i i : 1 i n L c(1 , . . . , n ) c(1 , . . . , n ) (Cong)

for any connective c &

& with arity n, with n 1. 
Remark 5.2.13 Because of the congruence property, any core fuzzy logic is also
algebraizable, and its equivalent algebraic semantics, the class L of L -algebras, is
a variety. As a consequence, the strong completeness Theorem 5.2.10 also holds for
any core fuzzy logic L w.r.t. the corresponding L -algebras. Moreover, any core
fuzzy logic L is strongly complete with respect to the class of L -chains.7 This
means that core fuzzy logics are semilinear.
It is possible to preserve semilinearity in extensions of core fuzzy logics which
have new (finitary) inference rules. Indeed, it is known that an extension L of a core
fuzzy logic is semilinear iff for each new finitary inference rule
1 . . . n
(R)

its corresponding -form
1 . . . n
(R )

is also derivable in L , where is an arbitrary formula.

The logic MTL (as well as the logic BL) does not satisfy the deduction meta-
theorem DMT. However, a weak form of DMT, called local DMT, holds. For every
formula , let 0 =  and, for n 0, n+1 = & n .
def def

Theorem 5.2.14 Let  {, } L& . Then:

, MTL iff  MTL n for some n 0.

The fact that n depends on each specific deduction justifies the name local DMT.
It is worth noting that any core fuzzy logic also satisfies local DMT.

7 An L -chain is a subalgebra of L which is linearly ordered, that is: any two elements of the algebra
are comparable w.r.t. the order .
196 5 LFIs based on Other Logics

It should be noticed that, according to the given definitions, core fuzzy logics
(including MTL) are not paraconsistent: for every formula , the set {, } is unsat-
isfiable, in accordance with Definition 5.2.8 and the definition of . Accordingly,
, L for every and , according to Definition 5.2.9. However, another con-
sequence relation can be discerned for these logics, based on the idea of considering
degrees of truth. This notion was originally proposed by Wjcicki for ukasiewicz
logics (see [32], 4.3.14, p. 288) and afterwards was generalized to other varieties of
residuated lattices in [33].

Definition 5.2.15 ([33]) Given a core fuzzy logic L , the logic L associated with
it is defined as follows:  |=L iff for every L -chain A, every a A, and every
valuation v over A, if a v( ) for every , then a v(). The logic L is a
logic preserving degrees of truth, i.e. the degree-preserving companion of L . 
n
If  = {1 , . . . , n } is finite, then  |=L iff i=1 v(i ) v(), for any valu-
ation v over any L -chain A.

Proposition 5.2.16 ([33]) Let L be a core fuzzy logic (and so MPis the unique infer-
ence rule). Then the logic L can be axiomatized by the Hilbert calculus obtained
from the axiom schemas of L by adding the following inference rules:

(Adj)

L
(MP-r )

Observe that, if the set of theorems of L is decidable, then the above constitutes
a recursive Hilbert-style axiomatization of L .
The last definition can be generalized as follows:

Definition 5.2.17 Let L be a core fuzzy logic with a set of new inference rules. Then
L is the Hilbert calculus formed by all the axiom schemas of L , the two inference
rules (Adj-) and (MP-r ) of Proposition 5.2.16 and the following restricted rules:
for each new inference rule 1 . . . n
(R)

of L (that is, different to MP), add the following inference rule to L :

L 1 . . . L n
(R-r )



Defining L in terms of L , the following result follows easily:

5.2 LFIs Based on Fuzzy Logics 197

Proposition 5.2.18 Let 1 , . . . , n , be a set of formulas of the language of a

core fuzzy logic which has possibly been extended with new inference rules. Then:

1 , . . . , n L L (1 n ) .

In particular, L and L have the same theorems.

Corollary 5.2.19 Let L be as above. If L

is a conservative extension of L , then
L
is also a conservative extension of L .

From the definitions and results above, the manner in which a core fuzzy logic
L (possibly extended with new inference rules) could be transformed into an LFI
is clear: firstly, the expansion L
, obtained by adding a consistency operator (or,
dually, an inconsistency operator ), together with suitable axioms and some possible
new inference rules, is analyzed. In order to ensure the preservation of algebraizability
w.r.t. the corresponding expanded algebras, while also preserving semilinearity, the
congruence property (Cong) (see Definition 5.2.12) must be guaranteed in L
with
respect to and , in accordance with Remark 5.2.13. Moreover, if L adds new
inference rules to MTL, the new logic L
must substitute each new rule (R) of L with
the corresponding rule (R ) in order to guarantee semilinearity (see Remark 5.2.13).
From this, the desired LFI is obtained by considering the logic L
associated with
L
as stated in Definition 5.2.17.

Remark 5.2.20 It should be clear that this process only works in non-SMTL-
extensions of MTL. Recall that the Strict monoidal t-norm based logic (SMTL)
is the axiomatic extension of MTL obtained by adding the axiom schema

( ) (Pseudo)

Recalling that = and taking into account Proposition 5.2.18, it is

def

clear that the degree-preserving companion of any extension of SMTL is not para-
consistent w.r.t. , and so these logics lie outside the scope of the present approach
to paraconsistent fuzzy logics.

From now on, due to Remark 5.2.20, core fuzzy logic will stand for non-SMTL
core fuzzy logic.

5.2.2 Fuzzy Logics with a Consistency Operator

Within the framework described in the previous subsection, a consistency opera-

tor defined in an L -algebra, where L is a core fuzzy logic (possibly containing
additional inference rules), should be an unary operator satisfying the following
conditions in any L -chain A: for some x A,
198 5 LFIs based on Other Logics

(i) x (x) = 0;
(ii) x (x) = 0; but
(iii) y y (y) = 0 for every y A.

The operator can be regarded as the algebraic counterpart of a consistency

operator in the strong sense of Definition 2.1.9. According to [28], from which
all the definitions and results of the rest of this section were taken, the intended
meaning of the truth-value (x) is the (fuzzy) degree of classicality (or reliability,
or robustness) of x with respect to the satisfaction of the explosion law x x = 0.
Let A and as above and let N (A) = {x A {1} : x = 0}. Clearly, either

N (A) = or N (A) [a, 1), (a, 1) where a = N (A).8 In this way, condition
(iii) implies (x) = 0 for x (0, 1), such that x / N (A). Therefore, any must
satisfy the following conditions in any L -chain A:

(x) > 0, if x {0, 1}
(x) = 0, if x (0, 1) N (A)

From here, the following postulates for such a consistency operator in non-SMTL
chains A are proposed:
(c1) If x x = 0, then (x) = 0;
(c2) If x {0, 1}, then (x) = 1;
(c3) If x = 0 and x y then, (x) (y).
Clause (c1) guarantees the validity of postulate (iii). In the classical (2-valued)
case, both truth-values 0 and 1 satisfy the explosion law x x = 0 and, then,
(x) = 1 for every truth-value x. Clause (c2) is justified by the fact that intends to
extend the classical case; another justification for (c2) is that 0 and 1 are classical
truth-values with fuzzy degree 1. Moreover, clause (c2) guarantees the satisfaction
of the aforementioned postulates (i) and (ii) for consistency operators. Clause (c3)
guarantees the coherence of : in N (A), the segment of the chain A where is
positive, the consistency operator is monotonic. This supports the idea that (x)
is the fuzzy degree of classicality, from the perspective of the explosion law, and so
the closer to 1 is x, the more classical it is.
This leads us to the following definition:

Definition 5.2.21 Let L be any non-SMTL extension of logic MTL defined by

a Hilbert calculus over a signature &
containing & . The logic L is the Hilbert

calculus over the signature & = &

{ } obtained from the calculus L by adding
the following:

def
8 Asusual in any partially ordered set L, [a, b) = {x L : a x < b}, where x < y denotes
x y and y  x. Analogously are defined the sets (a, b), (a, b] and [a, b].
5.2 LFIs Based on Fuzzy Logics 199

Axiom schemas:

( ) (AxC1)
 (AxC2a )
(AxC2b )

Inference rules:
( ) ( ( ))
(Cong) (Coh)
( ) ( )

The inference rule (Cong) guarantees the satisfaction of the property Cong and
so, as pointed out in the previous subsection, L is algebraizable in the sense of Blok
and Pigozzi, and its algebraic semantics is given by the L -algebras, defined below:

Definition 5.2.22 Let L as in Definition 5.2.21. An L -algebra is an algebra A

over the signature & , such that its reduct to &

is an L -algebra and the operation
: A A satisfies the following conditions: for all x, y, z A,
(1) x x (x) = 0
(2) (1) = (0) = 1
(3) if (x (x y)) z = 1, then ((x) (y)) z = 1 

Because of this, L is sound and strongly complete with respect to the quasi-
variety L of L -algebras. Let |=L be the consequence relation w.r.t. L -algebras.
Since the inference rules (Cong) and (Coh) are closed under -forms, we know that
L is also semilinear and, hence, is strongly complete with respect to the class of
L -chains:

Proposition 5.2.23 (Chain completeness) The logic L is strongly complete with

respect to the class of L -chains.

Conditions (1)(3) on in a linearly ordered L -algebra faithfully capture the

three intended properties (c1)(c3) that were required for such a operator at the
beginning of this section. In fact, the following result can be proven.

Proposition 5.2.24 Let A be an L -chain and let : A A be a mapping. Then

satisfies conditions (c1), (c2) and (c3) iff A expanded with the operator is an
L -chain.

Proof Firstly, observe that in any chain the following holds: (a) x y = 0 iff
x = 0 or y = 0; (b) x y = 1 iff either x = 1 or y = 1; (c) x y = 1 iff x y;
(d) x = 1 iff x = 0.
200 5 LFIs based on Other Logics

From left to right: Observe that (2) is identical to (c2). Suppose that satisfies (c1)
and let x A. If x x = 0, then x x (x) = 0. If x x = 0, then (x) = 0
by (c1) and, again, x x (x) = 0. Thus (1) holds. Finally, assume that sat-
isfies (c3) and let x, y, z A such that (x (x y)) z = 1. By item (b) (see
the beginning of the proof), either x (x y) = 1 or z = 1. If z = 1 then obvi-
ously ((x) (y)) z = 1. Otherwise, both x = 1 and (x y) = 1. By (c)
and (d), x = 0 and x y. So, by (c3), (x) (y). By (c), (x) (y) = 1 and
so ((x) (y)) z = 1, showing that (3) holds.
From right to left: Assume (1) and let x A such that x x = 0. By (1),
x x (x) = 0 and so, by (a), (x)=0. This shows that (c1) holds. Now,
assuming (3), suppose that x = 0 and x y. By (c) and (d), x = 1 and
(x y) = 1. Then x (x y) = 1. For z = 0 and using (3), it follows that
((x) (y)) 0 = 1. Hence, by (c), (x) (y), showing that (c3) holds. 
Examples 5.2.25 In [28], two paradigmatic examples are given. The first consists
of the logic of a t-norm which is an ordinal sum of a ukasiewicz component and
a Gdel component with an idempotent separating point 0 < a < 1. This is a non-
SMTL chain, denoted by G, and is such that N ( G) = [a, 1). The operator
in the corresponding standard algebra is any function : [0, 1] [0, 1], such that:
(i) (x) = 1 if x {0, 1}
(ii) (x) = 0 if x (0, a) (the segment where x x = 0),
(iii) (x) is non-decreasing in N ( G) = [a, 1) (the segment where x = 0).
This shows that there are as many consistency operators in this algebra as there are
non-decreasing functions over the interval [a, 1] with values in [0, 1].
The second example consists of the ukasiewicz logic , that is, the logic of the
ukasiewicz t-norm given by x y = max {x + y 1, 0}, which is complete with
def

respect to the standard chain [0, 1] . Given the fact that the negation is involutive in
this structure, then N ([0, 1] ) = , and there is thus a unique operator definable in
the ukasiewicz standard chain. It is defined as follows: (x) = 1 if x {0, 1}, and
(x) = 0 otherwise.
It is easy to prove that L conservatively extends L :
Proposition 5.2.26 Let L as in Definition 5.2.21. Then L is a conservative exten-
sion of it. That is, for every set of formulas  {} L&
, the following holds:

 L iff  L .

Proof The if part is trivial. For the only if part, suppose that  L . By com-
pleteness, there exists an L -chain A and a valuation v over A such that v[] {1}
and v() = 1. The chain A can be expanded to an L -chain A
by defining an opera-
tor : A A such that (1) = (0) = 1 and (x) = 0 for every x A {0, 1}. By
extending v to a valuation v
over A
such that v
() = (v
()) for every formula
, the algebra A
and the valuation v
show that  |=L . Therefore, by soundness,
 L . 
5.2 LFIs Based on Fuzzy Logics 201

As observed in all the previous examples, the -operators defined in L -chains

A are completely determined over the set {x A : x x = 0}, but they can
be defined in different ways in the segment where x = 0. We will now consider
consistency operators defined in logics whose associated chains have no elements
x < 1 such that x = 0; that is, in chains A where N (A) = ) or, in other words,
where x = 0 always implies x = 1. These logics, which are obtained from any L
by adding a suitable inference rule, will be denoted as L .
Definition 5.2.27 The logic L is obtained from L by adding the following rule:

()



It should be clear that L is complete with respect to the corresponding quasi-
variety L of L -algebras, that is, the class of L -algebras satisfying the quasi-
equation If x = 1 then x = 1, or equivalently the quasi-equation If x = 0
then x = 1. As observed in [28], L is not a variety. On the other hand, L is a
semi-linear logic:
Proposition 5.2.28 The rule

()

(where is an arbitrary formula) is derivable in L .

Proof As is a theorem of MTL, it follows that L ,
and so L ( ). Now, by using the rule (), we see that is
derived from premise in L . 
From the results in Sect. 5.2.1, it follows that:
Corollary 5.2.29 The logic L is semilinear and thus strongly complete with
respect to the class of L -chains.
The following step is to extend logic L by adding a consistency operator
to the signature, as well as the axioms (AxC1), (AxC2a ) and (AxC2b ) and the
inference rules (Cong) and (Coh) of Definition 5.2.21 . The resulting logic, called
L , is clearly sound and complete with respect to both the quasi-variety L
of
L -algebras and the class of chains of L
.
The schemes and inference rule noted below are, respectively, provable and deriv-
able in L :

(B1)
( ) ( ) (B2)
( ) (B3)
(B4)
202 5 LFIs based on Other Logics

(Nec)

Moreover, the schemes and rules above constitute a simpler axiomatization of

L .
Theorem 5.2.30 L can be axiomatized by adding the axiom schemas (B1)(B4)
and the inference rule (Nec) to the Hilbert calculus for L .
Proof Let L+ be the Hilbert calculus over the signature & {} obtained from
L by adding the axioms (B1)(B4) and the rule (Nec). It is easy to see that
axioms (B1)(B4) and the rule (Nec) are sound w.r.t. L -algebras. It is necessary
to prove that all the axioms of L are provable in the new system L+ , and that the
rules (Cong) and (Coh) are derivable in L+ . Axiom (AxC1) is clearly obtained from
(B1) since implies , and this last is equivalent to
(AxC1). Axiom (AxC2a ) follows easily from rule (Nec), while (AxC2b ) is in fact
(B4). Now we only have to prove that (Cong) and (Coh) are derivable in L+ .
Firstly, from ( ) , it follows that (( ) ) by rule (Nec); by (B3)
and MP, ( ) is obtained. Finally, by (B2), MP and the monotonicity of
, it follows that ( ) . Hence (Cong) is derivable.
Now, from ( ( )) , it follows that and ( ) .
Using () , one obtains and ( ) and thus ( )&(( ) )
as well. By the properties of &, we infer (&(( ) )) (&(( ) ))
and by MP and monotonicity of & and , we obtain . Using (Nec), it follows
that ( ) and, by axiom (B3) and MP, is inferred. Since MTL proves
q (q
q) for every variable q, q
then, by structurality, ( ) is
a theorem of L+ . Thus, by monotonicity of and MP, ( ) is obtained
from . This shows that rule (Coh) is also derivable in L+ . 
Given that L is chain-complete, determining what sort of operators are
definable in a L -chain A is an interesting task. However, since N (A) = by
definition, the operator required is completely determined and defined as follows:

1, if x {0, 1}
(x) =
0, otherwise

This means that there is only one consistency operator over each L -chain,
which is defined in the most possible intuitive way. Moreover, the consistency oper-
ator is crisp in such cases, that is, it only assumes the classical truth-values 0 and 1.
Remark 5.2.31 The relationship between LFIs and core fuzzy logics was investi-
gated for the first time in [29], by considering the algebras of the logic L (the
expansion of L with the Monteiro-Baaz  operator), and a consistency operator
was defined within these algebras as (x) = (x x). It was further proved
def

in [28] (where such ideas were generalized) that, in L -algebras, the  operator
is also definable as x = (x) x.
5.2 LFIs Based on Fuzzy Logics 203

From the last observation, the following result is easy to prove, using chain com-
pleteness for both logics.
Corollary 5.2.32 L -algebras and (L ) -algebras are termwise equivalent,
hence the logics L and (L ) themselves are equivalent.
Our next step is to analyze the crisp case in general, that is, not only in algebras
whose chains A are such that N (A) = . Since, in any chain, x {0, 1} iff x x =
1, we arrive at the following notion:
Definition 5.2.33 Let L and L as in Definition 5.2.21. The logic Lc is the Hilbert

calculus over the signature & obtained from L by adding the following axiom
schema:
(c)

An Lc -algebra is an L -algebra A such that (x) (x) = 1 for every x A. 

The logic Lc is an axiomatic extension of L and hence is algebraizable, with
equivalent algebraic semantics given by the quasi-variety Lc of Lc -algebras. It is
also semilinear and thus complete with respect the class of Lc -chains. By definition,
the operator is such that, in any Lc -chain A, (x) {0, 1} for every x A. This
implies that the set {x A {0} : (x) = 1} = N (A) {1} is an interval of the
type [a, 1] or (a, 1].
Consider now the logics corresponding to the minimal and maximal (w.r.t. the
pointwise order) consistency operators.
Definition 5.2.34 Let L and L as in Definition 5.2.21. The logic Lmin is the
Hilbert calculus obtained from L by adding the following axiom schema:

(AxMin)


Following the same reasoning as above, Lmin is sound and complete with respect
tothe class of Lmin -chains, that is, L -chains additionally satisfying the equation
x x (x) = 1. On the other hand, over such chains, the definition of the
operator is analogous to the case of L .
Proposition 5.2.35 The logic Lmin is sound and complete with respect to the class
of L -chains where the operator is the minimal one, which is defined as follows:

1, if x {0, 1}
(x) =
0, otherwise

Proof It is enough to prove that if the equation x x (x) = 1 holds in an L -

chain, then (x) = 0 when 0 < x < 1. Thus, let 0 < x < 1. If x > 0, then (x) =
0 by condition (1) of Definition 5.2.22. Contrarily, if x = 0, then (x) = 1,
by (AxMin), and so (x) = 0. This concludes the proof. 
204 5 LFIs based on Other Logics

In order to characterize the logic where the operator is maximal, an inference

rule is necessary instead of an axiom schema.
Definition 5.2.36 Let L and L as in Definition 5.2.21. The logic Lmax is the
Hilbert calculus obtained from L by adding the following inference rule:

( )


Once again, since rule ( ) is closed under disjunction, Lmax is sound and
complete with respect to Lmax -chains, that is, L -chains satisfying the following
condition: if x = 0, then x = 1. This produces a maximal operator :
Proposition 5.2.37 The logic Lmax is sound and complete with respect to the class
of L -chains where the operator is maximal, defined as follows:

1, if x x = 0
(x) =
0, otherwise

Proof Suppose that, in an Lmax -chain, x x > 0; then, (x) = 0 by condi-

tion (1) of Definition 5.2.22. Now, if x x = 0, there are then two possibili-
ties: x {0, 1} or x = 0. In the first case, (x) = 1 by condition (2) of Defin-
ition 5.2.22. In the second, x = 1 and so, by ( ) (with taking value 0), it
follows that (x) = 1. This concludes the proof. 
By considering the proof of conservativeness of L with respect to L in Propo-
sition 5.2.26, it is easy to see that the same kind of proof also applies to the logics
L , Lc , Lmin and Lmax . In this way, L is a conservative extension of L ,
while Lc , Lmin and Lmax are conservative extensions of L .
Finally, some thoughts on the systems in which we are currently interested, that
is, the extension of paraconsistent logics L with a consistency operator, , which
will be denoted by L . As mentioned in Sect. 5.2.1, these are the logics which are
in fact paraconsistent (and, moreover, strong LFIs) and not the ones studied in the
previous Sect. 5.2.2.
However, from the results in Sect. 5.2.1, we know how to obtain L from L .
In fact, it is enough to consider the same axioms of L and then add the inference
rules (Adj-) and (MP-r ) of Proposition 5.2.16, along with the following rules:

L ( ) L ( ( ))
(Cong-r ) (Coh-r )
( ) ( )

The corresponding Hilbert calculi for the logics (L ) , (Lc ) , (Lmin ) and
(Lmax ) are easily defined, using the same technique. Thus, (L ) is obtained
after additionally considering the inference rule:

L
(-r )

5.2 LFIs Based on Fuzzy Logics 205

while the logic (Lmax ) must include the following rule:

Lmax
( -r )

5.2.3 Propagation of Consistency and DAT

As analyzed in Sect. 3.6 of Chap. 3, the so-called propagation of consistency, first

introduced by N. da Costa for his paraconsistent systems, plays an important role
in the framework of LFIs. Definition 3.6.1 can be adapted to the present setting to
determine conditions which ensure its validity.
Definition 5.2.38 Let L be a paraconsistent fuzzy logic with a consistency operator
. We say that satisfies the propagation property in L with respect to a subsignature
of the signature of L if

1 , . . . , n L #(1 , . . . , n )

for every n-nary connective # and formulas 1 , . . . , n in L . 

The case n = 0 corresponds to a constant c, so the relation above can be expressed
by
L c.

The paraconsistent fuzzy logics studied in this section are the logics L defined
over the signature & = {, &, , } as well as some extensions of it where L is a
truth-preserving fuzzy logic. From the relation between truth-preserving and degree-
preserving fuzzy logics stated above, we can say that satisfies the propagation
property in a logic L with respect to some subsignature of & iff

L ( ) (#) for each binary # (Pr op)

Recalling that L , we can assume that does not belong to since it is already
propagated, justifying the definition of condition (Prop*).
Proposition 5.2.39 The following conditions hold:
(1) satisfies (Prop*) in any fuzzy logic L with respect to {, }.
(2) satisfies (Prop*) in any fuzzy logic of the families L , Lc , Lmin or Lmax
with respect to {, &, }.
Proof Because of the chain completeness of the logics involved, the proof is done
by algebraic means. Let A be a L -chain.
(1) The proof for the connective is a consequence of the non-decreasing property
of in A {0} and the fact that (0) = 1. For , the proof is easy. Remembering
206 5 LFIs based on Other Logics

that a residuated implication satisfies the inequality y x y and assuming y > 0,

we have (x) (y) (y) (x y). Now, if y = 0 then x y is the negation
x of x, and so the property can be proven by cases. If x {0, 1}, the property is
obvious. If x N (A), then x = 0 and so (x) = 1. Thus, the property holds. If
not, x x > 0 and, then, (x) = 0, again satisfying the property.
(2) It is enough to analyze the case of &. Observe that if x = 0 and y = 0,
(x & y) = 0 as well, as (x & y) = (x & y) 0 = x (y 0) = x y =
x 0 = x = 0. From here, the result follows as a consequence of the following
facts: (i) in the chains of the considered logics, the image of is {0, 1}, (ii) & is closed
for the classical truth-values {0, 1} and (iii) if N (A) = then, for all x N (A),
either (x) = 0 or (x) = 1. 
The first item of Proposition 5.2.39 cannot be improved in the sense that L does
not prove () () (&) (see the example presented in [28]).
It should be clear that the fuzzy logics analyzed here can be seen as subsystems
of CPL, identifying and & as two different aspects of the classical conjunction.
However, we can now analyze the question of recovering classical reasoning (that is,
inferences in CPL) by means of the consistency connective , as was done between
CPL and mbC with the Derivability Adjustment Theorem (or DAT) in Theorem 2.4.7
of Chap. 2. Moreover, thanks to the property of consistency propagation stated above,
a better DAT can be obtained by exclusively requiring consistency exclusively for
the relevant propositional variables, as was done in Theorem 3.6.10 for the system
mbCw1 as well as for its extensions, such as Cila.
From now on, we will investigate how to define a DAT for LFIs of the form
L defined over & . As observed above, has the propagation property iff certain
formulas are theorems in L . Thus, when propagation holds, we are looking for a
DAT, in terms of theoremhood, as follows:
 n 
 n

(PDAT ) CPL iff L pi iff pi L
i=1 i=1

where { p1 , . . . , pn } is the set of propositional variables occurring

n in . In the partic-
ular case when n = 0, that is, when only uses constants, i=1 pi is then defined
to be the constant .
Assuming that satisfies the propagation property in logics L , with respect to
the classical signature (see Proposition 5.2.39), we want to prove wheather or not
they have the property (PDAT*).
Remark 5.2.40 It is important to observe that, in general, logics L do not satisfy
(PDAT*). For instance, we know that CPL p p; however, L p ( p p),
that is, p ( p p) is not always a tautology over L -chains. In order to see
this, consider the L -chain defined on a L -chain A by taking as follows: (x) = 1
if x {0, 1} N (A), and (x) = 0 otherwise. From here, it is an easy task to prove
that if N (A) = then v( p ( p p)) = 1 for any valuation v on A, such that
v( p) N (A).
5.2 LFIs Based on Fuzzy Logics 207

The example of the last remark is paradigmatic, given that the axiom schema
is enough to produce the collapse between MTL-logic and classical logic
CPL. This justifies the following definition:

Definition 5.2.41 Let Ldat be the logic defined over & = & { } which is
obtained from L by adding the following axiom:

( ) (EM)


Following the same reasoning as in the previous cases, Ldat is algebraizable in

the sense of Blok and Pigozzi, and its algebraic semantics is given by the variety
of L -algebras, which will now be defined.
Ldat dat

Definition 5.2.42 A Ldat -algebra is a L -algebra A = A, , &, , , 0, 1 such

that (x) x x for every x A. 

It follows that Ldat is sound and strongly complete with respect to the variety
Ldat
and, additionally, to the chains of the variety. This is true because the logic is
an axiomatic extension and, thus, it is semilinear like L .
Moreover, since Ldat extends L , it follows that also satisfies the propagation
property in Ldat with respect to the classical signature.
However, as it was observed with respect to the deduction meta-theorem DMT,
only a weak form (the local DMT) holds in the logics being studied (recall Theo-
rem 5.2.14). Based on this, a slightly modified form of DAT is proven to hold for
Ldat :

Proposition 5.2.43 The logic Ldat satisfies the following form of DAT:
 n k


(PDAT ) CPL iff there is k 1 such that Ldat pi
i=1

where { p1 , . . . , pn } is the set of propositional variables occurring in and k , which

are defined as in Theorem 5.2.14.

Proof Let be a formula in the language of CPL such that { p1 , . . . , pn } is the set
of propositional variables occurring in it. Assuming that CPL , it is clear that

{ pi pi : i = 1, 2, . . . , n} L .

Indeed, for any valuation v on a chain, v( pi pi ) = 1 iff v( pi ) {0, 1}. By the

local DMT of L , there is some k 0 such that
 n k

L ( pi pi )
i=1
208 5 LFIs based on Other Logics

which of course, also holds in Ldat . Therefore, by axiom (EM), it follows that
 n k

Ldat pi
i=1

and then the property (PDAT ) holds.
n
Suppose now that Ldat ( i=1 pi )k for some k 1 and consider any val-
uation v on the 2-element Boolean algebra B2 . Given the fact that B2 can be consid-
n
ered as a Ldat -chain (by taking (0) = (1) = 1), then
n v(( i=1 pi ) ) = 1
k

by soundness of L . But then, v() = 1 since v( i=1 pi ) = 1. From here, is a

dat

CPL-tautology and so, by completeness of CPL, CPL as required. 

The previous result is immediately extendable to cases with a finite set of premises:
 k

m
 CPL iff there is k 1 such that  Ldat pi
i=1

where { p1 , . . . , pm } is the set of propositional variables occurring in  {}.

Finally, observe that axiom schema (EM) is a theorem of the logics L and
L since any instance of it is a tautology in the class of chains. Therefore, in these
min

logics, pi pi is derivable by EM and by the fact that pi is a tautology w.r.t.

chains, and is thus a theorem of L . From here, pi is equivalent to ( pi )k for any
k. Thus, the following corollary is obtained:

Corollary 5.2.44 Let  {} be a finite set of formulas in the language of CPL and
let { p1 , . . . , pm } be the set of propositional variables occurring in  {}. Then,
for any + {, min},
 n 

 CPL iff  L+ pi .
i=1

5.2.4 Fuzzy Logics with an Inconsistency Operator

As studied in Sect. 3.5 of Chap. 3, it is possible (and also expected) to consider, in

the place of (or together with) a consistency operator , an inconsistency operator ,
dual to , in such a way that the intended meaning of is .
Using the content of the previous sections concerning consistency operators as
our base, this section shows how to add inconsistency operators to L -algebras, as
well as to their logical counterparts.

Definition 5.2.45 Given an axiomatic extension L (which can also have new infer-
ence rules) of MTL over the signature &
, which is not SMTL, we define the logic
5.2 LFIs Based on Fuzzy Logics 209

L as the Hilbert calculus over the signature & = &

{ } obtained from L by
adding the following:

Axiom schemas:

( ) (AxC1
)
 (AxC2
a )
(AxC2
b )

Inference rules:
( ) ( ( ))
(Cong
) (Coh
)
( ) ( )


As in the case of L , due to the presence of the rule (Cong

), L is algebraizable in
the sense of Blok and Pigozzi and its algebraic semantics is given by the quasi-variety
L of L -algebras, which will be now be defined.

Definition 5.2.46 An L -algebra is an structure A over the signature & , such that

its reduct to the signature & is an L -algebra and the operation : A A satisfies
the following conditions for all x, y, z A:
(1) (x x) (x) = 1
(2) (1) = (0) = 0
(3) if (x (x y)) z = 1 then ((y) (x)) z = 1. 

Again, since the rules (Cong

) and (Con
) are closed under -forms, L is complete
with respect to the class of L -chains.
Using techniques similar to the ones presented in Sect. 3.5 of Chap. 3, namely, by
using a simplified version of the translation mappings + and from Definition 3.5.4,
it is possible to prove the intended duality between the operators and , along the
same lines as Theorem 3.5.11.

Proposition 5.2.47
(1) Let t be the mapping from the language of L to the language of L , which
replaces with . Then t is a translation from L to L , that is: for every set of
formulas  {} of L ,  L implies that t[] L t ().
(2) Let t
be the mapping from the language of L to the language of L , which
replaces with . Then t is a translation from L to L , that is: for every set of
formulas  {} of L , 
L
implies that t
[
] L t
(
).

Proof For both items, observe that it suffices to prove that: (a) the translation of each
axiom of the source logic can be derived in the target logic, and (b) the translation
of each inference rule of the source logic is an inference rule which is derivable in
210 5 LFIs based on Other Logics

the target logic. Here, the translation of an inference rule r : / using , is the
inference rule (r ) : ()/ (). In order to prove this, the following theorems of
MTL will be used for both items:
(i) ( ) ( )
(ii) ;
(iii) ( ) ( ).
(1) Let us analyze the translation of each axiom and inference rule of L , using t:
Axiom (AxC1): Let be an instance ( ) of (AxC1). Then t () is the
formula (t () t () t ()). Observe that, by (AxC1
), the formula (t ()
t ()) t () is a theorem of L . Then, by (ii) and monotonicity of , the formula
(t () t ()) t () is a theorem of L . But then, by (i), the formula t ()
is derivable in L .
Axiom (AxC2a ): The translation of this axiom is itself an axiom, namely (AxC
2a ).
Axiom (AxC2b ): The translation of this axiom is itself an axiom, namely (AxC
2b ).
Rule (Cong): Let / be an instance of (Cong), where is ( ) and
is ( ) . Then t () is (t () t ()) t () and t () is (t ()
t ()) t (). Assuming t (), we obtain, by (Cong
), (t () t ()) t ()
and then t () follows from (iii) (twice) and the properties of .
Rule (Coh): Let / be an instance of (Coh), where is ( ( ))
and is ( ) . Then t () is (t () (t () t ())) t () and t ()
is (t () t ()) t (). Assuming t (), we obtain (t () t ()) t ()
by (Coh). Then t () follows by (iii).
From this, it follows that if 1 . . . n is a derivation of from  in L , the sequence
t (1 ) . . . t (n ) can be transformed into a derivation of t () from t[] in L . This
proves that t is a translation from L to L .

(2) The proof is analogous to that of item (1). Details are left to the reader as an easy
exercise. 
The last result establishes a sort of equivalence between L and L , analogous
to the one obtained in Theorem 3.5.11 between mbCcew and mbC cew. However,
this result does not mean that both logics are equivalent in the stronger sense of Blok
and Pigozzi (see [34]). Indeed if is a formula of L , then it is not equivalent, in
general, to the formula t
(t ()). For instance, p is not equivalent to p (being
p a propositional variable). Analogously, p is not equivalent to p, and so a
formula of L is not equivalent, in general, to the formula t (t
()).
However, in the context of the logic Lc and its extensions, where is Boolean,
it is not hard to prove . It is possible, in this way, to establish the
equivalence between L and the corresponding logic L . In particular, if L is such
that the negation is involutive, then Lc coincides with L itself.
Consider the inconsistency counterparts of the logics L , Lc , Lmin or Lmax
defined as follows:
5.2 LFIs Based on Fuzzy Logics 211

L : is the extension of L with the rule from infer

Lc : is the axiomatic extension of L with the axiom schema
Lmax : is the axiomatic extension of L with the axiom schema
Lmin : is the extension of L with the rule from infer

The following equivalences among logics via the translations t and t

(denoted by
) can be thus established: L  L , Lc  Lc , Lmin  Lmax and Lmax 
Lmin .
As a consequence, the (quasi) varieties associated with pairs of equivalent logics
are term-wise equivalent.
As in the case of the consistency operators, it should be stressed that the paracon-
sistent versions of the above logics defined in terms of the inconsistency operator
correspond to their degree-preserving counterparts, namely the logics L , (L ) ,
(Lc ) , (Lmin ) and (Lmax ) .

5.3 A Modal LFI Based on Belnap and Dunns Logic BD

In his Ph.D. thesis of 1966 (see [35]), Dunn introduced the notion of proposition
surrogates, also called polarities (see [36]), as a set-theoretic device to represent De
Morgan Lattices. From this semantical framework, a four-valued logic was defined,
which was afterwards developed by Belnap in [37, 38], originating the famous logic
now known as Belnap-Dunns logic FOUR. As observed by Belnap, the features
of this logic allow to deal with lack of information (a sentence is neither true nor
false) or with excess of information (a sentence is both true and false), thus being
suitable to represent information systems under a computer science perspective. So,
besides the classical truth-values 0 and 1, there are the non-classical truth-values
N (representing neither, that is, lack of information) and B (representing both,
that is, excess of information). It is useful to consider N as and B as { f alse, tr ue},
while 0 represents { f alse} and 1 represents {tr ue}. These four truth-values define a
lattice that we call M4 , which is diplayed above.

In this lattice, infima interpret a conjunction while suprema interpret a disjunc-

tion , as expected:
212 5 LFIs based on Other Logics

1 B N 0 1 B N 0

1 1 B N 0 1 1 1 1 1
B 1 B 0 0 B 1 B 1 B
N 1 0 N 0 N 1 1 N N
0 0 0 0 0 0 1 B N 0

Belnap and Dunn enriched the lattice M4 with a negation defined as follows:

1 0
B B
N N
0 1

This negation is a De Morgan negation, since it satisfies the following, for every
x, y M4 (compare with the axioms of Nelsons logic N4, Definition 5.1.24):

x = x (DM1)
(x y) = x y (DM2)
(x y) = x y (DM3)

The logical matrix M4 , {1, B} over the algebra M4 = M4 , , , defines a logic
called BD, which is clearly paraconsistent and paracomplete:

p, p BD q and BD q q

if p and q are two different propositional variables. In fact, it is enough to consider

a valuation v over BD such that v( p) = B and v(q) = N .
The lattice structure above was called logical lattice by Belnap, based on the truth
order. There is another lattice structure than can be defined over M4 considered as
the power set of { f alse, tr ue}, with the inclusion order, called the knowledge order.
In this structure N = is the bottom element and B = { f alse, tr ue} is the top
def def

element, while 0 = { f alse} and 1 = {tr ue} are incomparable. This produces a
def def

lattice structure called approximation lattice, and so M4 is the domain of a bilattice

structure which generates the well-known logic FOUR. Bilattices were afterwards
studied in the context of Logic applied to computer science (see for instance [3941]).
Only the logical lattice structure of M4 will be relevant to our purposes.
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 213

The lattice structure M4 = M4 , , , , on which the logic BD is based, con-

stitutes a De Morgan lattice, that is, an algebra A, , , of type (2, 2, 1) such
that the reduct A, , is a distributive lattice and satisfies equations (DM1)-
(DM3) above. A De Morgan algebra is a De Morgan lattice such that the underlying
lattice is bounded. It is well-known that M4 generates the variety DM of De Morgan
lattices, while M4 , , , , 0, 1 generates the variety DMA of De Morgan alge-
bras. See [42] for a nice study of the logic BD of the matrix M4 , {1, B} from an
algebraic perspective in connection with De Morgan lattices. The system BD will
be taken up again in Sect. 5.3.3, when the logic preserving degrees of truth M4m of
the variety TMA of tetravalent modal algebras will be presented as a dC-system.

5.3.1 The Logic M4m of Tetravalent Modal Algebras

Before the investigations of Belnap and Dunn on the logic FOUR, in 1963
Monteiro has shown in [43] the independence of the axioms for the variety of 3-
valued ukasiewicz algebras proposed by his father, A. Monteiro. In order to prove
the independence of one of such axioms, Monteiro proposed the algebra

M4m = M4 , , , , , 0

where the reduct M4 , , , coincides with M4 , with 1 = 0, and where  is

def

defined as follows:

1 1
B 0
N 0
0 0

Motivated for such example, A. Monteiro envisaged that it could generate an

interesting class of algebras generalizing 3-valued ukasiewicz algebras, that he
called in 1978 tetravalent modal algebras. This class of algebras was firstly studied
by his Ph.D. student I. Loureiro (see for instance [44]) and afterwards by several
others as it will be clear in this chapter.
In formal terms, a tetravalent modal algebra (TMA) is an algebra U = A, , ,
, , 0 of type (2, 2, 1, 1, 0) where the reduct U = A, , , , 0, 1 is a De Mor-
gan algebra (with 1 = 0), such that the unary operation  satisfies the following
def

two axioms:
214 5 LFIs based on Other Logics

x x = 1 (TMA1)
x x = x x (TMA2)

for every x A. The class of all tetravalent modal algebras constitute a variety
denoted by TMA, which is generated precisely by M4m .
The logic(s) associated to TMA are now analyzed. In Sect. 5.2 we distinguished,
following the approach in [33] (introduced in [32]), between two different semantical
consequence relations associated to a class of algebras: the truth-preserving semantics
(recall Definition 5.2.9) and the consequence relation preserving degrees of truth (see
Definition 5.2.15), and so both approaches can be related to TMA. However, as in
the case of MTL-algebras, the truth-preserving semantics (which, in the present case,
would correspond to the matrix logic characterized by the class of logical matrices
U, {1} such that U is a TMA) produces a logic which is clearly not paraconsistent: in
a TMA it is impossible to have x = 1 and x = 1 simultaneously and so the negation
is explosive. On the other hand, the consequence relation preserving degrees of
truth produces an LFI, as we shall see in Theorem 5.3.6 below, and so this is the
semantical perspective that will be adopted from now on.
Let  = {, , , , } be the signature for TMAs.

Definition 5.3.1 The logic preserving degrees of truth of the variety TMA is the
logic LTMA over L whose consequence relation |=TMA is defined as follows:
for every set  {} L ,  |=TMA if and only if, there is some finite set
0  such that, for every U TMA and for every homomorphism h : L U,

{h( ) : 0 } h(). In particular, |=TMA if and only if h() = 1 for
every U TMA and for every homomorphism h : L U. 

Definition 5.3.2 The 4-valued modal logic M4m is the logic over L whose conse-
quence relation |= M4m is defined as follows: for every set  {} L ,  |= M4m
if and only if, thereis some finite set 0  such that, for every homomorphism
h : L M4m , {h( ) : 0 } h(). In particular, |= M4m if and only
if h() = 1 for every homomorphism h : L M4m . 

Given that M4m generates the variety TMA, it is immediate to prove the following:

Proposition 5.3.3 The logic M4m coincides with the logic LTMA . That is: for every
set  {} L ,  |= M4m if and only if  |=TMA .

As proven by Font and Rius in [45], the logic M4m is in fact a matrix logic
expanding Belnap and Dunns logic BD. We will give now a direct proof of this fact,
taken from [46]. We begin by stating a extremely useful technical lemma about the
algebra M4m .

Lemma 5.3.4 Let h : L M4m be a homomorphism and let V

be a set of propo-
sitional variables. Consider the homomorphism h
: L M4m such that, for
every p V
,
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 215


h( p) if h( p) {0, 1}

h ( p) = N if h( p) = B


B if h( p) = N .

Then, for every L depending on the variables in V

:

h() if h() {0, 1}
h
() = N if h() = B

B if h() = N .

Proof By observing the operations in M4m , the proof follows easily by induction on
the complexity of .9 The details are left to the reader as an exercise. 
Theorem 5.3.5 For every set of formulas  {} L ,

 |= M4m if and only if  |=M B ,

where M B = M4m , {B, 1} .

def

Proof Observe that, by definition of |= M4m ,  |= M4m if and only if |= M4m , for
some formula . In fact, it is enough to take the conjunction of the elements of the
finite subset 0 of  such that 0 |= M4m , if 0 = ; otherwise, we take as .
The same reasoning applies to |=M B . It is, thus, enough to consider inferences of the
form |= M4m and |=M B .
Thus, suppose that |= M4m and let h : L M4m be a homomorphism
such that h() {1, B}. Since h() h(), we have that h() {1, B}. From this,
|=M B .
Conversely, suppose that |=M B and let h : L M4m be a homomor-
phism. If h() = 0 then h() h(). If h() = B then h() {1, B} whence, by
hypothesis, h() {1, B} and, therefore, h() h(). If h() = N , let h
as in the
above lemma. Then, h
() = B {1, B} and so h
() {1, B}, by hypothesis. From
this, h() {1, N } and then h() h(). Finally, if h() = 1 then h() {1, B},
by hypothesis. Suppose that h() = B; then, by considering h
as above, h
() = 1
and h
() = N / {1, B}, a contradiction. Therefore h() = 1 whence h() h().
The previous analysis shows that |= M4m . 
It follows from the previous theorem that M4m (and so LTMA ) is both paraconsis-
tent and paracomplete (the proof is identical to the proof for BD presented above).
Moreover, as it was observed in [47], it is in fact an LFI.
Theorem 5.3.6 Given a propositional variable p, let ( p) = ( p p), and let
def

( p) = { p}. Then the logic M4m is a strong LFI (w.r.t. and ( p)).

9 The complexity l() of a formula in L is defined as expected, in a similar way to Defini-
tions 2.1.11 and 2.5.1, namely: l() = 1 for V ar {}; l(#) = l() + 1 for # {, }; and
l( # ) = l() + l() + 1 for # {, }.
216 5 LFIs based on Other Logics

Proof Recalling Remark 2.1.10 of Chap. 2, let p and q be two different proposi-
tional variables. If h 1 is a homomorphism such that h 1 ( p) = B and h 1 (q) = 0 then
h 1 ( p) = h 1 ( p) = B {B, 1} but h 1 (q)
/ {B, 1}, showing that p, p |= M4m q and
so clause (i.a) of Remark 2.1.10 is satisfied. Considering now a homomorphism h 2
such that h 2 ( p) = 1 and h 2 (q) = 0 we show that p, p |= M4m q and then clause (i.b)
of Remark 2.1.10 is satisfied. Now, let h 3 such that h 3 ( p) = 0 and h 3 (q) = 0; then,
it follows that p, p |= M4m q and so clause (i.c) of Remark 2.1.10 is also satisfied.
Finally, by analyzing the truth-tables of M4m , it is clear that for no homomorphism
h and formula is the case that h(), h() and h() belong simultaneously to
{B, 1}, proving that clause (ii) of Remark 2.1.10 is satisfied. This shows that M4m is
a strong LFI w.r.t. and , in the sense of Definition 2.1.9. 

The last result proves that M4m or, equivalently, LTMA , is a (strong) LFI which
is an expansion of Belnap and Dunns logic BD by a necessitation operator . The
truth-table for the consistency operator in the algebra M4m is defined as follows:

1 1
B 0
N 0
0 1

In Sect. 5.3.3 it will be proven that M4m , seen as an LFI, is a dC-system based on the
logic preserving degrees of truth LBDL of the variety BDL of bounded distributive
lattices.

Proposition 5.3.7 The operator p = ( p p) is the unique such that M4m is

def

an LFI w.r.t. and p.

Proof Let
p be a consistency operator defined in M4m . That is,
p is a formula in
L depending exclusively on the propositional variable p satisfying the clauses of
Definition 2.1.7 of LFIs. Let us denote also by
the operator
: M4 M4 obtained

p in the algebra M4m . It follows that, for every x,
by interpreting the formula
x x
(x) 0, N (because of Item (iii) of Definition 2.1.7). But, if x x

(x) = N for some x then it is possible to have y y

(y) = B for some other

y, by Lemma 5.3.4, a contradiction. Therefore x

x (x) = 0
for every x. In
particular, N
(N ) = 0, whence

(N )
0, B . Since 0, N , 1 is a subalgebra
of M4m it follows that
(N ) 0, N , 1 , therefore
(N ) = 0. Analogously it is
proved that
(B) = 0. Thus, the truth-table of any consistency operator
p defined
in M4m must be as follows:
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 217

1 a
B 0
N 0
0 b

By Item (ii.a) of Definition 2.1.7 there is some x such that x

(x) = 0. Hence,
it follows that a = 0 (because of x = 1). By turn, x x = 0 for some x, by
Item

(ii.b) of Definition 2.1.7. From this, b = 0 (because of x = 0). Given that
0, 1 is a subalgebra of M4m , it follows that a = b = 1. Hence,
= . That is, p
is the unique consistency operator definable in M4m . 

Remark 5.3.8 It is easy to prove that, by expanding BD with a bottom element 0

and with the consistency operator with a truth-table as above, the operator  can
be defined in such algebra as (x) = (x) x, and so the class of TMAs could
def

be defined in terms of a primitive consistency operator, provided that a suitable

axiomatization is given for in the signature {, , , , }. It is worth noting that
this situation is entirely analogous to the proposal in [29] (see Remark 5.2.31 above)
concerning LFIs based on core fuzzy logics. Indeed, in that paper the consistency
operator was defined in the algebras of the logic L (that is, the expansion of L with
the Monteiro-Baaz  operator, which is similar to ) exactly in the same manner
as it is defined in TMAs: (x) = (x x). Additionally, the operator  can be
def

defined from , in L -algebras, as (x) = (x) x. These similarities suggest

def

that it is possible to apply the same ideas, as well as the generalization presented
in [28] and described in Sect. 5.2 above, to a wide class of algebras and its associated
logics.

Concerning the presentations of M4m by means of syntactical deductive systems,

Font and Rius introduced in [45] a sequent calculus for this logic, called G. The
propositional logic associated to this calculus was called T ML, defined as follows:
for every set  {} L ,  T ML iff there exists a finite set 0  such
that the sequent 0  has a derivation in G. It was proved in [45] that the logic
T ML defined by the sequent calculus G coincides with M4m , and so G consti-
tutes a proof-theoretic counterpart of M4m . Figallo has shown in [46] that G does
not satisfy the important property of Cut-elimination. In order to circumvent this
difficulty, he introduced an hypersequent calculus adequate for M4m , which enjoys
Cut-elimination. Also in [46] he proposed a sound and complete tableau system for
M4m . Finally, in [47] (see also [46]) it was introduced a very simple Hilbert-style
axiomatization for M4m , which is sound and complete. This system will be analyzed
in Sect. 5.3.5.
218 5 LFIs based on Other Logics

5.3.2 M4m as an LFI

A short analysis of M4m , seen as an LFI, will be presented in this subsection.

Proposition 5.3.9 The logic M4m is an LFI in which:

(i) |= M4m ( ) and so axiom (ciw) is not valid;
(ii) |= M4m and so axiom (ci), presented as a rule, is not valid;
(iii) ( ) |= M4m and so axiom (cl), presented as a rule, is not valid.

Proof Let p and q be two different propositional variables, and let h be a homomor-
phism such that h( p) = N and h(q) = B. Then:
(i) h( p ( p p)) = 0 N = N / {B, 1}, whence M4m does not satisfy
axiom (ciw).
(ii) Observe that h( p) = 1, while h( p p) = N / {B, 1}.
(iii) Observe that h(q q) = B, while h(q) = 0 / {B, 1}. 

Besides the proposition above, Figallo developed in [46] a brief but interesting
study of M4m under the point of view of LFIs, whose main results are reproduced
below.
Concerning the propagation properties of consistency operator of M4m , he obtained
the following result:

Theorem 5.3.10 The logic M4m satisfies the following:

(i) |= M4m ,
(ii) |= M4m ,
(iii) |= M4m and |= M4m ,
(iv) , |= M4m (#) for # {, },
(v) |= M4m n , for all n 0 (in particular, |= M4m ). In this way, M4m vali-
dates all the axioms (ccn ) of the logic mCi (recall Theorem 3.5.15(iii), Chap. 3).

Proof It follows by considering the truth-tables of M4m . Details are left as an exercise
to the reader. 

Observe that, since M4m does not validate axiom (ciw), by Proposition 5.3.9(i), the
results stated in Sect. 3.6 of Chap. 3 cannot be used in order to prove Theorem 5.3.10.
As usual in the framework of LFIs (recall Sect. 3.5), it is possible to define an
inconsistency operator on M4m in the following way:

= .
def

In this way is equivalent to ( ), where =  defines, as usual

def

in modal logic, a possibility operator, with the following truth-tables:

5.3 A Modal LFI Based on Belnap and Dunns Logic BD 219

1 1 1 0
B 1 B 1
N 1 N 1
0 0 0 0

The proof of the following result is immediate.

Theorem 5.3.11 In M4m it holds:

(i) |= M4m but |= M4m ,
(ii) |= M4m and |= M4m ,
(iii) (#) |= M4m for # {, }, but the converse does not hold.

The last result shows that the concept of inconsistency, on the one hand, and
contradiction, on the other, can be disassociated in the logic M4m of TMA-algebras.
This is a valuable feature in the universe of LFIs, just enjoyed by mbC among the
logics studied in Chap. 3. Despite the fact of M4m not being functionally complete
(see comment after Proposition 5.3.16 below), it enjoys a great expressive power. For
instance, in M4m it is possible to speak about the classical truth-values (0 and 1),
as well as to identify the non-classical ones, namely N and B. Thus, for instance,
is satisfied exactly when just assumes the truth-values 0 and 1. On the other
hand, asserts, on the contrary, that the truth-value of is either N or B.
The following interaction law between the inconsistency and consistency opera-
tors of M4m holds:
, ( ) |= M4m (Iprop)

This property, together with others mentioned here, suggests that M4m , seen as an
LFI, can be suitable for concrete applications, such as inconsistent databases or
paraconsistent logic programming (once the contrapositive implication is taken into
consideration, see Sect. 5.3.4 below).
As it was done with other LFIs, it is possible to recover classical logic inside
M4m . Thus, consider the presentation of classical propositional logic CPL over the
signature  cl = {, , } (we are intentionally using the same symbols for the
common connectives of M4m and CPL). Let L cl be the algebra of formulas of CPL
generated by such signature. Then, we have the following Derivability Adjustment
Theorem (DAT) with respect to CPL.
220 5 LFIs based on Other Logics

Theorem 5.3.12 Let  {} be a finite set of formulas in L cl . Then,

 CPL iff , p1 , . . . , pn |= M4m

where { p1 , . . . , pn } is the set of propositional variables ocurring in  {}.

Proof () Let  = {1 , . . . , k } L cl and let h : L cl Acl 2 be a homomor-

phism, where Acl 2 is the Boolean algebra with domain {0, 1} over the signature  cl .
Since A2 can be expanded to a TMA-subalgebra of M4m (by putting (x) = x for
cl

all x), h can be regarded as a homomorphism h : L M4m . Then, by hypothesis

we have that
k n
h( i p j ) h().
i=1 j=1

By recalling that originally h : L cl Acl

2 , it follows that h( p j ) {0, 1} for all
j such that 1 j n. Thus, by definition of , h( p j ) = h( p j ) = 1 for all j
k
such that 1 j n. Therefore, if h(i ) = 1 for every 1 i k then h( i=1
 i
n
j=1 p j ) = 1, whence h() = 1. This means that  CPL as required.
() The proof is similar. 

Recall from [5, 6] (see also Remark 5.1.15 above) that a propositional logic L is
boldly paraconsistent if there is not a formula ( p1 , . . . , pn ) satisfying the following:
(i) L (1 , . . . , n ) for some 1 , . . . , n , and
(ii) ,  L (1 , . . . , n ) for every , 1 , . . . , n .
Finally, it was proven in [46] that M4m is a boldly paraconsistent logic. In order
to prove this, he obtained a previous useful result:

Lemma 5.3.13 Let and two formulas in L without common propositional
variables. If |= M4m then |= M4m or |= M4m .

Proof Assuming that and do not have common propositional variables, suppose
that |= M4m , |= M4m and |= M4m . Then, there are homomorphisms h, h

:
L M4m such that h() = 0 (since |= M4m ) and h

() = 1 (since |= M4m ).

Given that and do not have common propositional variables, it can be assumed,
without loss of generality, that h = h

. From this, h() = 0 and h() = 1. But

|= M4m and so h() h(). Therefore, h() = 1. If h() = B, then h() = B.
Let us define h
on the set of propositional variables occurring in as in Lemma 5.3.4;
then h
() = N . The mapping h
can be extended to the set of propositional variables
occurring in by setting h
( p) = h( p). Then, h
() = h() = B and so h
() 
h
(), a contradiction. The case h() = N is treated analogously. 

Corollary 5.3.14 The logic M4m is boldly paraconsistent.

5.3 A Modal LFI Based on Belnap and Dunns Logic BD 221

Proof Suppose that M4m is not boldly paraconsistent. Then, there exists a formula
( p1 , . . . , pn ) satisfying the clauses (i) and (ii) above, for the logic M4m . Let p be
a propositional variable such that p / { p1 , . . . , pn }. By clause (ii), ( p p) |= M4m
( p1 , . . . , pn ), where p does not occur in . Since ( p p) |= M4m it follows that
|= M4m ( p1 , . . . , pn ), by Lemma 5.3.13. Therefore |= M4m (1 , . . . , n ) for every
1 , . . . , n , which contradicts (i). From this, it follows that M4m is boldly paracon-
sistent. 
Remark 5.3.15 As observed in Remark 5.1.15, a logic which is not boldly para-
consistent is not a genuine paraconsistent logic: from a contradiction it is pos-
sible to derive all the instances of some schema. As proved in Theorem 5.1.14,
Johanssons minimal logic is paraconsistent but it is not boldly paraconsistent since
, Min for all and all . We have just shown, in Corollary 5.3.14, that
M4m is a genuine paraconsistent logic. On the other hand, as we shall see in Propo-
sition 5.3.16 below, M4m is an LFI in which the (classical) strong negation is not
definable. This is an interesting feature that reveals the importance of looking at M4m
under the perspective of paraconsistency.
A Boolean complement defined in M4m , that is, satisfying the equations

x x = 0 (BN1)
x x = 1 (BN2)

for every x M4 , should have the following truth-table:

1 0
B N
N B
0 1

In [48] it was shown that the algebra M4m cannot define a Boolean complement as
above:
Proposition 5.3.16 It is not possible to define an operator in the algebra M4m
such that conditions (BN1) and (BN2) hold.
Proof Since {0, 1, B} is (the domain of) a subalgebra of M4m , the result follows
easily. 
The last result shows that the algebra M4m is not functionally complete (that is, its
operators cannot define every function f : M4n M4 , for every n 0). Moreover,
the expansion of M4m by adding a Boolean complement as above produces a
222 5 LFIs based on Other Logics

symmetric (or involutive) Boolean algebra which, by its turn, generates the class of
models characterizing a normal extension of modal logic S5 (see [48]).

5.3.3 M4m as a dC-System

In this section it will be shown in Theorem 5.3.22 another interesting feature of M4m
seen as an LFI: it is a dC-system based on the logic preserving degrees of truth
LBDL of the variety BDL of bounded distributive lattices (BDLs). In order to do this,
consider the variety TMA of TMAs defined over the signature   def
=  {  } =
{, , , , , }. The logic preserving degrees of truth LTMA of the variety
TMA is defined by adapting the one for LTMA (recall Definition 5.3.1) to the
expanded signature. Clearly, TMA is generated by M 4m = M4 , , , , , 0, 1 ,

and so the logic LTMA coincides with the logic M4m , which is defined by adapting

Definition 5.3.2 of M4m to the extended language. Of course, both LTMA and M4m
coincide with the logic generated by the matrix MB = M 4m , {B, 1} .
def

Consider now the signature BDL = {, , , } for BDLs.

Definition 5.3.17 The logic preserving degrees of truth of the variety BDL is the
logic LBDL over LBDL whose consequence relation |=BDL is defined as follows:
for every set  {} LBDL ,  |=BDL if and only if, there is some finite set
0  such that, for every U BDL and for every homomorphism h : LBDL U,

{h( ) : 0 } h(). In particular, |=BDL if and only if h() = 1 for
every U BDL and for every homomorphism h : LBDL U. 
Recall that when a negation satisfying equations

x = x (DM1)
(x y) = x y (DM2)
(x y) = x y (DM3)

is added to a distributive lattice, it becomes a De Morgan lattice. The structure

M4 = M4 , , , generates the variety DML of De Morgan lattices (DMLs)
over the signature DML = {, , } for DMLs.
Definition 5.3.18 The logic preserving degrees of truth of the variety DML is the
logic LDML over LDML whose consequence relation |=DML is defined as follows:
for every set  {} LDML ,  |=DML if and only if, there is some finite non-
empty subset 0 = {1 , . . . , n } of  such that, for every U DML and for every
homomorphism h : LDML U, h(1 ) h(n ) h(). 
As in the case of TMAs, since M4 generates DML, it is enough to consider
U = M4 in Definition 5.3.18 instead of testing with all the DMLs U. Moreover, the
following result was proved in [42]:
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 223

Proposition 5.3.19 The logic LDML coincides with the logic BD generated by the
logical matrix M4 , {1, B} .

Recall now that a De Morgan algebra is a De Morgan lattice such that the under-
lying lattice is bounded. It is well-known that M,
4 = M4 , , , , 0, 1 generates
the variety DMA of De Morgan algebras (DMAs). Let DMA = {, , , , } be
the signature for DMAs.

Definition 5.3.20 The logic preserving degrees of truth of the variety DMA is the
logic LDMA over LDMA whose consequence relation |=DMA is defined as follows:
for every set  {} LDMA ,  |=DMA if and only if, there is some finite set
0  such that, for every U DMA and for every homomorphism h : LDMA U,

{h( ) : 0 } h(). In particular, |=DMA if and only if h() = 1 for
every U DMA and for every homomorphism h : LDMA U. 

Now, recalling that p = ( p p) defines a consistency operator on each

def

TMA, the following results can be stated:


Theorem 5.3.21 The logic M4m (or, equivalently, MB or, equivalently, LTMA ) of

the variety of TMAs over the signature  is a C-system based on LBDL w.r.t.
and p.

Proof In order to prove that the logic LTMA conservatively extends the logic LBDL ,
observe that the variety BDL is
generated by B2 , the BDL -reduct of the Boolean
algebra A2 with domain 0, 1 . Thus, suppose that  {} LBDL is such that
 |=BDL . Then, for everyfinite subset 0 of  there exists a homomorphism
h 0 : LBDL B2 such that {h 0 ( ) : 0 }  h 0 (). Fix a finite subset 0 of
 and a homomorphism h 0 : LBDL B2 such that {h 0 ( ) : 0 }  h 0 ().
By employing a technique similar to the one used in the proof of Lemma 3.3.15,
the homomorphism h 0 : LBDL B2 can be extended to a homomorphism h :
LTMA A 
2 , where A2 is the TMA obtained from  A2 by adding the operation
x = x for every x. This homomorphism is such that {h( ) : 0 }  h()
and so  |=TMA . Therefore, LTMA is a conservative extension of LBDL .
By adapting the proof of Theorem 5.3.6 to the extended signature, it follows easily

that the logic M4m is a strong LFI w.r.t. and p. Clearly, LBDL is not trivial. Thus,
it remains to prove that cannot be defined in LBDL . Let ( p) be a formula over
BDL which depends exclusively on the propositional variable p. By induction on

the complexity of ( p), it can be proven that, in M4m , either ( p) is equivalent to
p, or ( p) is equivalent to , or ( p) is equivalent to  (exercise to the reader).

Therefore, ( p) is not equivalent to p in M4m . 

Theorem 5.3.22 The logic M4m (or, equivalently, MB or, equivalently, LTMA ) of

the variety of TMAs over the signature  is a dC-system based on LBDL w.r.t.
and p.

Proof In Proposition 5.3.7 it was proved that p = ( p p) is the unique con-


sistency operator definable in M4m . The same argument can be applied to M4m . It
224 5 LFIs based on Other Logics

will now be proved that the operator p cannot be defined in the algebra M 4m using
the signature DMA = {, , , , }. Let ( p) be a formula over DMA which
depends exclusively on the propositional variable p. By induction on the complexity
of ( p), it can be proven that the truth-table of ( p) in M 4m coincides with the
truth-table in M4m of one of the following formulas: p, , , p, p p, p p
(exercise to the reader). From this, it follows that it is impossible to define p in the
algebra M 4m using such signature. This shows that Item (iii.b) of Definition 3.3.5
is satisfied. Analogously it can be proved that Item (iii.a) of Definition 3.3.5 is also
satisfied. The rest of the proof follows from Theorem 5.3.21. 
The last result constitutes an interesting example of a dC-system which lies outside
the scope of the systems studied in Chap. 3.

5.3.4 The Contrapositive Implication

As it can be observed, the original signature of TMAs (and so the signature of

logic M4m ) does not include an implication as a primitive connective. It is a natural
question to ask how to define a binary operator in TMAs, in terms of the others, with
the behavior of a deductive implication.
Several implication connectives for TMAs were proposed in the literature. I.
Loureiro proposed in [44] the following implication operator for TMAs:

x y = x y.
def

By its turn, Figallo and Landini introduced in [49] an interesting implication operator
for TMAs, called contrapositive implication, that can be defined as follows (see [47]):

x ! y = (x y) (y x) ((x y) (x y)).

def

The truth-table in M4m of the contrapositive implication is as follows:

1 B N 0

1 1 B N 0
B 1 1 N B
N 1 B 1 N
0 1 1 1 1

The main feature of the contrapositive implication is that it internalizes the con-
sequence relation (whenever just one premise is considered), as we shall see in
Theorem 5.3.24. Another important aspect of the contrapositive implication is that
all the operations of the TMAs can be defined in terms of ! and 0. In fact:
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 225

Proposition 5.3.23 (see [49]) In every TMA it holds:

(i) 1 = (0 ! 0),
(ii) x = (x ! 0),
(iii) x y = (x ! y) ! y,
(iv) x y = (x y),
(v) x = (x ! x).
Therefore, ! and 0 are enough to generate all the operations of a given TMA.

Let TMAc be the variety of TMAs expressed in the signature  c = {!, }

(see [47]). From the previous result, we can reformulate M4m and its logic M4m ,
as well as the logic LTMA , in the signature  c . We will denote by Mc4m , M4m
c
and
LTMAc the corresponding version of M4m , M4m and LTMA , respectively, over  c .
Obviously M4m c
coincides with LTMAc , since Mc4m generates the variety TMAc .
c
A weak version of the Deduction meta-theorem DMT can be proved in M4m with
respect to the contrapositive implication.

Theorem 5.3.24 ([49]) Let , L c . Then:

c iff
|= M4m |= M4m
c ! .

This means that the contrapositive implication ! internalizes the consequence rela-
c
tion of M4m whenever just one premise is considered. From the algebraic point of
view, ! internalizes the partial order of TMAs.
It is not possible to improve Theorem 5.3.24:
c
Proposition 5.3.25 ([47]) In M4m , both directions of DMT fail if more than one
premise is considered. In more precise terms:
(i) , |= M4m
c does not imply that |= c ! ,
M4m
c ! does not imply that , |= c .
(ii) |= M4m M4m

Proof (i) Consider the following formulas of L c (here, the connectives not belong-
ing to  c must be regarded as abbreviations, according to Proposition 5.3.23):
= p q ( p ! q) p, = q and = , where p and q are two differ-
def def def

ent propositional variables. Then h( ) = 0 = h( ), for every homomorphism

h : L c Mc4m . From is, , |= M4m
c . Consider now a homomorphism h such that

h( p) = N and h(q) = B. Then h() = N and h() = B and so N = h()  h( !

) = B ! 0 = B. Thus, |= M4mc ! .

(ii) Consider = p, = p and = , where p is a propositional variable.

def def def

c ! , since |= c . Let h : L c M
Then |= M4m M4m 
c
4m be a homomorphism
c .
such that h( p) = N . Then h( ) = N  0 = h( ) and so , |= M4m 
226 5 LFIs based on Other Logics

c
5.3.5 A Hilbert-Style Axiomatization of M4m

Because of the nice properties of the contrapositive implication !, in [47] (see

also [46]) a Hilbert-style axiomatization of the 4-valued modal logic M4m over the
signature  c was proposed, that is, an axiomatization of the logic M4m c
.
Previous to the definition of the calculus TM4 for M4m (which was called H4m
c

in [47]), some abbreviations in the algebra of formulas L c , which are defined accord-
ing to the ones introduced in Proposition 5.3.23, need to be introduced. By  we
mean the formula ! ; by we mean the formula ! (thus,  denotes );
denotes ( ! ) ! ; denotes ( ); and  is an abbreviation
for ( ! ).

Definition 5.3.26 Let TM4 be the Hilbert calculus defined over the language L c
as follows:

Axiom schemas:

( ! ) ! (A1)
! ( ! ) (A2)
( ) ! ( ) (A3)
! (A4)

( ! ( ! )) (A5)
(( ) ! ) ! (( ! ) ( ! )) (A6)
( ! ( ! )) ! ( ! ( ! )) (A7)
( ! ) ! (( ! ) ! (( ! ) ! ( ! ))) (A8)
( ! ) ! (( ! ) ! ( ! )) (A9)
 ! (A10)

Inference rules:

!
(MP) (Conj) (Nec)


Since the calculus TM4 intends to reflect the consequence relation preserving
degrees of truth instead of the truth-preserving semantics for TMA, the usual notion
of derivation in a Hilbert calculus does not work for TM4. In the case of core fuzzy
logics (recall Sect. 5.2), the inference rules were rewritten in order to deal with that
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 227

semantics (recall Proposition 5.2.16 and Definition 5.2.17), while the usual notion
of derivation captured the truth-preserving semantics over those algebras. In the case
of TM4, the usual notion of derivation reflects the truth-preserving semantics for
TMA, which defines a second logic associated to TMAs, as shown in [47]. On the
other hand, a suitable modification of the notion of derivation, usual in the context of
modal logics, will be sufficient to the purposes of capturing the consequence relation
preserving degrees of truth (see [47]).

Definition 5.3.27 (Derivations in TM4)

(1) A derivation in TM4 of a formula L c is a finite sequence of formulas
1 . . . n such that n is and every i is either an instance of an axiom, or it is the
consequence of the instance of some inference rule whose premises appear in the
sequence 1 . . . i1 . We say that is derivable in TM4, and we write TM4 , if
there exists a derivation of in TM4.

(2) Let  {} be a set of formulas in L c . We say that is derivable in TM4 from

, and we write  TM4 , if either TM4 or there exists a finite, nonempty subset
{1 , . . . , n } of  such that the formula

(1 (2 (. . . (n1 n ) . . .))) !

is derivable in TM4. 

Remark 5.3.28 It follows from Definition 5.3.27 that TM4 iff TM4 , for
every formula . Also, it is easy to prove that TM4 iff TM4 ! , showing
that a weak version of the deduction meta-theorem MTD holds in TM4, by the
very definition of derivability. This is the syntactical counterpart of Theorem 5.3.24.
However, as stated in Proposition 5.3.25, the general version of the Deduction meta-
c
theorem does not hold in M4m . As mentioned above, the proposed notion of derivation
from premises in TM4 defined in terms of theoremhood is analogous to the usual
one used in modal logics. It is worth noting that TM4 is a Tarskian, finitary and
structural consequence relation, as we shall see in Remark 5.3.33.

By using (Nec), (A10) and (MP), it is easy to prove that TM4 satisfies the fol-
lowing fundamental meta-theorem:

Proposition 5.3.29 For every formula in L c :

TM4 if and only if TM4 .

The proof of completeness of TM4 follows from the usual Lindenbaum-Tarski

construction. In this particular case, it is enough to consider the empty theory.
The detailed proof of the following results can be found in [47].

Proposition 5.3.30 Let TM4 L c L c be the relation defined as follows:

228 5 LFIs based on Other Logics

TM4 = (, ) L c L c : TM4 ! and TM4 ! .
def

Then TM4 is a logical congruence in TM4 (recall Definition 4.3.3).

Theorem 5.3.31 The Lindenbaum algebra L c /TM4 of TM4 is a tetravalent
modal algebra defined over the signature  c , in which the operators are given as
follows: || ! || = | ! | and 0 = ||, where | | denotes the equivalence
def def

class of the formula , for every L c .

Theorem 5.3.32 [Soundness and Completeness of TM4] The following conditions
are equivalent, for every subset  {} of L c :
(i)  TM4 ,
c .
(ii)  |= M4m
c
Proof As observed after Proposition 5.3.23, the logic M4m coincides with the logic
LTMAc . This fact will be used in the rest of the proof.

(i) (ii) (Soundness): It is immediate to check that every axiom of TM4 is valid in
c
M4m . On the other hand, if an instance of the premises of an inference rule is valid
c c
in M4m then the respective conclusion is also valid in M4m .
Now, assume that  TM4 . If TM4 , consider a derivation 1 . . . k of in
TM4. By induction on the lenght k of the derivation it is easy to prove that is
valid in M4mc
, by the observations above. On the other hand, if TM4 then, by
Definition 5.3.27, there exists a finite nonempty subset 0 = {1 , . . . , n } of  such
that TM4 , where is the formula

(1 (2 (. . . (n1 n ) . . .))) ! .

As observed above, is valid in M4m

c
, and so 0 |= M4m
c . Therefore,  |= c ,
M4m
proving the soundness of TM4.

(ii) (i) (Completeness): Suppose that  |= M4m c . Then there exists a finite subset

{1 , . . . , n } of  such that {1 , . . . , n } |= M4m

c .

Case 1: If n = 0 then |= M4mc and so h() = 1 for every homomorphism h : L c


U and every U TMAc . In particular, h() = 1 for every homomorphism h : L c
L c /TM4 , by Theorem 5.3.31. Let h : L c L c /TM4 be the canonical map
given by h() = ||, for every formula . Then h is a homomorphism and so || = 1.
In particular TM4  ! . But it is easy to prove that  is a theorem of TM4, and
then TM4 , by (MP). From this  TM4 , by definition of derivation in TM4.
n
Case 2: If n > 0, let be the formula 1 n . Given that h( ) = i=1 h(i )
for every homomorphism h : L c Mc4m , it follows that |= M4m c , because we

are assuming that {1 , . . . , n } |= M4m

c . From this, we get |= c ! . But then,
M4m
by the Case 1, it can be concluded that TM4 ! . Finally,  TM4 , by definition
of derivation in TM4. 
5.3 A Modal LFI Based on Belnap and Dunns Logic BD 229

Remark 5.3.33 As an immediate corollary of the soundness and completeness The-

orem 5.3.32, it follows that the consequence relation TM4 generated by TM4 intro-
duced in Definition 5.3.26(2) is Tarskian, finitary and structural, that is, standard (see
Definition 2.1.4).
It is possible to give a direct proof of this fact without using the soundness and
completeness theorem. This can be done by adapting to TM4, together with the pair
(!, ), a general result (Theorem 2.10.2) concerning entailment systems proved
in [32]. The details of this alternative proof are left to the interested reader.

5.4 Paraconsistent Modalities, Consistency

and Determinedness

Finally, in this section some LFIs based on positive modal logics (called anodic
systems) will be briefly described. The resulting systems are called cathodic modal
systems.
Modal logics and paraconsistent logics are cousins. In 1948, while attempting to
answer a question posed by J. Lukasiewicz, S. Jaskowski presented a first formal
system for a paraconsistent logic with his discussive logic (recall the beginning of
Sect. 4.4.3). Interestingly enough, his logic was framed in terms of modalities, and
later on it was proved to be a particular case of the family of LFIs. Paraconsistent
negation can also be regarded as a kind of modal operator, considering the fact that
the classical negation for posibility (and, a fortiori, for necessity) has a paraconsistent
behavior. Namely, the operator = defines a paraconsistent negation where,
def

as usual, denotes the classical negation. This relationship has been studied in [50],
both with respect to the standard modal logic S5 and to four-valued modal logics (see
also [51]). It is worth noting that the fact that defines a paraconsistent negation
was already observed in 1987 in [52], when a Kripke-style semantics was proposed
for Settes logic P1 (recall Sect. 4.4.4) based on Kripke frames for the modal logic T.
The consistency operator of LFIs can also be conceived with a modal flavor: a
modal version of the propositional consistency operator in the LFIs symbolized as
", can be defined in any normal non-degenerate modal logic (given that  is either
primitive or definable in the language) by

" := 

or, equivalently, by " = , in the cases where  is defined as

def

(recalling that here means ).

It is also possible (as in [53]) to start from a purely positive modal logic of
consistency, having in the language the classical propositional connectives , ,
230 5 LFIs based on Other Logics

and , plus a consistency connective as primitive, interpreted in a Kripke-style

relational model as:

M, w |= " iff M, w |= implies: for all w

, if wRw
then M, w
|= .

These definitions depend, of course, on the interpretation of , as well as on the

interpretation of implication (and, in certain cases, negation as well). This makes
clear that the modal-defined notions of consistency are relative to the underlying
logic, as investigated in [54], where some ways of defining consistency and negation
when expressed in modal terms are explored.
A more intimate relationship between LFIs and their modal versions is inves-
tigated by Bueno-Soler in [55], where the so-called anodic modal systems (purely
positive modal systems) introduced in [56] are extended by adding certain paracon-
sistent axioms based on LFIs, defining a class of modal systems called cathodic
modal systems. This class consists of modal paraconsistent systems, dealing with
certain kinds of conflicting modal situations. A semantic interpretation is also given
in [55], where it is shown that the cathodic systems can be semantically characterized
in two different ways: by means of Kripke-style semantics and by means of modal
possible-translations semantics.
In order to offer a brief sketch of the main ideas behind the cathodic modal
systems, let  = {, , , , , } be the signature where cathodic systems
are constructed. The consistency connective of the LFIs plays a crucial role in
the systems, as it will be combined with modal operators. Consider the following
propositional axioms and rules:
 
(Ax1)
      
(Ax2)
  
(Ax3)
 
(Ax4)
 
(Ax5)
 
( ) (( ) ) (AxImp3)
( ) (Ax10)
( ( )) (bc1)
(cf)
( ) (ci)

(MP)

5.4 Paraconsistent Modalities, Consistency and Determinedness 231

The system PC is a fragment of the propositional calculus PC, introduced by

Henkin in [57], and is composed by the axioms (Ax1), (Ax2), (AxImp3) and by
the rule (MP). Observe that PC is defined over the signature  which contains
exclusively the implication . The disjunction connective is defined in PC as:

( ) = ( ) .
def

The (constructive) completeness of PC is easily proved with respect to PC -

valuations, which are functions v : L {0, 1} satisfying the following clause:

(vImp) v( ) = 1 iff v() = 0 or v() = 1.

The system PC, is obtained from PC by adding the conjunction (obtaining

the signature  ) and the schema axioms (Ax3), (Ax4) and (Ax5). It is sound
and complete w.r.t. valuations v : L {0, 1} satisfying clause (vImp) plus the
following:
(vAnd) v( ) = 1 iff v() = 1 and v() = 1.

The main useful results valid in PC, are the Deduction meta-theorem (DMT)
and the distributivity of over .
The minimal anodic modal systems, K and K, , are obtained, respectively,
from PC and PC, by adding an unary connective  together with the following
axiom and inference rule:

( ) ( ) (K)


(Nec)
 
Both systems are sound and complete w.r.t. a suitable Kripke semantics (see [56]).
Observe that these systems are positive, and so they can be used as a basis for modal
LFIs, as we shall see below.
The dual modality must be introduced separately in anodic modal systems,
because of the lack of a (dualizing) negation. Thus, the following additional axioms
should be considered:

( ) ( ) (K1)
( ) ( ) (K2)
( ) ( ) (K3)

The bi-modal system obtained from K, by adding to the signature plus the
axioms above is denoted by K,, , and it can also be characterized by a Kripke
semantics. All details about the anodic modal systems appear in [56].
232 5 LFIs based on Other Logics

From PC, it is possible to obtain several well-known paraconsistent systems:

the system PI (recall Sect. 3.3 of Chap. 3), introduced by Batens in [58], can be
obtained from PC, by adding the axiom (Ax10); the system mbC is obtained
from PI by adding a consistency operator plus the schema axiom (bc1)10 ; and,
as usual, systems bC and Ci are obtained from mbC by adding the schema axioms
(bC) and (Ci), respectively.
As analyzed in Sect. 3.3, system PI (also called mbC ) fails to be an LFI, since
it cannot define the consistency operator (recall Proposition 3.3.19): it is a conse-
quence of the fact that it is impossible to express inside that logic, as proved in
Proposition 3.3.18.
Recall from Sect. 2.4 of Chap. 2 that a classical negation (or strong negation) can
be defined in mbC, for every , as

= ( ( )).
def

From this definition all the relevant properties of classical negation are derivable,
which is useful in showing several intuitive results about cathodic systems involving
classical negation.
A PI-valuation is a function v : LPI {0, 1} (where PI is the signature of PI
without disjunction) satisfying the clauses (vImp) and (vAnd) plus the following:
(vNeg) v() = 0 implies v() = 1.
It is known that PI is sound and complete w.r.t. PI-valuations (recall Proposi-
tion 3.3.14). By the very definition, and taking into consideration that disjunction is
a defined connective, the mbC-valuations (recall Definition 2.2.1 from Chap. 2) are
PI-valuations satisfying, additionally, the following:
(vCon) v() = 1 implies v() = 0 or v() = 0.

Valuations for bC and Ci are obtained from mbC-valuations as defined in Chap. 3.

Another semantic characterization for the paraconsistent systems PI, mbC, bC and
Ci can be attained w.r.t. possible-translations semantics (see Sect. 6.8 of Chap. 6).
The systems PI, mbC, bC and Ci are used in [55] to define the classes of cathodic
systems.
The minimal normal cathodic modal system11 PI, is obtained from K,,
by adding a negation plus axiom (Ax10). Thus, it is a paraconsistent modal logic
whose positive (or anodic) modal base is K,, . Analogously, the paraconsistent

10 To be rigorous, it is obtained a version of PI and mbC in a signature without disjunction, where

can be expressed by ( ) , as mentioned above. It can be proved that the axioms
for given in Definition 2.1.12, as well as axiom (Ax8), are derivable in PC , and so they hold in
this version of mbC, as expected.
11 A modal system is classified as normal if it contains the Distribution Axiom (K) and the Neces-

sitation Rule (Nec) among its axioms and rules, and as minimal if it has only (K) as a modal axiom
and only (Nec) as a modal rule. In this sense, the above defined systems are the minimal normal
modal extension of each propositional basis.
5.4 Paraconsistent Modalities, Consistency and Determinedness 233

bi-modal systems mbC, , bC, , and Ci, can be defined from the positive
modal logic K,, by adding, besides a negation , a consistency operator plus
the corresponding axioms, as observed above.
Since, as noted, from the system mbC onwards it is possible to define a classical
negation , then the possibility operator is definable in such systems from the
necessity operator , as is usual in modal logic:

= 
def

The axioms (K1)(K3) are thus innocuous in mbC, , bC, , and Ci, , as the
reader can verify. Therefore, only the system PI, is indeed a genuine bi-modal
system.
Proofs of completeness and incompleteness for the classes of cathodic systems
extended with each specific modal axiom in the standard Lemmon-Scott schema
Gk,l,m,n :
k l m n (Gk,l,m,n )

(where m and n mean the iteration of modalities), are given in detail in [55].
Modal logics have also been built on top of other paraconsistent logics, as in [16],
Nelsons logic N4 (analyzed in Sect. 5.1.3 above) is taken as a non-modal system on
which a paraconsistent modal logic is defined, and studied in detail.
Negationless (i.e., positive) normal modal logics with box-like and diamond-like
operators are studied in [59], and gradually extended by adding a modal paraconsis-
tent negation and a modal paracomplete negation, and also by adding modal opera-
tors that internalize appropriate notions of consistency and determinedness into the
object-language level.
Deontic logics with paraconsistent genetics have been also studied in the literature
since a long time (see, for instance, [60]), and deontic counterparts of LFIs, the log-
ics of deontic (in)consistency (LDIs), have been introduced in [61]. These logics are
shown to be able to handle deontic paradoxes, as the well-known Chisholms para-
dox. Since contradictory obligations do not trivialize such LDIs, several paradoxes
involving conflicting obligations are dissolved (see [62]).

References

1. Mendelson, Elliot. 1987. Introduction to Mathematical Logic, 3rd ed. New York: Chapman &
Hall.
2. Johansson, Ingebrigt. 1937. Der Minimalkalkl, ein reduzierter intuitionistischer Formalismus
(in German). Compositio Mathematica 4: 119136. http://www.numdam.org/item?id=CM_
1937__4__119_0.
3. Kolmogorov, Andrei N. 1925. O printsipe tertium non datur (in Russian). Matmaticeskij
Sbornik 32(4): 646667. Translated to English as On the principle of excluded middle.
In [Heijenoort, Jean van, ed. 1967. From Frege to Gdel; A Source Book in Mathematical
Logic, 18791931. Harvard University Press], pp. 414437.
234 5 LFIs based on Other Logics

4. Rasiowa, Helena. 1974. An algebraic approach to non-classical logics. Studies in Logic and
the Foundations of Mathematics, vol. 78. Amsterdam: North-Holland.
5. Carnielli, Walter A, and Joo Marcos. A taxonomy of C-systems. In [Carnielli, Walter A.,
Marcelo E. Coniglio, and Itala M. L. DOttaviano, eds. 2002. Paraconsistency: The Logical
Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP
2000), Vol. 228 of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New
York], pp. 194.
6. Carnielli, Walter A, Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Incon-
sistency. In Handbook of Philosophical Logic (2nd edition), ed. Dov M. Gabbay and Franz
Guenthner, vol. 14, 193. Springer. doi:10.1007/978-1-4020-6324-4_1.
7. Humberstone, Lloyd. 2011. The connectives. Cambridge, MA: MIT Press.
8. Nelson, David. 1949. Constructible falsity. The Journal of Symbolic Logic 14(1): 1626.
9. Nelson, David. 1959. Negation and separation of concepts in constructive systems. In Construc-
tivity in Mathematics. Proceedings of the colloquium held at Amsterdam, 1957. Studies in Logic
and the Foundations of Mathematics, ed. Arend Heyting, vol. 39, 208225. North-Holland.
10. Almukdad, Ahmad, and David Nelson. 1984. Constructible falsity and inexact predicates. The
Journal of Symbolic Logic 49(1): 231233.
11. Odintsov, Sergei P. 2003. Algebraic semantics for paraconsistent Nelsons logic. Journal of
Logic and Computation 13(4): 453468.
12. Thomason, Richmond H. 1969. A semantical study of constructible falsity. Zeitschrift fr
mathematische Logik und Grundlagen der Mathematik 15: 247257.
13. Odintsov, Sergei P. 2004. On representation of N4-lattices. Studia Logica 76(3): 385405.
14. Fidel, Manuel M. 1978. An algebraic study of a propositional system of Nelson. In Mathe-
matical logic. Proceedings of the first Brazilian conference on mathematical logic, Campinas
1977, eds. Ayda I. Arruda, Newton C. A. da Costa and Rolando Chuaqui, Lecture Notes in
Pure and Applied Mathematics, vol 39, 99117. Marcel Dekker.
15. Vakarelov, Dimiter. 1977. Notes on N-lattices and constructive logic with strong negation.
Studia Logica 36(12): 109125.
16. Rivieccio, Umberto. 2011. Paraconsistent modal logics. Electronic Notes in Theoretical Com-
puter Science 278: 173186.
17. Rivieccio, Umberto. 2014. Implicative twist-structures. Algebra Universalis 71(2): 155186.
18. Ono, Hiroakira, and Umberto Rivieccio. 2014. Modal twist-structures over residuated lattices.
Logic Journal of the IGPL 22(3): 440457.
19. Odintsov, Sergei P. 2008. Constructive negations and paraconsistency. Trends in Logic, vol.
26. Springer.
20. Fidel, Manuel M. An algebraic study of logic with constructive negation. In [Arruda, Ayda I.,
Newton C. A. da Costa, and Antonio M. A. Sette, eds. 1980. Proceedings of the Third Brazilian
Conference on Mathematical Logic, Recife 1979. Sociedade Brasileira de Logica, Campinas],
pp. 119129.
21. Fidel, Manuel M. 1977. The decidability of the calculi Cn . Reports on Mathematical Logic 8:
3140.
22. Galatos, Nikolaos, and James G. Raftery. 2004. Adding involution to residuated structures.
Studia Logica 77(2): 181207.
23. Busaniche, Manuela, and Roberto Cignoli. 2009. Residuated lattices as an algebraic semantics
for paraconsistent Nelsons logic. Journal of Logic and Computation 19(6): 10191029.
24. Ramos, Fernando M., and Vctor L. Fernndez. 2009. Twist-structures semantics for the logics
of the hierarchy I n P k . Journal of Applied Non-Classical Logics 19(2): 183209.
25. Fernndez, Vctor L., and Carina Murciano. 2013. A generalization of twist-structures seman-
tics for n-valued logics. CLE e-Prints 13(1). http://www.cle.unicamp.br/e-prints/vol_13,n_1,
2013.html.
26. Hjek, Petr. 1998. Metamathematics of Fuzzy Logic. Trends in Logic, vol. 4. Dordrecht: Kluwer
Academic Publishers.
27. Esteva, Francesc, and Llus Godo. 2001. Monoidal t-norm based logic: Towards a logic for
left-continuous t-norms. Fuzzy Sets and Systems 124(3): 271288.
References 235

28. Coniglio, Marcelo E., Francesc Esteva, and Llus Godo. 2014. Logics of formal inconsistency
arising from systems of fuzzy logic. Logic Journal of the IGPL 22(6): 880904. doi:10.1093/
jigpal/jzu016.
29. Ertola, Rodolfo, Francesc Esteva, Tommaso Flaminio, Llus Godo, and Carles Noguera. 2015.
Paraconsistency properties in degree-preserving fuzzy logics. Soft Computing 19(3): 531546.
doi:10.1007/s00500-014-1489-0.
30. Cintula, Petr, Petr Hjek, and Carles Noguera, eds. 2011. Handbook of Mathematical Fuzzy
Logic (in 3 volumes), vols. 3739. Studies in Logic, Mathematical Logic and Foundations.
London: College Publications.
31. Fodor, Jnos. 1995. Nilpotent minimum and related connectives for fuzzy logic. In Fuzzy
systems, 1995. International joint conference of the fourth IEEE international conference on
Fuzzy systems and the second international fuzzy engineering symposium, 20772082.
32. Wjcicki, Ryszard. 1988. Theory of logical calculi: Basic theory of consequence operations.
Synthese Library, vol. 199. Dordrecht: Kluwer Academic Publishers.
33. Bou, Flix, Francesc Esteva, Josep Maria Font, ngel Gil, Llus Godo, Antoni Torrens, and
Ventura Verd. 2009. Logics preserving degrees of truth from varieties of residuated lattices.
Journal of Logic and Computation 19(6): 10311069.
34. Blok, Willem J. and Don Pigozzi. 2001. Abstract algebraic logic and the deduction theorem.
http://www.math.iastate.edu/dpigozzi/papers/aaldedth.pdf.
35. Michael Dunn, J. 1966. The algebra of intensional logics. Ph.D. thesis, University of Pittsburgh,
USA.
36. Michael Dunn, J., and Greg Restall. 2002. Relevance Logic. In Handbook of Philosophical
Logic (2nd edition), eds. Dov M. Gabbay and Franz Guenthner, vol. 6, 1136. Kluwer Academic
Publishers.
37. Belnap, Nuel D. 1976. How a computer should think. In Gilbert Ryle, editor, Contemporary
Aspects of Philosophy, pages 3056. Oriel Press.
38. Belnap, Nuel D. 1977. A useful four-valued logic. In Modern Uses of Multiple-Valued Logic:
Invited Papers from the Fifth International Symposium on Multiple-Valued Logic held at Indiana
University, Bloomington, Indiana, May 13-16, 1975, eds. J. Michael Dunn and George Epstein,
Episteme, vol. 2, 537. Reidel.
39. Ginsberg, Matthew L. 1988. Multivalued logics: A uniform approach to inference in artificial
intelligence. Computational Intelligence 4: 265316.
40. Fitting, Melvin. Bilattices in logic programming. In [Epstein, George, ed. 1990. Proceedings
of the Twentieth International Symposium on Multiple-Valued Logic, Charlotte, NC, USA. The
IEEE Computer Society Press], pp. 238246.
41. Arieli, Ofer, and Arnon Avron. 1996. Reasoning with logical bilattices. Journal of Logic,
Language and Information 5(1): 2563.
42. Font, Josep Maria. 1997. Belnaps four-valued logic and De Morgan lattices. Logic Journal of
the IGPL 5(3): 129. doi:10.1093/jigpal/5.3.1-e.
43. Monteiro, Luiz. 1963. Axiomes indpendants pour les algbres de ukasiewicz trivalentes (in
French). Bulletin de la Societ des Sciences Mathmatiques et Physiques de la R. P. Roumanie,
Nouvelle Srie, 7: 199202.
44. Loureiro, Isabel. 1983. lgebras Modais Tetravalentes (Tetravalent modal Algebras, in Por-
tuguese). Ph.D. thesis, Faculdade de Cincias, Universidade de Lisboa, Portugal.
45. Font, Josep Maria, and Miquel Rius. 2000. An abstract algebraic logic approach to tetravalent
modal logics. The Journal of Symbolic Logic 65(2): 481518.
46. Figallo, Martn. 2013. Hipersecuentes y la Lgica Tetravalente Modal T ML (Hypersequents
and the Tetravalent Modal Logic T ML, in Spanish). Ph.D. thesis, Universidad Nacional del
Sur, Argentina.
47. Coniglio, Marcelo E., and Martn Figallo. 2014. Hilbert-style presentations of two logics asso-
ciated to tetravalent modal algebras. Studia Logica 102(3): 525539. doi:10.1007/s11225-013-
9489-0.
48. Coniglio, Marcelo E., and Martn Figallo. 2014. On a four-valued modal logic with deductive
implication. Bulletin of the Section of Logic 43(12): 118.
236 5 LFIs based on Other Logics

49. Figallo, Aldo V., and Paolo Landini. 1995. On generalized I-algebras and 4-valued modal
algebras. Reports on Mathematical Logic 29: 318.
50. Bziau, Jean-Yves. 2002. S5 is a paraconsistent logic and so is first-order classical logic.
Logical Studies 9: 301309.
51. Bziau, Jean-Yves. 2005. Paraconsistent logic from a modal viewpoint. Journal of Applied
Logic 3: 714.
52. de Arajo, Ana L, Elias H. Alves, and Jos A. D. Guerzoni. 1987. Some relations between
modal and paraconsistent logic. The Journal of Non-Classical Logic, 4(2): 3344. http://www.
cle.unicamp.br/jancl/.
53. Marcos, Joo. 2005. Nearly every normal modal logic is paranormal. Logique et Analyse
48(189192): 279300.
54. Bueno-Soler, Juliana, and Walter A. Carnielli. 2014. Experimenting with consistency. CLE
e-Prints 14(3). http://www.cle.unicamp.br/e-prints/vol_14,n_3,2014.html.
55. Bueno-Soler, Juliana. 2010. Two semantical approaches to paraconsistent modalities. Logica
Universalis, 4(1): 137160. http://www.cle.unicamp.br/e-prints/vol_9,n_6,2009.html.
56. Bueno-Soler, Juliana. 2009. Completeness and incompleteness for anodic modal logics. Journal
of Applied Non-Classical Logic, 4(5): 291310. http://www.cle.unicamp.br/e-prints/vol_9,n_
5,2009.html.
57. Henkin, Leon. 1949. The completeness of the first-order functional calculus. The Journal of
Symbolic Logic 14(3): 159166.
58. Batens, Diderik. 1980. Paraconsistent extensional propositional logics. Logique et Analyse
9091: 195234.
59. Dod, Adriano, and Joo Marcos. 2014. Negative modalities, consistency and determinedness.
Electronic Notes in Theoretical Computer Science 300: 2145. doi:10.1016/j.entcs.2013.12.
010.
60. da Costa, Newton C.A., and Walter A. Carnielli. 1986. Paraconsistent deontic logics.
PhilosophiaPhilosophical Quarterly of Israel 16(34): 293305.
61. Coniglio, Marcelo E. 2009. Logics of deontic inconsistency. Revista Brasileira de Filosofia,
233: 162186. http://www.cle.unicamp.br/e-prints/vol_7,n_4,2007.html.
62. Coniglio, Marcelo E., and Newton M. Peron. 2009. A paraconsistentist approach to Chisholms
paradox. Principia 13(3): 299326.
Chapter 6
Semantics of Non-deterministic Character
for LFIs

The present chapter is devoted to analize alternative semantics for the traditional
LFIs presented in Chaps. 2 and 3. The non-standard semantical frameworks proposed
here were developed by diverse authors in order to deal with the hard task to give a
semantical interpretation to this kind of logic.
By adapting the proof of a result for da Costas logic C1 due to Mortensen (see
Theorem 3, p. 698 in [1]) it can be proven that neither Cila, the version of C1
defined over the signature  (see Remark 3.6.3 in Chap. 3) nor its subsystems, such
as mbC, admit a non-trivial logical congruence. In view of such circumstances,
these systems are not algebraizable even in the broader sense of Blok and Pigozzi
(see [26]). An alternative proof of the above mentioned result for C1 can be found
in [7], where a (finite) logical matrix which is a model of C1 is exhibited, in such
a way that it possesses two different C1 -filters which are mapped by the so-called
Leibiniz operator (see [2]) to the same congruence (namely, the identity relation).
Consequently, the Leibiniz operator on this algebra is not injective, hence C1 is not
algebraizable in Blok and Pigozzis sense.1 This argument can be easily generalized
to Cila and its subsystems such as mbC (see Theorem 3.83, p. 80 in [8]).
Because of this kind of result, several LFIs resist bravely to the semantical analy-
sis based on standard tools, like categorial or algebraic semantics. This is why the
development of alternative semantical techniques for these logics is an indispensable
task. The main feature presented by all the semantical approaches examined in this
chapter is its non-deterministic character. Non-determinism can be manifested by
using relations instead of functions to interpret a non-truth-functional connective,
as in the case of F-structures (see Sect. 6.1). Alternatively, non-deterministic logical
matrices can be defined (in which some connectives are interpreted as multioperators
instead of operators, see Sects. 6.36.6). In turn, possible-translations semantics, in
which a logic is translated into a family of simpler logics, generalizes the previous
non-deterministic semantical approaches (see Sect. 6.8).

1 Asproved in [2] (Theorem 4.2, p. 35), if a logic is algebraizable then its Leibniz operator is
necessarily injective.
Springer International Publishing Switzerland 2016 237
W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_6
238 6 Semantics of Non-deterministic Character for LFIs

Some of the original results presented in this chapter were adapted from the
preprint [9].

6.1 Fidel Structures for mbC

In this section, a semantics of Fidel structures will be proposed for the basic logic
mbC, together with the corresponding soundness and completeness theorem.
Recall the signatures  = {, , , , } and + = {, , } from Defin-
ition 2.1.3 of Chap. 2. In Sect. 5.1.3 of Chap. 5 implicative lattices were consid-
A = A, , , , 1 where A, , , 1 is a lattice with top
ered: they are algebras
element 1 such that {c A : a c b} exists for every a, b A. The binary

operator (implication) is such that a b = {c A : a c b} for every
def

a, b A. If A has also a bottom element 0, it is a Heyting algebra. The following

result is well-known:

Proposition 6.1.1 Let A = A, , , , 0, 1 be a Heyting algebra, that is, an

implicative lattice with a bottom element 0. Then, A is a Boolean algebra if and only
if a (a b) = 1 for every a, b A.

From the last proposition, and taking into account that mbC can be seen as an
extension of CPL (recall Sect. 2.5 of Chap. 2), we propose now an adaptation of the
semantics introduced by M. Fidel in 1979 for system N , whose S. Odintsovs version
for system N4 was briefly analized here in Sect. 5.1.3 of Chap. 5:

Definition 6.1.2 (Fidel structures for mbC) A Fidel structure for mbC, or simply
an F-structure for mbC, is a triple
 
E = A, {Na }aA , {Oa }aA

such that A = A, , , , 0, 1 is a Boolean algebra (see Proposition 6.1.1) and

where {Na }aA and {Oa }aA are families of nonempty subsets of A such that, for
every a, b A, the following holds:
(F1) if b Na then a b = 1;
(F2) if b Na then BCab
= , where
 
BCab = c Oa : a b c = 0 .

The intuitive reading for b Na and c Oa is that b is a possible negation a

of a, and that c is a possible consistency a of a coherent with b Na . This is
justified by the following definition:
6.1 Fidel Structures for mbC 239
 
Definition 6.1.3 A valuation over an F-structure E = A, {Na }aA , {Oa }aA for
mbC is a map v : L A satisfying the following properties, for every formu-
las and :
(1) v(#) = v()#v(), for # {, , };
(2) v() Nv() ;
(3) v() BCv()v() .


It is worth noting that, since v() Nv() (by item (2) of the last definition)
then BCv()v()
= , by item (F2) of Definition 6.1.2, and so item (3) of the last
definition is well-defined.
The semantical consequence relation associated to F-structures for mbC is thus
defined as expected:
 
Definition 6.1.4 Let  {} L , and let E = A, {Na }aA , {Oa }aA be a Fidel
structure for mbC. We say that is a semantical consequence of  (w.r.t. Fidel
structures for mbC), denoted by  mbC
F , if, for every F-structure E for mbC and
for every valuation v over E: v() = 1 whenever v( ) = 1 for every . 

Example 6.1.5 Let us briefly analyze the finite family of F-structures for mbC over
the two-elements Boolean algebra A2 with domain A2 = {0, 1}. Each of such struc-
tures is of the form  
E = A2 , N0 , N1 , O0 , O1 

such that
= Ni , Oi A2 (for i = 0, 1). Since 0 / N0 (because of item (F1) of
Definition 6.1.2) then N0 = {1}. On the other hand, there are three possibilities for
defining N1 and Ok (for k = 0, 1), namely: N10 = Ok0 = {0}, N11 = Ok1 = {1}, and
N12 = Ok2 = {0, 1}. By item (F2) of Definition 6.1.2, if 1 N1 then 0 O1 and so
O1
= {1} in this case. Thus, all the admissible F-structures for mbC over A2 are the
following:
 j 
E0i j = A2 , N0 , N10 , O0i , O1  , for 0 i, j 2;
 j 
Eki j = A2 , N0 , N1k , O0i , O1  , for 1 k 2, 0 i 2, j {0, 2}.

It will be shown in Theorem 6.2.30 below that the finite family of Fidel structures
for mbC over A2 described above semantically characterizes the logic mbC, and
the same result can be extended to other systems studied in this chapter.

The soundness of mbC w.r.t. Fidel structures is easy to prove:

Theorem 6.1.6 (Soundness of mbC w.r.t. Fidel structures) Let  {} L be a

set of formulas. Then,
 mbC implies  mbCF .
240 6 Semantics of Non-deterministic Character for LFIs
 
Proof Let E = A, {Na }aA , {Oa }aA be a Fidel structure for mbC, and let v :
L A be a valuation over it. If is an instance of an schema axiom of mbC,
it can be proven that v() = 1. In fact, if is an instance of an axiom of mbC
other than (Ax10) or (bc1), then it is immediate that v() is a term of the Boolean
algebra A which is always 1, by item (1) of Definition 6.1.3. For example, if is
an instance of axiom (Ax1), v() is a term of the form x (y x), which is
always 1. Now, if is an instance of axiom (Ax10) then v() Nv() ,
by item (2) of Definition 6.1.3. But then, by clause (F1) of Definition 6.1.2 and
item (1) of Definition 6.1.3, v() = v( ) = v() v() = 1. Finally, if is
an instance ( ( )) of axiom (bc1) then v() v() v() =
v() v() v() = 0, by item (3) of Definition 6.1.3, and by commutativity
and associativity of the infimum . From this, and by using item (1) of Definition 6.1.3
and the algebraic properties of Boolean algebras,

v() = v() (v() (v() v( ))) = (v() v() v()) v( ) = 1

since 0 x = 1 for every x, in every Boolean algebra.

Concerning MP observe that, if v() = 1 and v( ) = 1 then v() = 1, by
item (1) of Definition 6.1.3 and the algebraic properties of Boolean algebras.
Suppose, now, that  mbC . From all the facts stated above it is easy to prove,
by induction on the length of a derivation in mbC of from  that, if v( ) = 1 for
every , then v() = 1. This shows that  mbC F , as required. 

In order to prove completeness, some notions and results will be useful.

Let  be a non-trivial a theory in mbC (that is, a set of formulas  L such
that  mbC for some formula ). Consider the following relation mbC between
formulas in L :

mbC
 iff  mbC and  mbC .

It is easy to prove that mbC

 is an equivalence relation.
 Moreover, it ismbC
a congruence

w.r.t. the connectives of + of CPL+ . Let /mbC = L  : be the
  
equivalence class of L and let L /mbC 
= /mbC 
: L be the set of
all the equivalence classes. As  is a congruence w.r.t. + , the operations
mbC

/mbC # /mbC = (#)/mbC for # {, , }

def

  

are well-defined. Moreover, the following holds:

Proposition 6.1.7 The structure

def  
AmbC
 = L /mbC

, , , , 0 , 1

where 0 = ( p1 p1 p1 )/mbC and 1 = ( p1 p1 )/mbC

def def

 
, is a Boolean
algebra.
6.1 Fidel Structures for mbC 241

Proof As mentioned
 in Sect. 5.1.3, it was proved
 in [10] that the Lindenbaum
F-structure L / , , , , ( p1 p1 )/ for N4 is an implicative lattice. But,
over + , N4 coincides with IPL+ , while mbC coincides with CPL+ (which is an
axiomatic extension of IPL+ ). Consequently, and since p1 p1   p1 p1
mbC

(because of axiom (Ax10)), it follows by the very definitions that L /mbC , , ,

 
, 1 is an implicative lattice. Moreover, 0 is a bottom element of it, as an
immediate consequence of axiom (bc1) of mbC (Recall Definition 2.1.12) and
the definition of the operations in AmbC
 . Finally, because of axiom (Ax9) and the
definition of the operations in AmbC
 , it follows that AmbC
 is a Boolean algebra,
by Proposition 6.1.1. 
Definition 6.1.8 (Lindenbaum F-structures for mbC) Let  and AmbC as above.
The Lindenbaum F-structure for mbC over  is defined as follows:
 
EmbC = AmbC
 , {Na }aL /mbC , {Oa }aL /mbC
 

where, for each a L /mbC


,
   
Na = ()/mbC : a and Oa = ()/mbC : a .
def def

 


Proposition 6.1.9 For every non-trivial theory , the structure EmbC is an
F-structure for mbC.
Proof Let a L /mbC

. Then, a = /mbC
, for some formula . Suppose that
()/mbC

N a . Then, a, that is, mbC
 , and so a = /mbC

= /mbC

.
From this,

a ()/mbC

= /mbC

()/mbC

= ( )/mbC

= 1

and so condition (F1) of Definition 6.1.2 is verified.

Finally, suppose that ()/mbC

Na . Then, a and so a = /mbC

, while
()/mbC

O a is such that
 
a ()/mbC ()/mbC = /mbC ( )/mbC = ( )/mbC .
    

But mbC  p1 p1 p1 , therefore condition (F2) of Defin-

ition 6.1.2 is also verified. 
We are now ready to prove completeness:
Theorem 6.1.10 (Completeness of mbC w.r.t. Fidel structures) Let  {} L
be a set of formulas. Then,

 mbC
F implies  mbC .
242 6 Semantics of Non-deterministic Character for LFIs

Proof Assume that  mbC

F . By Proposition 6.1.9, EmbC is an F-structure for mbC.
Consider now the mapping v : L L /mbC 
given by v() = /mbC 
. Then v is a
valuation over E , by the very definition of A and the last part of the proof of
mbC mbC

Proposition 6.1.9. Moreover, v() = 1 iff  mbC , for every formula . Hence
v() = 1 for every  and so, by hypothesis, it follows that v() = 1 . That is,
 mbC . 

6.2 Fidel Structures for Some Extensions of mbC

The semantics of F-structures provided for mbC can be extended to all the axiomatic
extensions of mbC defined in Chap. 3. In this section we show how to define such
structures for just some paradigmatic extensions of mbC.
We begin with the simplest extension of mbC, the logic mbCciw introduced in
Definition 3.1.1, which is obtained from mbC by adding the axiom schema

( ) (ciw)

Definition 6.2.1 (Fidel structures for mbCciw) A Fidel structure for mbCciw, or
simply an F-structure for mbCciw, is an F-structure for mbC
 
E = A, {Na }aA , {Oa }aA

(recall Definition 6.1.2) satisfying, additionally, the following clause for every
a, b A:
(F3) if b Na then
 CIWab
= , where 
CIWab = c Oa : c (a b) = 1 .


 
Definition 6.2.2 Let E = A, {Na }aA , {Oa }aA be an F-structure for mbCciw. A
valuation over E is a map v : L A satisfying the clauses (1)(3) of Defini-
tion 6.1.3, plus the following property, for every formula :

(4) v() CIWv()v() .

Given that v() Nv() (by item (2) of Definition 6.1.3) then the set CIWv()v()
is nonempty, by item (F3) of Definition 6.2.1, and so item (4) of the last definition is
well-defined.
The semantical consequence relation mbCciw
F associated to F-structures for mbC-
ciw is defined analogously to the case of mbC
F for mbC, but now using valuations
over F-structures for mbCciw.
6.2 Fidel Structures for Some Extensions of mbC 243

The soundness of mbCciw w.r.t. Fidel structures is, again, easy to prove:
Theorem 6.2.3 (Soundness of mbCciw w.r.t. Fidel structures) Let  {} L
be a set of formulas. Then,

 mbCciw implies  mbCciw

F .

Proof The proof follows easily from that for mbC with minor modifications: in order
to prove that v() = 1 for every instance of axiom (bc1) (the only axiom of mbC
involving the consistency operator), it is enough to observe that any valuation over
a Fidel structure for mbCciw also satisfies condition (3) of Definition 6.1.3. From
this, the proof is the same that the one for Theorem 6.1.6.
Finally, if is an instance ( ) of axiom (ciw) then

v() = v( ( )) = v() (v() v()) = 1,

by items (1) and (4) of Definition 6.2.2. This completes the proof. 
In order to prove completeness, the Lindenbaum F-structure EmbCciw for mbCciw
over  is defined as in Definition 6.1.8, but now with respect to the logic mbCciw.
Proposition 6.2.4 For every non-trivial theory , the structure EmbCciw is an F-
structure for mbCciw.
Proof It is enough to prove that EmbCciw verifies condition (F3) of Definition 6.2.1.
Thus, suppose that b Na , where b = ()/mbC 
, and consider c = ()/mbC 
.
Then, c Oa . Since a = /mbC
then it is easy to see that c C I W ab , because
of axiom (ciw). This shows that EmbCciw satisfies (F3) and so it is an F-structure for
mbCciw. 
From this, completeness is immediate:
Theorem 6.2.5 (Completeness of mbCciw w.r.t. Fidel structures) Let  {}
L be a set of formulas. Then,

 mbCciw
F implies  mbCciw .

Proof Assuming that  mbCciwF , then EmbCciw is an F-structure for mbCciw, by

Proposition 6.2.4. It is enough to verify that the mapping v : L L /mbCciw

given
by v() = /mbCciw

is a valuation over E
mbCciw
. But this is immediate, by the very
definition of A
mbCciw
and the proof of Proposition 6.2.4. From this, it follows that
 mbCciw , as in the proof of Theorem 6.1.10. 
Observe that the definition of the sets C I Wab of clause (F3) describes the require-
ment for satisfying axiom (ciw). Let us analyze now another extension of mbC,
the logic mbCci (recall Definition 3.1.7), obtained from mbC by adding the axiom
schema:
244 6 Semantics of Non-deterministic Character for LFIs

( ) (ci)

In order to satisfy this axiom by suitable F-structures, the condition on the com-
panion sets Oa is necessarily a bit more complicated, given that a negation is applied
to a consistency operator in the formula .

Definition 6.2.6 (Fidel structures for mbCci) A Fidel structure for mbCci, or sim-
ply an F-structure for mbCci, is an F-structure for mbC
 
E = A, {Na }aA , {Oa }aA

satisfying, additionally, the following property:

(F4) if b Na then C Iab

= , where 
CIab = c Oa : d (a b) = 1 for some d Nc .


 
Definition 6.2.7 Let E = A, {Na }aA , {Oa }aA be an F-structure for mbCci. A
valuation over E is a map v : L A satisfying the properties (1)(3) of Defini-
tion 6.1.3 plus the following, for every formula :

(5) v() C Iv()v() ; 

(6) v() d Nv() : d (v() v()) = 1 .

Given that v() Nv() (by item (2) of Definition 6.1.3) then C Iv()v()
= ,
by item (F4) of Definition 6.2.6, and so item (5) of the last definition is well-defined.
On the other hand, since v() C Iv()v() (by item  (5) of the previous defini-
tion) then d Nv() : d (v() v()) = 1
= , by definition of the set
C Iv()v() , and then condition (6) of the last definition is well-defined.
The semantical consequence relation mbCci
F associated to F-structures for mbCci
is defined analogously to the case of mbCF for mbC, but now using valuations over
F-structures for mbCci.
The soundness of mbCci w.r.t. Fidel structures follows straightforwardly:

Theorem 6.2.8 (Soundness of mbCci w.r.t. Fidel structures) Let  {} L be

a set of formulas. Then,

 mbCci implies  mbCci

F .

Proof As it was done with mbCciw, it is enough to analyze the validity of

axiom (bc1), as well as the validity of the new axiom, in this case (ci). The proof
of the validity of (bc1) is identical to that for mbC, given that condition (3) of
Definition 6.1.3 holds.
6.2 Fidel Structures for Some Extensions of mbC 245

Finally, if is an instance ( ) of axiom (ci) then

v() = v( ( )) = v() (v() v()) = 1,

by items (1) and (5) of Definition 6.2.7. This completes the proof. 

As in the case of mbCciw, in order to prove completeness, the Lindenbaum

F-structure EmbCci for mbCci over  is defined as in Definition 6.1.8, but now with
respect to the logic mbCci.

Proposition 6.2.9 For every non-trivial theory , the structure EmbCci is an

F-structure for mbCci.

Proof It is sufficient to prove that the structure EmbCci satisfies condition (F4) of
Definition 6.2.6. Thus, suppose that b Na . Then, b = ()/mbC 
for some a
and so a = /mbC
. Let c = ()/ mbC and d = ()/ mbC . Then c Oa and d
 
Nc . Because of axiom (ci), d (a b) = 1 , where 1 = ( p1 p1 )/mbC 
. This
means that c C Iab , that is, C Iab
= , and then EmbCci satisfies (F4). Therefore it
is an F-structure for mbCci. 

From this, completeness is immediate:

Theorem 6.2.10 (Completeness of mbCci w.r.t. Fidel structures) Let  {} L

be a set of formulas. Then,

 mbCci
F implies  mbCci .

Proof Suppose that  mbCci

F . Then, EmbCci is an F-structure for mbCci, by Propo-
sition 6.2.9. As in the proof for mbCciw, it is enough to prove that the mapping
v : L L /mbCci

such that v() = /mbCci

is a valuation over EmbCci . But this is
immediate, by the definition of AmbCci
and the proof of Proposition 6.2.9. Therefore
 mbCci , as in the proof of Theorem 6.1.10. 

The following system to be analyzed is mbCcl (recall Definition 3.3.26), obtained

from mbC by adding the axiom schema:

( ) (cl)

Definition 6.2.11 (Fidel structures for mbCcl) A Fidel structure for mbCcl, or
simply an F-structure for mbCcl, is an F-structure for mbC
 
E = A, {Na }aA , {Oa }aA
246 6 Semantics of Non-deterministic Character for LFIs

satisfying, additionally, the following property:

(F5) if b Nathen CLab

= , where 
CLab = c Oa : d c = 1 for some d Nab .


 
Definition 6.2.12 Let E = A, {Na }aA , {Oa }aA be an F-structure for mbCcl. A
valuation over E is a map v : L A satisfying the properties (1)(3) of Defini-
tion 6.1.3 plus the following, for every formula :

 ;
(7) v() CLv()v() 
(8) v(( )) d Nv()v() : d v() = 1 .

As in the previous cases, it is easy to see that clauses (7) and (8) are well-defined.
The semantical consequence relation mbCclF associated to F-structures for mbCcl is
defined analogously to the previous cases.
By observing that v(( ) ) = 1 for every valuation v over an F-
structure for mbCcl, and for every formula , the soundness of mbCcl w.r.t. Fidel
structures can be proved as above:

Theorem 6.2.13 (Soundness of mbCcl w.r.t. Fidel structures) Let  {} L be

a set of formulas. Then,

 mbCcl implies  mbCcl

F .

Consider now the Lindenbaum F-structure EmbCcl for mbCcl over , defined
similarly to the previous cases. The following result is proved as above.

Proposition 6.2.14 For every non-trivial theory , the structure EmbCcl is an

F-structure for mbCcl.

Proof Given a non-trivial theory  in mbCcl, the Lindenbaum F-structure for mbCcl
over , denoted by EmbCcl , is defined analogously to EmbC (recall Definition 6.1.8)
with the obvious modifications. Since mbCcl extends mbC, EmbCcl is an F-structure
for mbC. If a L /mbCcl

then, a = /mbCcl

, for some formula . Let b Na . Then,
b = ()/mbCcl

where a, that is, mbCcl
 , and so a = /mbCcl

and a b =
( )/mbCcl

. Thus, ()/ mbCcl Oa and (( ))/ mbCcl Nab are such

that (( ))/mbCcl

()/mbCcl 
= 1 , because of axiom (cl). This shows that
EmbCcl satisfies property (F5) of Definition 6.2.11, being therefore an F-structure for
mbCcl. 

From this, the proof of completeness is obtained as in the previous cases:

6.2 Fidel Structures for Some Extensions of mbC 247

Theorem 6.2.15 (Completeness of mbCcl w.r.t. Fidel structures) Let  {} L

be a set of formulas. Then,

 mbCcl
F implies  mbCcl .

Proof Consider the canonical mapping v : L L /mbCcl

given by v() = /mbCcl

.
Since v(( )) v() = 1 , the mapping v is a valuation over E mbCcl
. The
rest of the proof is analogous to that of Theorem 6.1.10 above. 

It is easy to prove that mbCcl is sound and complete w.r.t. Fidel-structures over the
2-elements Boolean algebra A2 (and, of course, the same holds for mbC, mbCciw
and mbCci). Let mbCcl F2 be the restriction of the consequence relation mbCcl
F to
F-structures over A2 . That is, for every set of formulas  {} L ,  mbCcl F2
iff, for every F-structure E for mbCcl over A2 and for every valuation v over E, if
v[] {1} then v() = 1.

Theorem 6.2.16 (Soundness and Completeness of mbCcl w.r.t. Fidel structures

over A2 ) Let  {} L be a set of formulas. Then,

 mbCcl iff  mbCcl

F2 .

Proof Suppose that  mbCcl . Then,  mbCcl

F , by Theorem 6.2.13. In particular,
 mbCcl
F2 .
Now, suppose that  mbCcl . By completeness of mbCcl w.r.t.
mbCcl-valuations, there exists an mbCcl-valuation v : L {0, 1} (recall Defi-
nition 3.3.26) such that v[] {1} and v() = 0.
Consider the structure Ev = A2 , N0 , N1 , O0 , O1  such that, for every a
{0, 1},
Na = {v() : v() = a}

Oa = {v() : v() = a} .

Then, N0 = O0 = {1}, N1 = {0, 1} and 1 O1 , because of the properties of v.

Moreover, v() Nv() and v() Ov() for every formula . Let us prove
that Ev is an F-structure for mbCcl over A2 . It is easy to see that items (F1)
and (F2) of Definition 6.1.2 are satisfied: if b Na then there exists such that
a = v() and b = v(), and so a b = v() v() = 1. Let c = v(). Then
c Oa such that a b c = 0. That is, Ev is an F-structure for mbC over A2 .
Now, given a, b, c as above, let d = v(( )). Then, d Nab such that
d c = v(( ) ) = 1. Therefore item (F5) of Definition 6.2.11 is sat-
isfied and so Ev is an F-structure for mbCcl over A2 . It is immediate to see that v is
a valuation over Ev such that, by hypothesis, v[] {1} and v() = 0. This shows
that  mbCcl
F2 , as required. 
248 6 Semantics of Non-deterministic Character for LFIs

Consider now the following law of double-negation:

(cf)

As mentioned in Sect. 3.5 of Chap. 3, the following systems involving (cf) have
been investigated in the literature:
- The system bC is obtained from mbC by adding axiom (cf).
- The system Ci is obtained from mbC by adding axioms (ci) and (cf).
- The system Cil is obtained from Ci by adding axiom (cl).
It is easy to define F-structures adequate to the logic bC:

Definition 6.2.17 (Fidel structures for bC) A Fidel structure for bC, or simply an
F-structure for bC, is an F-structure for mbC
 
E = A, {Na }aA , {Oa }aA

satisfying, additionally, the following property:

(F6) if b Nathen C Fab

= , where
C Fab = c Nb : c a = 1 .


 
Definition 6.2.18 Let E = A, {Na }aA , {Oa }aA be an F-structure for bC. A valu-
ation over E is a map v : L A satisfying the properties (1)(3) of Definition 6.1.3
plus the following, for every formula :

(9) v() C Fv()v() .

As in the previous cases, we obtain the following:

Theorem 6.2.19 (Soundness and completeness of bC w.r.t. Fidel structures) Let

 {} L be a set of formulas. Then,

F if and only if  bC .

 bC

Concerning the logic Ci, it is enough to combine the previous results:

Definition 6.2.20 (Fidel structures for Ci) A Fidel structure for Ci, or simply an
F-structure for Ci, is an structure
 
E = A, {Na }aA , {Oa }aA
6.2 Fidel Structures for Some Extensions of mbC 249

which is simultaneously an F-structure for mbCci and for bC. That is, the structure
A = A, , , , 0, 1 is a Boolean algebra, and {Na }aA and {Oa }aA are families
of nonempty subsets of A satisfying the clauses (F1) and (F2) of Definition 6.1.2,
(F4) of Definition 6.2.6, and (F6) of Definition 6.2.17. 
 
Definition 6.2.21 Let E = A, {Na }aA , {Oa }aA be an F-structure for Ci. A valua-
tion over E is a map v : L A satisfying the properties (1)(3) of Definition 6.1.3,
(5) and (6) of Definition 6.2.7, and (9) of Definition 6.2.18. 
Theorem 6.2.22 (Soundness and completeness of Ci w.r.t. Fidel structures) Let
 {} L be a set of formulas. Then,

F if and only if  Ci .

 Ci

The logic Cil is analyzed similarly:

Definition 6.2.23 (Fidel structures for Cil) A Fidel structure for Cil, or simply an
F-structure for Cil, is an F-structure for Ci
 
E = A, {Na }aA , {Oa }aA

which is also an F-structure for mbCcl. That is, the families {Na }aA and {Oa }aA
satisfy, additionally, the clause (F5) of Definition 6.2.11. 
 
Definition 6.2.24 Let E = A, {Na }aA , {Oa }aA be an F-structure for Cil. A valu-
ation over E is a valuation considering E as an F-structure for Ci satisfying, besides
the properties of Definition 6.2.21, the properties (7)(8) of Definition 6.2.12. 
Theorem 6.2.25 (Soundness and completeness of Cil w.r.t. Fidel structures) Let
 {} L be a set of formulas. Then,

F if and only if  Cil .

 Cil

Consider now the axioms for propagation of consistency analyzed in Sect. 3.6 of
Chap. 3. We will only consider here the propagation axioms for the connectives in
the signature + of CPL+ , namely

( ) ( ) (ca )
( ) ( ) (ca )
( ) ( ) (ca )

This is justified by the fact that, in extensions of Cil, the propagation of over
the connectives and is provable, and so it is not necessary to postulate it.
Recall from Remark 3.6.3 that the logic Cila is obtained from Cil by adding axiom
schemas (ca ), (ca ) and (ca ). This logic can be seen as the version of da Costas
system C1 (see Definition 3.7.4) over the signature  (see [8, 11]).
250 6 Semantics of Non-deterministic Character for LFIs

Definition 6.2.26 (Fidel structures for Cila) A Fidel structure for Cila, or simply
an F-structure for Cila, is an F-structure for Cil
 
E = A, {Na }aA , {Oa }aA

in which the families {Na }aA and {Oa }aA satisfy, additionally, the following clauses,
for # {, , }:

(F7)# if a  Oa and 
 b Ob then C
#
Aaba  b
= , where

C Aaba  b = c Oa#b : (a b ) c = 1 .
# 


 
Definition 6.2.27 Let E = A, {Na }aA , {Oa }aA be an F-structure for Cila. A val-
uation over E is a valuation considering E as an F-structure for Cil satisfying, besides
the properties of Definition 6.2.24, the following properties for every formulas and
, and for # {, , }:

(10)# v((#)) C A#v()v()v()v() .


Theorem 6.2.28 (Soundness and completeness of Cila w.r.t. Fidel structures) Let
 {} L be a set of formulas. Then,

 Cila
F if and only if  Cila .

Remark 6.2.29 In [12] it was proposed an alternative definition of semantics of F-

structures for the family of da Costas systems Cn . In particular, a new semantics of
F-structures for C1 was given, constituting an alternative to the one introduced above
for Cila (taking into account that, rigorously speaking, Cila and C1 are not the same
logic).
Finally, it is clear that Theorem 6.2.16 can be generalized to all the systems
analyzed in this section:
Theorem 6.2.30 (Soundness and Completeness of LFIs w.r.t. Fidel structures
over A2 ) Let L {mbC, mbCciw, mbCci, mbCcl, bC, Ci, Cil, Cila}, and let 
{} L be a set of formulas. Then,

 L iff  L
F2

where L F2 denotes the consequence relation generated by Fidel structures for L

over A2 .
Clearly, the last result has as an immediate consequence the decidability of all
these system.
6.2 Fidel Structures for Some Extensions of mbC 251

Corollary 6.2.31 Let L {mbC, mbCciw, mbCci, mbCcl, bC, Ci, Cil, Cila}.
The consequence relation of L (by using finite premises) is decidable by means
of Fidel structures over A2 .

6.3 Non-deterministic Matrices

This section briefly describes another non-deterministic semantical framework

suitable to deal with LFIs, the non-deterministic matrices (or Nmatrices). Non-
deterministic matrices were introduced in 1962 by Rescher in [13], under the name of
quasi-truth-functional systems. Afterwards, this notion was independently proposed
by Kearns (in 1981, see [14]) and Ivlev (in 1988, see [15]) as a way to overcome
Dugundjis results by means of an alternative to Kripke semantics. Kearns introduces
four-valued Nmatrices with a restriction on the valuations (called level valuations)
for several modal systems, while Ivlev proposes four-valued Nmatrices for weaker
versions of modal systems T and S5. These results have been revisited in [16], with
the addition of some generalizations.
The Nmatrix semantics was reintroduced (together with the terminology non-
deterministic matrices and Nmatrices) by Avron and Lev in [17] (see also [18]),
and afterwards intensively developed, from the point of view of applications, by A.
Avron and his collaborators.
As we shall in the next sections, Nmatrices are closely related to possible-
translations semantics and swap structures, to be defined below.
An Nmatrix is a multialgebra (also called hyperstructure or hyperalgebra or non-
deterministic algebra) together with a set of designated values of its domain.
Non-deterministic algebras are algebraic structures having at least one multivalued
operation (that is, a multifunction), which is called an hyperoperation. This kind of
structure constitute an useful tool in computer science in order to deal with non-
deterministic processes (non-deterministic automaton, for instance). Historically,
multivalued functions were firstly proposed in mathematics in the 18th century, within
the study of the logarithm of negative numbers (see Sect. 9.2.7 of Chap. 9).
The study of hyperstructuresfrom the point of view of Abstract Algebrabegan
in 1934 when F. Marty proposed the notion of hypergroups (see [19]). Since then,
several hyperstructures have been proposed in the literature besides hypergroups:
hyperlattices, hyperrings, hyperfields, and hyperalgebras in general.
Recall that, given two sets A and B, a multivalued function f from A to B is a
function f : A (B) in the usual sense. If f and g are multivalued functions from
A to B and from B to C, respectively, then its composite g f is the multivalued
function from A to C given by
 
(g f )(a) = g(b) : b f (a)
def

for every a A. The composition of multifunctions is associative. Of course a func-

tion f : A B can be seen as a multifunction from A to B such that the set f (a)
is a singleton, for every a A.
252 6 Semantics of Non-deterministic Character for LFIs

Nmatrices constitute a natural application of the theory of multialgebras to Logic.

The formal definition of Nmatrices proposed in [17] is as follows (compare with
Definition 4.1.2)
Definition 6.3.1 Let  be a propositional signature. A non-deterministic matrix (or
Nmatrix) for  is a structure M = A, D, O such that A is a nonempty set (of
truth-values), D A is a nonempty set (of designated values) and O is a mapping
which assigns to every n-ary connective c of  a function cM : An (A)
{}.2 A valuation over M is a mapping v : L A such that v(c(1 , . . . , n ))
cM (v(1 ), . . . , v(n )) for every n-ary connective c of , i L (for i = 1, . . . , n)
and n N. The consequence relation |=M induced by M is defined as follows: let
 {} L ; then  |=M if, for every valuation v over M, v[] D implies
v() D. A logic L over  is sound (respectively, complete) for M if, for every
 {} L it holds:  L only if (respectively, if)  |=M . The logic L is
adequate for M if it is sound and complete for M. 
Remark 6.3.2 Given an Nmatrix M = A, D, O, observe that it has associated an
unique multialgebra A = A, O. Accordingly, any Nmatrix M will be alternatively
defined as M = A, D, where A is a multialgebra and D is a nonempty subset of
the domain of A such that any multioperation in A always returns a nonempty set of
values for any input.
Example 6.3.3 In [20] Avron proposed an Nmatrix for mbC called M5B , defined as
follows: its domain A is  
BA2 = T, t, t0 , F, f 0
 
while the set D of designated values is {T, t, t0 }. Let N D = F, f 0 be the set
of non-designated truth-values; finally, the multioperations of M5B are defined as
follows:
Table 6.1

2 Or, equivalently, a multivalued function cM from An to A such that cM (a)

= , for every a A.
6.3 Non-deterministic Matrices 253

It was proved in [20] that mbC is adequate for M5B .

Example 6.3.4 Consider the following non-deterministic matrix MCi proposed
in [21] for the logic Ci: A = {T, t, F}; D = {T, t}; and O is defined by the multi-
operations below.

Then MCi semantically characterizes the logic Ci, as shown in [21].

Recall the Dugundji-like theorems for LFIs obtained in Sect. 4.2 of Chap. 4,
showing that most of the LFIs studied in the previous chapters cannot be char-
acterizable by a single finite logic matrix. Under the light of these limitations, the
characterization of several LFIs by single finite Nmatrices obtained by Avron and
his collaborators shows that this semantical framework is a valuable tool, and consti-
tutes an elegant and natural generalization of the matrix semantics. Avron himself,
in [22], has met some limitations on the expressive powers of Nmatrices: it is possible
to establish a Dugundji-like theorem for some LFIs, even with respect to Nmatrices.
Theorem 6.3.5 No logic between mbCcl and Cila can be characterized by a single
finite Nmatrix.3
In the next section one more non-deterministic semantical approach to LFIs will
be introduced, the so-called swap structures. These structures are inspired by twist
structures, and they are representable by Fidel structures, but the latter are more
general than the former, as we shall see. Basically, swap structures are multial-
gebras defined over suitable subsets of B 3 , where B is any Boolean algebra, and
so they naturally determine families of Nmatrices which semantically characterize
several LFIs.

6.4 Swap Structures for mbC

In [10], Odintsov shows that Fidel structures and twist structures for N4 are equivalent
in a strong sense: for every F-structure E for N4, and for every valuation v over it, there

3 In fact, Avrons original result encompasses even more systems, not considered in this book.
254 6 Semantics of Non-deterministic Character for LFIs

exists a twist-structure E and a valuation v over it such that 1 (v ()) = v(),
for every formula . Conversely, given a twist structure B for N4 and a valuation v
over it, there exists an F-structure EB for N4, and a valuation vF over it such that
1 (v()) = vF (), for every formula . This pair of results is the key for the proof
of soundness and completeness of N4 w.r.t. twist structures, from the corresponding
adequacy result for F-structures.
In this section we will adapt this techique to obtain a class of hyperstructures for
mbC resembling twist structures, which induces an adequate semantics, in terms of
non-deterministic matrices, for this logic. We do not speak about twist semantics
directly, since a key feature of this kind of structure, namely the use of the opposite
order of the underlying lattice (which justifies its name) is missing in the present
structures. In addition, swap structures are hyperalgebras instead of ordinary alge-
bras. Moreover, the elements of the swap structures for mbC are triples instead of
pairs, since we have now two (logically independent) non-truth-functional connec-
tives instead of just one. We will return to this point later, on Remark 6.4.4.
From now on, given a set A and the cartesian product A A A with canonical
projections 1 , 2 and 3 , the following notation will be used: xi will stand for i (x)
for any x A A A and i = 1, 2, 3.
Definition 6.4.1 Let A = A, , , , 0, 1 be a Boolean algebra (see Proposi-
tion 6.1.1), and let

BA = {x A A A : x1 x2 = 1 and x1 x2 x3 = 0} .

A swap structure for mbC over A is any multialgebra over  (see Sect. 6.3)

B = B, , , , , 

such that B BA and where the multioperations are defined as follows, for every x
and y in B:
(i) x#y = {z B : z 1 = x1 #y1 }, for each # {, , };
def

(ii) x = {z B : z 1 = x2 };
def

(iii) x = {z B : z 1 = x3 }.
def

The unique swap structure for mbC over A with domain BA will be denoted
by BA . 
Remark 6.4.2
(1) It is worth noting that, given a Boolean algebra A and a A, always exist some
b, c A such that a b = 1 and a b c = 0; for instance, b can be taken as
the Boolean complement of a and then c can be arbitrary. Thus, the output of the
multioperators are nonempty sets and so it is always possible to construct the swap
structure BA . Moreover, B = {(1, 0, 1), (0, 1, 1)} is the domain of the unique swap
structure for mbC defined over a given Boolean algebra.
(2) Observe that swap structures over the two-elements Boolean algebra A2 are
defined over subsets of
6.4 Swap Structures for mbC 255
 
BA2 = (1, 1, 0), (1, 0, 1), (1, 0, 0), (0, 1, 1), (0, 1, 0) .

The elements of a given swap structure are called snapshots. This terminology
is inspired by its use in computer systems to refer to states. Accordingly, a triple
(a, b, c) of a swap structure B keeps track simultaneously of the truth-value a of a
given formula , the value b of , and the value c of . We will return to this point
in Remark 6.4.4 below.
Given that any swap structure is a multialgebra, the consequence relation over
swap structures will be defined by means of non-deterministic matrices, in analogy
with the corresponding notion for twist structures (recall Definition 5.1.36).

Definition 6.4.3 (Nmatrices associated to swap structures for mbC)

(1) Given a Boolean algebra A and a swap structure B for mbC over A with domain
B, let  
D B = x B : x1 = 1 .

The non-deterministic matrix associated to B is MmbC (B) = B, D B ,

def

where |=MmbC (B) is the corresponding consequence relation (recall Definition 6.3.1).4
The Nmatrix associated to BA will be denoted by MmbC A . In particular
MmbC = B , .
def
A2 A2 D B A2
(2) Let KmbC be the class of all the Nmatrices defined by swap structures for mbC,
that is:
 
KmbC = MmbC (B) : B is a swap structure for mbC over A, for some A .

The semantical consequence relation w.r.t. swap structures for mbC is the conse-
quence relation generated by the class KmbC of Nmatrices. That is, if  {} is a set
of formulas of mbC, then:  |=KmbC iff  |=MmbC (B) , for every swap structure B
for mbC. 

Given a swap structure B for mbC, a valuation over B is a valuation over the
Nmatrix MmbC (B), recall Definition 6.3.1. By the definition above, it is immediate
to see that  |=KmbC iff, for every swap structure B and for every valuation v over
B: 1 (v()) = 1 whenever 1 (v( )) = 1 for every . Observe that this notion
is similar to the original definition of twist structures semantics for N4 presented in
Definition 5.1.36, but now considering Nmatrices instead of matrices.

Remark 6.4.4
(1) A few words are needed concerning the structures defined above. Firstly, they
are not twist structures: as mentioned, the dual order structure of A is irrelevant
in this context, and so no truth-value is twisted here by means of the negation of
A. However, we swap in a certain manner the components of the snapshots by
using the negation and the consistency operator (in the first case, we swap the second

4 From this point on, the notation for Nmatrices established in Remark 6.3.2 will be used.
256 6 Semantics of Non-deterministic Character for LFIs

component putting it in the first place; in the second case, the third component is
reallocated to the first place). This justifies the name adopted for these multialgebras.
Despite the differences, there are some analogies between swap structures and
the twist structures already proposed in the literature for other logics. Moreover,
the analogy with twist structures is additionally supported by the fact, to be proved
below, that they are essentially equivalent to the F-structures for mbC proposed in
Sect. 6.1. This feature is analogous to the exact correspondence between F-structures
and twist structures for N4 pointed out at the beginning of this section.
(2) The intuitive reading of a snapshot (x, y, z) of BA is as follows: x can be seen as
the truth-value of a sentence ; y is a possible truth-value for the sentence , while
z is a possible truth-value for the sentence . We could say that only the first coor-
dinate of a snapshot in B has a (deterministic, or expected) algebraic behavior, while
the other two coordinates have a non-deterministic (or unexected) behavior. Given a
snapshot (x, y, z), if x represents the truth-value of then y, the first coordinate of
every snapshot in (x, y, z), represents a possible truth-value of while z, the first
coordinate of every snapshot of (x, y, z), represents a possible truth-value of ,
justifying so the definition of and in B adopted here. The fact that swap struc-
tures are multialgebras reflects the non- determinateness (or non-truth-functionality)
underlying any semantics of the logic mbC, and so we justifiably speak about a
possible truth-value of and instead of the truth-value of these formulas:
by fixing a truth-value for , all the admissible snapshots will produce the plethora
of possibilities for the truth-values of and . In [20], Avron already considered
triples of this kind in order to obtain a semantics of Nmatrices for several LFIs, and
so this notion of snapshots is not new. However, as we shall see at the end of Sect. 6.6,
algebraic structures defined by snapshots formed by n-tuples of formulas (in which
twist structures constitute a particular case) were already proposed by Fidel in [23].
(3) Recall from Remark 6.4.2(2) that
 
BA2 = (1, 1, 0), (1, 0, 1), (1, 0, 0), (0, 1, 1), (0, 1, 0) .

The first three values are designated (since the first coordinate is 1), while the others
are non-designated. Observe that the five truth-values correspond to the five valua-
tions v1 -v5 shown in the Table 2.1 displayed just before Theorem 2.3.2, in Chap. 2.
The truth-value t = (1, 1, 0) reflects a scenario (or valuation) in which a proposition
and its negation are both true, and so is false (according to the axiom bc1).
By its turn, the truth-value T = (1, 0, 1) is classically true: it says that is true,
is false and so is true. But there exists another possibility in mbC, the weakest LFI
presented in Chap. 2: the truth-value t0 = (1, 0, 0), which says that is true and
is false, but even so is false. As observed in Remark 3.1.5 of Chap. 3, the system
mbCciw and its extensions avoid this situation, and so a swap structure over A2 for
these logics would omit this truth-value, together with its dual f 0 = (0, 1, 0). Besides
f 0 , the non-designated truth-value F = (0, 1, 1) (dual to T ) is classical. Observe
that the snapshots in BA2 were renamed using the same names as the truth-values of
Avrons characteristic
 Nmatrix M5B for mbC presented in Example 6.3.3. This is why
we write BA2 = T, t, t0 , F, f 0 in that example. Moreover, it is easy to check that
6.4 Swap Structures for mbC 257
 
M5B coincides with MmbCA2 and so D = DBA2 = {T, t, t0 }, while N D = F, f 0 is
the set of non-designated truth-values. Recall that M5B was described in Table 6.1 of
Example 6.3.3.

It is worth noting that the Nmatrices generated by swap structures for mbC over
a non-trivial Boolean algebra A (that is, with 1
= 0) are non-trivial, that is, the set
of designated truth-values is a proper and nonempty subset of the domain:

Proposition 6.4.5 Let A be a Boolean algebra with domain A where 1

= 0, and let
B be a swap structure for mbC over A with domain B. Then:
(i) 1 [B] = {a A : (a, b, c) B for some b, c A}, with the operations inher-
ited from B, is a sub-Boolean algebra of A.
(ii)
= D B
= B and so MmbC (B) generates a non-trivial logic.

Proof (i) Firstly observe that, by the definition of the multioperations in B, 1 [x#y] =
{x1 #y1 } for # {, , }. Then, 1 [B] is closed under the operations , and
of A: if a, b 1 [B] then a#b 1 [B], for # {, , }.
Let x be a snapshot in B. By the definition of the multioperations in B, 1 [x
x] = {x1 x1 } = {1} and so 1 belongs to 1 [B]. Additionally, 1 [x x x] =
{x1 x2 x3 } = {0} and so 0 belong to 1 [B].5 Now, let a 1 [B], and let x be
a snapshot in B such that x1 = a. Then, Z = x (x x x) is a subset of B
such that 1 [Z ] = {a 0} and so a 0 belongs to 1 [B].6 This shows that 1 [B]
is also closed by Boolean complements and so it is a sub-Boolean algebra of A.
(ii) As shown in item (i), if x B then
= x x D B , and so D B
= . Now, let
z be a snapshot in B. If z 1
= 1 or z 2
= 1 then either z / D B or z D B = ,
where z
= ; in both cases, D B
= B. Now, suppose that z 1 = z 2 = 1. Since
z 1 z 2 z 3 = 0 then z 3 = 0. Therefore, z is a nonempty subset of B such that
1 [z] = {0} and so z D B = . This shows that always is the case that D B
= B, as
required. 

From now on, A will denote a Boolean

 algebra with domain
 A where 1
= 0.
Consider now an F-structure E = A, {Na }aA , {Oa }aA for mbC. Let BE be the
unique swap structure BA for mbC with domain BA , and let v be a valuation over E.
By Definition 6.1.3, (v(), v(), v()) BA for every formula . From this,
consider the mapping vS : L BA such that vS () = (v(), v(), v()), for
every formula . Observe that
(i) vS (#) = (v(#), v((#)), v((#))) vS ()#vS (),
for # {, , };
(ii) vS () = (v(), v(), v()) vS ();
(iii) vS () = (v(), v(), v()) vS ().

5 Observe that the expression x x x corresponds to the output of a composition of multifunc-

tions (where is written in infix notation) as defined in the beginning of Sect. 6.3.
6 Again, notice that the expression x (x x x) corresponds to the output of a composition

of multifunctions.
258 6 Semantics of Non-deterministic Character for LFIs

Thus, vS is a valuation over BE such that vS () DBA iff v() = 1, for every for-
mula .
Conversely, let B = B, , , , ,  be a swap structure for mbC. By Propo-
sition 6.4.5 (i), the set A = 1 [B] is the domain of a sub-Boolean algebra A of A.
Define a structure  
EB = A , {NaB }aA , {OaB }aA

such that, for every a A ,

NaB = {b A : (a, b, c) B for some c A },

OaB = {c A : (a, b, c) B for some b A }.

The sets above are always nonempty, by the definition of and . Since B BA ,
it is clear that EB is an F-structure for mbC. Now, given a valuation v over B,
define a mapping vF : L A as follows: vF () = 1 (v()), for every formula .
If , L then

vF (#) = 1 (v(#)) 1 [v()#v()]

= {1 (v())#1 (v())} = {vF ()#vF ()}

for # {, , }. Thus, vF (#) = vF ()#vF () and so vF satisfies clause (1) of

Definition 6.1.3. On the other hand,

vF () = 1 (v()) 1 [v()] = {2 (v())}.

Then, vF () = 2 (v()). But (1 (v()), 2 (v()), 3 (v())) B and so

2 (v()) NB1 (v()) = NvBF () , by definition of EB . Then vF () NvBF () . This
shows that vF satisfies clause (2) of Definition 6.1.3. Finally,

vF () = 1 (v()) 1 [v()] = {3 (v())}.

By a similar argument 3 (v()) BC1 (v())2 (v()) = BCvF ()vF () , showing that
vF satisfies clause (3) of Definition 6.1.3. That is, vF is a valuation over EB such that,
by the very definitions, v() D B iff vF () = 1, for every formula .
We have, thus, proved the following result:

Proposition 6.4.6 Let A be a Boolean algebra with domain A where 1

= 0.
(1) For every F-structure E for mbC based on A, and for every valuation v over E,
there exists a swap structure BE for mbC with domain BA and a valuation vS over
BE such that vS () DBA iff v() = 1, for every formula .
(2) For every swap structure B for mbC based on A, and for every valuation v
over B, there exist an F-structure EB for mbC and a valuation vF over EB such that
v() D B iff vF () = 1, for every formula .
6.4 Swap Structures for mbC 259

From this, we can easily prove the equivalence between F-structures semantics
and swap structures semantics for mbC:
Theorem 6.4.7 (Equivalence between F-structures and swap structures semantics
for mbC) Let  {} be a set of formulas in L . Then:

 mbC
F if and only if  |=KmbC .

Proof Suppose that  mbC F . Let B be a swap structure for mbC and let v be a
valuation over B such that v( ) D B for every . By Proposition 6.4.6(2) there
exist an F-structure EB for mbC and a valuation vF over EB such that v() D B
iff vF () = 1, for every formula . Then, vF ( ) = 1 for every  and so, by
hypothesis, vF () = 1. Therefore v() D B , proving that  |=KmbC .
The converse is proved analogously, but now using Proposition 6.4.6(1). 
Theorem 6.4.8 (Soundness and completeness of mbC w.r.t. swap structures) Let
 {} be a set of formulas in L . Then:

 mbC if and only if  |=KmbC .

Proof It is a direct consequence of Theorems 6.1.6, 6.1.10 and 6.4.7. 

By adapting the proof of Proposition 6.4.6(1) presented above, we can relate the
valuation semantics for mbC introduced in Definition 2.2.1 of Chap. 2 with swap
structures, in a clear way. Essentially, we will prove that valuation semantics corre-
sponds to the particular case of the (truth-functional) valuation semantics over the
Nmatrix MmbC A2 = BA2 , DBA2  associated to the unique swap structure BA2 defined
over the Boolean algebra A2 with domain A2 = {0, 1} (recall Definition 6.4.3).
Theorem 6.4.9 Let v be a valuation for mbC in the sense of Definition 2.2.1, and let
A2 be the Boolean algebra with domain A2 = {0, 1}. Then, v induces a valuation vS
over the Nmatrix MmbC
A2 such that vS () DBA2 iff v() = 1, for every formula .

Proof Given a valuation v for mbC, observe that (v(), v(), v()) BA2 for
every formula , by the properties of v (recall Definition 2.2.1). Define now the
mapping vS such that vS () = (v(), v(), v()), for every formula . Clearly,
(i) vS (#) = (v(#), v((#)), v((#))) vS ()#vS (),
for # {, , };
(ii) vS () = (v(), v(), v()) vS ();
(iii) vS () = (v(), v(), v()) vS ().
From this, it follows that vS is a valuation over the Nmatrix MmbC
A2 such that vS ()
DBA2 iff v() = 1, for every formula . 
Corollary 6.4.10 (Soundness and completeness of mbC w.r.t. the Nmatrix MmbC
A2
over A2 ) Let  {} be a set of formulas in L . Then:

 mbC if and only if  |=MmbC

A
.
2
260 6 Semantics of Non-deterministic Character for LFIs

Proof The only if part is a direct consequence of Theorem 6.4.8. For the if
part, suppose that  |=MmbCA2
, and let v be a valuation for mbC (in the sense of
Definition 2.2.1) such that v[] {1}. By Theorem 6.4.9, there exist a valuation vS
over the Nmatrix MmbCA2 such that vS () DBA2 iff v() = 1, for every formula .
Thus, vS ( ) DBA2 for every  and so vS () DBA2 , by hypothesis. But then
v() = 1. This shows that  mbC and so  mbC , by completeness of mbC
w.r.t. valuations. 

As observed in Remark 6.4.4(3), M5B coincides with the characteristic Nmatrix

MmbC
A2 presented in Table 6.1 of Example 6.3.3. The previous corollary, thus, is
nothing more than Avrons result presented in [20], reintroduced in the context of
swap structures. As an immediate consequence of this it follows:

Corollary 6.4.11 The consequence relation of mbC (by using finite premises) is
decidable by means of the unique swap structure over A2 with domain BA2 .

In the next section, swap structures will be defined for some important extensions
of mbC.

6.5 Swap Structures for Some Extensions of mbC

As it was done in Sect. 6.2 with the semantics of F-structures, in this section we will
extend the swap structures introduced above for some paradigmatic extensions of
mbC.
The first system to be analyzed is mbCciw. Recall from Definitions 6.2.1 and 6.2.2
the notions of F-structures and its valuations. It is worth noting that a snapshot
(a, b, c) for a swap structure for mbCciw should satisfy, besides the conditions
a b = 1 and (a b) c = 0, the following one: c (a b) = 1, which corre-
sponds to axiom (ciw). But, by definition of Boolean complement, this means that
c = (a b), where denotes the Boolean complement in the Boolean algebra A.
From this, the space of snapshots for mbCciw (and its extensions) can be described
by pairs (a, b) such that a b = 1, assuming tacitly that the third component is
(a b). This motivates the following definition:

Definition 6.5.1 Let A = A, , , , 0, 1 be a Boolean algebra, and let

A = {x A A : x 1 x 2 = 1} .
Bciw

A swap structure for mbCciw over A is any multialgebra over  (see Sect. 6.3)

B = B, , , , , 
6.5 Swap Structures for Some Extensions of mbC 261

such that B Bciw

A and where the multioperations are defined as follows, for every
x and y in B:
(i) x#y = {z B : z 1 = x1 #y1 }, for each # {, , };
def

(ii) x = {z B : z 1 = x2 };
def

(iii) x = {z B : z 1 = (x1 x2 )}.

def

Observe that B = {(1, 0), (0, 1)} is the domain of a swap structure for mbC-
ciw over any Boolean algebra, where the multioperations are defined as in Defini-
tion 6.5.1. This means that the notion of swap structures for mbCciw is nonvoid.
The concept of non-deterministic matrix semantics of swap structures for mbCciw
is similar to that of Definition 6.4.3, by considering now Nmatrices

MmbCciw (B) = B, D B 

def

defined from swap structures B for mbCciw with consequence relation denoted by
|=MmbCciw (B) . The class of all such Nmatrices is denoted by KmbCciw . The semantical
consequence relation w.r.t. swap structures for mbCciw is the consequence relation
|=KmbCciw generated by the class KmbCciw of Nmatrices. Given that any swap structure
for mbCciw can be seen as a swap structure for mbC (by extending each snapshot
(a, b) to the snapshot (a, b, (a b))), Proposition 6.4.5 states that the Nmatrices
generated by swap structures for mbCciw over a non-trivial Boolean algebra A are
non-trivial, that is, the set of designated truth-values is a proper and nonempty subset
of the domain. 
Given a non-trivial Boolean algebra A, consider an F-structure E = A, {Na }aA ,
{Oa }aA for mbCciw, and let v be a valuation over E. Let BE be the unique swap
structure BA mbCciw
for mbCciw with domain Bciw A . Clearly (v(), v()) BA ,
ciw

for every formula . Using this, consider the mapping vS : L BA such thatciw

vS () = (v(), v()), for every formula . Thus, by adapting the proof for mbC,
it can be seen that vS is a valuation over BE . The only difference concerns the
multioperator . By item (4) of Definition 6.2.2, v() (v() v()) = 1 and
so v() = (v() v()), which guarantees that vS () (vS ()). Finally, it
follows by the very definition that vS () DBciw A
iff v() = 1, for every formula .
Conversely, given a swap structure B = B, , , , ,  for mbCciw and a
valuation v over B, consider the structure
 
EB = A , {NaB }aA , {OaB }aA

such that A = 1 [B] and, for every a A ,

NaB = {b A : (a, b) B},

OaB = {(a b) A : (a, b) B for some b A }.

262 6 Semantics of Non-deterministic Character for LFIs

The sets above are always nonempty, by the definition of and . Since B Bciw A ,
and by the definition of the sets OaB , it is clear that EB is an F-structure for
mbCciw. Define now a mapping vF : L A as follows: vF () = 1 (v()), for
every formula . As it was proved for mbC, vF satisfies clauses (1)(3) of Def-
inition 6.1.3. Since (1 (v()), 2 (v())) B, vF () = 2 (v()) and vF () =
(1 (v()) 2 (v())), then vF () C I WvF ()vF () . From this, vF also satisfies
clause (4) of Definition 6.2.2 and then it is a valuation over EB such that v() D B
iff vF () = 1, for every formula .
We have proved the following result:

Proposition 6.5.2
(1) For every F-structure E for mbCciw and for every valuation v over E there exists
a swap structure BE for mbCciw with domain Bciw A and a valuation vS over BE such
that vS () DBciw
A
iff v() = 1, for every formula .
(2) For every swap structure B for mbCciw and for every valuation v over B there
exist an F-structure EB for mbCciw and a valuation vF over EB such that v() D B
iff vF () = 1, for every formula .

The following results are now easily proved, as in the previous section:

Theorem 6.5.3 (Equivalence between F-structures and swap structures semantics

for mbCciw) Let  {} be a set of formulas in L . Then:

 mbCciw
F if and only if  |=KmbCciw .

Theorem 6.5.4 (Soundness and completeness of mbCciw w.r.t. swap structures)

Let  {} be a set of formulas in L . Then:

 mbCciw if and only if  |=KmbCciw .

Additionally, the proof of Theorem 6.4.9 can be adapted to mbCciw, which will
lead to a characterization of mbCciw in terms of a single finite Nmatrix (see Corol-
lary 6.5.6 below). Firstly observe that, by definition,
 
A2 = (1, 1), (1, 0), (0, 1) .
Bciw

Consider now the Nmatrix MmbCciw

A2 induced by BAmbCciw
2
, the unique swap structure
for mbCciw with domain BA2 . The set of designated values is DBciw
ciw
A2
= {t, T }, where
t = (1, 1) and T = (1, 0), while F = (0, 1) is the only non-designated truth-value.
The multioperations are defined as follows:
6.5 Swap Structures for Some Extensions of mbC 263

Table 6.2

Theorem 6.5.5 Let v be a valuation for mbCciw, that is, a valuation for mbC
in the sense of Definition 2.2.1 which also satisfies clause (vCiw) (recall Sect. 3.1
of Chap. 3). Then, v induces a valuation vS over the Nmatrix MmbCciw
A2 such that
vS () DBciw
A
iff v() = 1, for every formula .
2

Proof The proof is just an adaptation of the one presented above for Theorem 6.4.9.
Thus, observe that, given a valuation v for mbCciw, the pair (v(), v()) is an
A2 , for every formula . Define now the mapping vS such that vS () =
snapshot in Bciw
(v(), v()), for every formula . From similar arguments than the ones given
in the proof of Theorem 6.4.9, it follows that vS is a valuation over the Nmatrix
MmbCciw
A2 . The consistency operator must be analyzed in a different way: from the
fact that v satisfies clause (vCiw), it follows that v() = (v() v()) and so
vS () (vS ()). Clearly, vS () DBciwA
iff v() = 1, for every formula . 
2

Corollary 6.5.6 (Soundness and completeness of mbCciw w.r.t. the Nmatrix

MmbCciw
A2 over A2 ) Let  {} be a set of formulas in L . Then:

 mbCciw if and only if  |=MmbCciw

A
.
2

Proof It is analogous to the proof of Corollary 6.4.10, but now using Theorem 6.5.4
(for the only if part) and Theorem 6.5.5, together with the completeness of mbCciw
w.r.t. valuations (for the if part). 

The notion of swap structures for mbCci (recall Definition 3.1.7) will now be
analyzed. In this logic, which extends mbCciw, the truth-value of can be
obtained from the truth-values of and , by considering the valuation seman-
tics introduced in Definition 3.1.7 of Chap. 3. Indeed, by the properties of such val-
uations, v() = v() v() while v() = (v() v()). In particular,
v() = v().

Definition 6.5.7 Let A = A, , , , 0, 1 be a Boolean algebra. A swap struc-

ture for mbCci over A is structure
264 6 Semantics of Non-deterministic Character for LFIs

B = B, , , , , 

satisfying all the properties of a swap structure for mbCciw over A (recall Defini-
tion 6.5.1), with the following restriction:

(x) = {((x1 x2 ), x1 x2 )}
def

for every x B. 

This means that, when considering swap structures for mbCci, the consistency
operator is interpreted as a function instead as a multifunction. The following result
justifies the restriction imposed to the consistency operator:

Proposition 6.5.8 Let B be a swap structure for mbCci. Then, for every x B 7 :

1 [x (x x)] = {1} .

Proof The easy proof is left as an exercise to the reader. 

Obviously, a swap structure for mbCci is also a swap structure for mbCciw, but
the converse is not necessarily true. The set B = {(1, 0), (0, 1)} is the domain of a
swap structure for mbCci over any Boolean algebra, where the multioperations are
defined as in Definition 6.5.7. Thus, always exist swap structures for mbCci. The
unique swap structure for mbCci with domain Bciw A will be denoted by BA
mbCci
.
The Nmatrix semantics of swap structures for mbCci is similar to that of Defini-
tion 6.4.3, by considering now Nmatrices defined from swap structures for mbCci

MmbCci (B) = B, D B 

def

whose consequence relation is denoted by |=MmbCci (B) . The class of all such Nmatrices
is denoted by KmbCci . Thus, the semantical consequence relation w.r.t. swap structures
for mbCci is the consequence relation |=KmbCci generated by the class KmbCci of non-
deterministic matrices. By adapting Proposition 6.4.5, it follows that the Nmatrices
generated by swap structures for mbCci over a non-trivial Boolean algebra A are
non-trivial, since the set of designated truth-values is a proper and nonempty subset
of the domain. The Nmatrix associated to BA mbCci
will be denoted by MmbCci
A . In
 
particular MA2 = BA2 , DBciw
mbCci def mbCci
A2
.
The following fundamental proposition can be obtained by extending the previous
results for mbC and mbCciw:

Proposition 6.5.9 Let A be a non-trivial Boolean algebra.

7 Observe again that the expression x (x x) corresponds to the output of a composition

of multifunctions (where and are written in infix notation) as defined in the beginning of
Sect. 6.3.
6.5 Swap Structures for Some Extensions of mbC 265

(1) For every F-structure E for mbCci over A, and for every valuation v over E,
there exists a swap structure BE for mbCci with domain Bciw A and a valuation vS
over BE such that vS () DBciw
A
iff v() = 1, for every formula .
(2) For every swap structure B for mbCci over A, and for every valuation v over
B, there exist an F-structure EB for mbCci and a valuation vF over EB such that
v() D B iff vF () = 1, for every formula .
 
Proof (1) Let E = A, {Na }aA , {Oa }aA be an F-structure for mbCci, and let v
be a valuation over E. Let BE be the unique swap structure BA mbCci
for mbCci
with domain BA . Clearly, (v(), v()) BA , for every formula . This allows
ciw ciw

to define the mapping vS : L Bciw A such that vS () = (v(), v()), for every
formula . Observe that, by condition (6) of Definition 6.2.7, v() (v()
v()) = 1 and so v() (v() v()) = 1, where denotes the Boolean
complement in A. Given that v() v() = 1 then v() v(), by defi-
nition of , whence v() (v() v()) = 1, by monotonicity of the supremum
. But v() (v() v()) = 0 and so v() = (v() v()), by defini-
tion of Boolean complement . This guarantees that vS () (vS ()) and so,
analogous to the proof for mbC, it can be seen that vS is a valuation over BE such
that vS () DBciw
A
iff v() = 1, for every formula .
(2) Consider a swap structure B = B, , , , ,  for mbCci and a valuation v
over B, and define the structure
 
EB = A , {NaB }aA , {OaB }aA

as in the previous section, recalling that A = 1 [B]. By the proof of Proposi-

tion 6.5.2, it follows that it is an F-structure for mbCciw.
Now, let b NaB . By definition of NaB , (a, b) B, and so c = (a b) OaB .
Since B is a swap structure for mbCci and (a, b) B then (c, a b) (a, b) B
and so a b NcB such that (a b) (a b) = 1. This means that EB satisfies
condition (F4) of Definition 6.2.6 and so it is an F-structure for mbCci.
Finally, consider the mapping vF : L A given by vF () = 1 (v()), for
def

every formula . As it was proved above, vF satisfies clauses (1)(3) of Defini-

tion 6.1.3. Given that (1 (v()), 2 (v())) B, vF () = 2 (v()) and vF () =
(1 (v()) 2 (v())), then vF satisfies clauses (5) and (6) of Definition 6.2.7.
Therefore, it is a valuation over EB such that v() D B iff vF () = 1, for every
formula . 

It is now straightforward to prove, as it was done in the previous section, the

following:

Theorem 6.5.10 (Equivalence between F-structures and swap structures semantics

for mbCci) Let  {} be a set of formulas in L . Then:

 mbCci
F if and only if  |=KmbCci .
266 6 Semantics of Non-deterministic Character for LFIs

Theorem 6.5.11 (Soundness and completeness of mbCci w.r.t. swap structures) Let
 {} be a set of formulas in L . Then:

 mbCci if and only if  |=KmbCci .

The proofs of Theorems 6.4.9 and 6.5.5 can be now adapted to mbCci, so obtaining
a characterization of mbCci in terms of a single finite Nmatrix (see Corollary 6.5.13
below). Recall that, by definition,
 
A2 = (1, 1), (1, 0), (0, 1) .
Bciw

Let MmbCci
A2 be the Nmatrix induced by BA mbCci
2
, the unique swap structure for mbCci
with domain BA2 . The set of designated values is DBciw
ciw
A2
= {t, T }, where t = (1, 1)
and T = (1, 0), while F = (0, 1) is the only non-designated truth-value. The multi-
operations of MmbCci
A2 are the same than those for MmbCciw
A2 , with just one restric-
tion: when x {T, F} then x = {T }, according to Definition 6.5.7. In other words,
the multioperations of MmbCci
A2 are the same than those for MmbCciw
A2 displayed in
Table 6.2, with exception of the multioperator , which is defined as follows:

That is, is a function instead of a multifunction. The complete definition of the

multioperations in the Nmatrix MmbCci
A2 are displayed below (Table 6.3).

Table 6.3
6.5 Swap Structures for Some Extensions of mbC 267

Theorem 6.5.12 Let v be a valuation for mbCci in the sense of Definition 3.1.7
of Chap. 3. Then, v induces a valuation vS over the Nmatrix MmbCci
A2 such that
vS () DBciw
A
iff v() = 1, for every formula .
2

Proof The proof is just an adaptation of the respective ones presented above for
Theorems 6.4.9 and 6.5.5. Let us begin by observing that, given a valuation v
for mbCci, the pair (v(), v()) is an snapshot in Bciw A2 , for every formula .
Consider the mapping vS such that vS () = (v(), v()), for every formula .
By the proof of Theorem 6.5.5, it follows that vS is a valuation over the Nmatrix
MmbCciw
A2 . Now, suppose that vS () = T . Then v() = 1 and v() = 0, whence
v() = (v() v()) = 1 and v() = v() v() = 0. Then vS () =
T whenever vS () = T . Analogously, it can be proven that vS () = T whenever
vS () = F. This means that vS is in fact a valuation over the Nmatrix MmbCci A2 .
Clearly, vS () DBciw
A
iff v() = 1, for every formula . 
2

From this, the following result follows easily:

Corollary 6.5.13 (Soundness and completeness of mbCci w.r.t. the Nmatrix MmbCci
A2
over A2 ) Let  {} be a set of formulas in L . Then:

 mbCci if and only if  |=MmbCci

A
.
2

Consider now the system bC, which is obtained from mbC by adding axiom (cf):

(cf)

Definition 6.5.14 Let A = A, , , , 0, 1 be a Boolean algebra. A swap struc-

ture for bC over A is a swap structure

B = B, , , , , 

for mbC over A (recall Definition 6.4.1) with the following restriction:

x = {z B : z 1 = x2 and z 2 x1 }
def

for every x B. 
Clearly, for every Boolean agebra A there exists a unique swap structure BA bC

for bC with domain BA . For any swap structure B for bC, the Nmatrix MbC (B) =
def

B, D B  is defined as in the previous cases. Let KbC be the class of all the Nmatrices
of the form MbC (B). The semantical consequence relation w.r.t. swap structures for
bC is the consequence relation |=KbC generated by the class KbC of Nmatrices. As in
the previous cases, it can be proven that the Nmatrices generated by swap structures
for bC over a non-trivial Boolean algebra A are non-trivial, since the set of designated
truth-values is a proper and nonempty subset of the domain. The Nmatrix associated
 bC 
to BAbC
will be denoted by MbC A . In particular MA2 = BA2 , DBA2 .
bC def
268 6 Semantics of Non-deterministic Character for LFIs

Proposition 6.5.15 Let B be a swap structure for bC. Then, for every x B,

1 [x x] = {1} .

Proof Let z x. Then, there exists y x such that z y and so z 1 = y2 .

Since y2 x1 it follows that z 1 x1 . Hence, z 1 x1 = 1. The proof concludes by
observing that 1 [x x] = {z 1 x1 : z x}. 

As in the previous cases, we obtain the following:

Proposition 6.5.16 Let A be a Boolean algebra with domain A where 1

= 0.
(1) For every F-structure E for bC based on A, and for every valuation v over E,
there exists a swap structure BE for bC with domain BA and a valuation vS over BE
such that vS () DBA iff v() = 1, for every formula .
(2) For every swap structure B for bC based on A, and for every valuation v over B,
there exist an F-structure EB for bC and a valuation vF over EB such that v() D B
iff vF () = 1, for every formula .
 
Proof (1) Let E = A, {Na }aA , {Oa }aA be an F-structure for mbC together with
a valuation v on it, and let BE be the unique swap structure BA bC
for bC with
domain BA . By Definitions 6.1.3 and 6.2.18, (v(), v(), v()) BA for every
formula . This allows to define the mapping vS : L BA such that vS () =
(v(), v(), v()), for every formula . Clearly,
(i) vS (#) = (v(#), v((#)), v((#))) vS ()#vS (),
for # {, , };
(ii) vS () = (v(), v(), v()) vS ().
Additionally,

vS () = (v(), v(), v()) {z BA : 1 (z) = 2 (vS ())} .

Thus, in order to prove that vS () vS () it is enough to prove that 2 (vS ())

1 (vS ()), that is, v() v(). But this is a consequence of item (9) of Def-
inition 6.2.18. That being so, it follows that vS is a valuation over BE such that
vS () DBA iff v() = 1, for every formula .
(2) Let B = B, , , , ,  be a swap structure for bC based on the Boolean
algebra A. By Proposition 6.4.5 (i), the set A = 1 [B] is the domain of a sub-Boolean
algebra A of A. Consider a structure
 
EB = A , {NaB }aA , {OaB }aA

where, for every a A ,

NaB = {b A : (a, b, c) B for some c A },

OaB = {c A : (a, b, c) B for some b A }.

6.5 Swap Structures for Some Extensions of mbC 269

Because of the definition of and , the sets above are always nonempty. Since
B BA , it follows that EB is an F-structure for mbC. If b NaB then (a, b, c) B
for some c. Let x = (a, b, c) and y x. Then y = (b, d, c ) for some d and c
such that d a. This means that d NbB and d a = 1, and so C Fab
= . Hence,
condition (F6) of Definition 6.2.17 is satisfied, showing that EB is an F-structure
for bC.
Consider now a valuation v over B, and define the mapping vF : L A
such that vF () = 1 (v()), for every formula . Analogous to the case of mbC,
it can be proven that vF satisfies clauses
 (1)(3) of Definition 6.1.3.
 By Propo-
sition 6.5.15, 1 [v() v()] = z 1 1 (v()) : z v() = {1}. Since
v() v(), it follows that 1 (v()) 1 (v()) = 1. That is,
vF () vF () = 1. Thus, vF is a valuation over EB such that, by the very defi-
nitions, v() D B iff vF () = 1, for every formula . 
From this, the following results can be proven easily:
Theorem 6.5.17 (Equivalence between F-structures and swap structures semantics
for bC) Let  {} be a set of formulas in L . Then:

F if and only if  |=KbC .

 bC

Theorem 6.5.18 (Soundness and completeness of bC w.r.t. swap structures) Let

 {} be a set of formulas in L . Then:

 bC if and only if  |=KbC .

As in the previous cases, the adequacy of bC w.r.t. the 5-valued Nmatrix MbCA2 =
BA
bC
2
,DBA2  associated to the unique swap structure BA bC
2
defined over the Boolean
algebra A2 with domain A2 = {0, 1}, can be stated. This can be proven easily by
adapting the proof of Proposition 6.5.16(1) presented above.
Theorem 6.5.19 Let v be a bC-valuation, that is, an mbC-valuation in the sense of
Definition 2.2.1 satisfying, additionally, clause (vCf): v() = 1 implies v() = 1,
for every formula . Then, v induces a valuation vS over the Nmatrix MbC A2 such that
vS () DBA2 iff v() = 1, for every formula .
Proof Given a bC-valuation v, it is clear that (v(), v(), v()) BA2 for every
formula , by the properties of v (recall Definition 2.2.1, and clause (vCf )). Consider
the mapping vS such that vS () = (v(), v(), v()), for every formula . It is
immediate to see that
(i) vS (#) = (v(#), v((#)), v((#))) vS ()#vS (),
for # {, , };
(ii) vS () = (v(), v(), v()) vS ().
Now, vS () = (v(), v(), v()) is such that v() v() and then
vS () vS (), for every . Therefore, vS is a valuation over the Nmatrix MbC
A2
such that vS () DBA2 iff v() = 1, for every formula . 
270 6 Semantics of Non-deterministic Character for LFIs

Corollary 6.5.20 (Soundness and completeness of bC w.r.t. the Nmatrix MbC

A2 over
A2 ) Let  {} be a set of formulas in L . Then:

 bC if and only if  |=MbC

A
.
2

Proof The only if part follows directly from Theorem 6.5.18. For the if part,
assume that  |=MbC
A
, and let v be a bC-valuation such that v[] {1}. By Theo-
2
rem 6.5.19, there exist a valuation vS over the Nmatrix MbC
A2 such that vS () DBA2
iff v() = 1, for every formula . Thus, vS ( ) DBA2 for every  whence
vS () DBA2 , by hypothesis. This means that v() = 1, showing that  bC . By
completeness of bC w.r.t. bC-valuations (see [11]),  bC . 

The Nmatrix MbC A2 is similar to MA2 , that is, the Nmatrix M5 presented in
mbC B

Table 6.1 of Example 6.3.3, but with the following restriction: the sets F and f 0
must coincide with the set {T, t0 } (we invite the reader to discover why these are the
unique
 restrictions to be imposed). Thus, recalling that D = {T, t, t0 } and N D =
F, f 0 , the multioperations for the Nmatrix MbC A2 are described in Table 6.1 of
Example 6.3.3, but where the negation is defined as follows:

T ND
t D
t0 ND
F {T, t0 }
f0 {T, t0 }

Finally, the system Ci, which is obtained from mbC by adding axioms (ci) and (cf),
will be also examined.

Definition 6.5.21 Let A = A, , , , 0, 1 be a Boolean algebra. A swap struc-

ture for Ci over A is any multialgebra over  (see Sect. 6.3)

B = B, , , , , 

such that B Bciw

A and where the multioperations are defined as follows, for every
x and y in B:
(i) x#y = {z B : z 1 = x1 #y1 }, for each # {, , };
def

(ii) x = {z B : z 1 = x2 and z 2 x1 };
def

(iii) x = {((x1 x2 ), x1 x2 )}.

def


6.5 Swap Structures for Some Extensions of mbC 271

 structure B forCi, the Nmatrix MCi (B) = B, D B  is defined as

def
For any swap
usual: D B = x B : x1 = 1 . Let K be the class of all the Nmatrices of the
Ci

form MCi (B). The semantical consequence relation w.r.t. swap structures for Ci is
the consequence relation |=KCi generated by the class KCi of Nmatrices. The proof
of the following results is routine, and left to the reader:

Theorem 6.5.22 (Equivalence between F-structures and swap structures semantics

for Ci) Let  {} be a set of formulas in L . Then:

F if and only if  |=KCi .

 Ci

Theorem 6.5.23 (Soundness and completeness of Ci w.r.t. swap structures) Let 

{} be a set of formulas in L . Then:

 Ci if and only if  |=KCi .

Let A2 be the two-elements Boolean algebra. Clearly, there is just one swap struc-
ture BA
Ci
2
for Ci with domain Bciw
A2 , namely the multialgebra obtained from BA2
mbCciw

(whose multioperations are displayed in Table 6.2), with exception of the multioper-
ators and , which are defined as follows:

Observe that is a function instead of a multifunction. The complete definition

of the multioperations in the swap structure BA
Ci
2
are displayed below (Table 6.4).

Table 6.4
272 6 Semantics of Non-deterministic Character for LFIs

Let MCiA2 be the Nmatrix induced by BA2 . The set of designated values is DBciw
Ci
A2
=
{t, T }. Then, the following result is easily obtained form the previous results:
Theorem 6.5.24 (Soundness and completeness of Ci w.r.t. the Nmatrix MCi
A2 ) Let
 {} be a set of formulas in L . Then:

 Ci if and only if  |=MCi

A
.
2

It is worth observing that the 3-valued Nmatrix MCi A2 obtained for Ci coincides
with the 3-valued characteristic matrix MCi for Ci proposed in [21] and reproduced
above in Example 6.3.4. Thus, Theorem 6.5.24 coincides with Avrons result pre-
sented in [21], but presented in the context of swap structures.
In the next section the axiom cl will be analyzed, showing that it is much more
complicated than the others investigated up to now. Moreover, apparently mbCcl is
not able to be treated under the perspective of the swap structures presented here.

6.6 Axiom (cl) and Uncharacterizability by Finite

Nmatrices

Recall from Definition 3.3.26 the logic mbCcl obtained from mbC by adding the
axiom schema
( ) (cl)

This logic, which extends mbCciw, is the least extension of mbC such that the
truth-value of can be expressed in terms of the truth-value of a formula con-
structed by the other connectives, by considering the valuation semantics introduced
in Definition 3.3.26 of Chap. 3. Specifically, the truth-value of is given by the
truth-value of the formula ( ). Thus, by the properties of such valuations,
v(( )) = v() = (v() v()).
As it will be argued in this section, axiom (cl) is, by nature, different from any other
investigated up to now: the extensions of mbC by adding axiom (cl) (and, of course,
if no other axioms powerful enough to get truth-functionality are present) produce
systems which cannot be characterized by a finite single Nmatrix, as discovered by
Avron in [22] (recall Theorem 6.3.5 above). This constitutes an interesting Dugundji-
like result for LFIs with respect to Nmatrices. Probably related to this fact, the
techniques developed in the previous sections for swap structures do not work for
mbCcl: the natural definition of swap structures formed by snapshots (a, b) such
that a represents the truth-value of and b represents a possible truth-value for
(assuming that (a b) represents ) is too tight, and the wide non-determinism
inherent to any semantics for mbCcl is lost in some way. This claim will be clear
from the analysis below.
Let A = A, , , , 0, 1 be a Boolean algebra. A swap structure for mbCcl
over A should be defined in two steps: firstly, it should be considered a multialgebra
6.6 Axiom (cl) and Uncharacterizability by Finite Nmatrices 273

B0 = B, , , , 

such that B Bciw

A and, for every x and y in B:

(i) x#y = {z B : z 1 = x1 #y1 }, for each # {, };

def

(ii) x {z B : z 1 = x2 };
(iii) x = {z B : z 1 = (x1 x2 )}.
def

After defining the negation (and observing that the consistency operator should
be defined as in mbCciw), in a second step the conjunction multioperator should
be defined, in such a way that every instance of axiom (cl) be valid. Obviously,
x y {z B : z 1 = x1 y1 }, and the question is how to restrict the conjunction
in order to validate (cl).
Thus, given B0 as above, defining a conjunction over B0 in order to validate
axiom (cl) is equivalent to having a conjunction as above such that, for every
x B,

1 [(x x) x] = {z 1 w1 : z (x x) and w x} = {1} .

Suppose that is defined over B0 such that (cl) is valid, and let x B and
z (x x). Then, there exists y (x x) such that z 1 = y2 . On the other
hand, if w x then w1 = (x1 x2 ), by definition of . Since axiom (cl) is
valid, z 1 w1 = 1. That is, z 1 w1 or, equivalently, y2 (x1 x2 ). There-
fore, y2 (x1 x2 ) = 0. Since y (x x) then y1 = x1 u 1 for some u x.
Then, u 1 = x2 whence y1 = x1 x2 . Given that y B then y2 y1 = 1, that is,
y2 (x1 x2 ) = 1. Using the definition of Boolean complement it follows that
y2 = (x1 x2 ). Therefore, any multioperation over B0 satisfying (cl) should be
defined as follows:

{(x1 x2 , (x1 x2 ))} if y x

xy=
{z B : z 1 = x1 y1 } otherwise.

This restriction is incompatible with the amount of freedom required by valua-

tions over F-structures for mbCcl (recall Definition 6.2.12) and, in particular, by
mbCcl-valuations (recall Definition 3.3.26). Indeed, in order to prove the equiv-
alence between F-structures and swap structures, we proceed as follows: given
an F-structure E for mbCcl and a valuation v over E, a mapping vS : L Bciw A is
defined such that vS () = (v(), v()), for every . Now, suppose that and are
formulas such that v() = v() and v() = v(). Then, vS () = vS () and
so vS () vS () = vS () vS () {(v() v(), (v() v()))}, since
vS () vS (). Given that vS ( ) = (v( ), v(( ))) vS () vS (), it
follows that v(( )) = (v() v()) = v( ). Thus, v((
)) = 1. In particular, consider the 2-valued Boolean algebra A2 and let v be an
mbCcl-valuation in the sense of Definition 3.3.26 such that v() = v() = 1 and
v() = v() = 1 for some formulas and (by simplicity, we can take
and as being two different propositional variables). Suppose, additionally, that
274 6 Semantics of Non-deterministic Character for LFIs

v() = 1. This means that t t, recalling that t = (1, 1) and the definition of vS .
By the requirements over , it follows that t t = {T }, where T = (1, 0). From this,
v(( )) = 0, and v(( )) = 1 for every and as above. It is clear that
this requirement is excessive for mbCcl-valuations: it is always possible to define,
for two different propositional variables and , an mbCcl-valuation v such that
v() = v() = v() = 1 and v() = v() = 1 but v(( )) = 1. This
shows that the frameworks of swap structures and finite Nmatrices, as studied above,
are not enough to characterize mbCcl.

Remark 6.6.1 From the analysis above, some considerations can be made about the
status of axiom (cl) with respect to Nmatrizes.
In a non-deterministic matrix semantics, a multioperation is associated to each
connective of the language. Given a valuation v over the Nmatrix, in order to evaluate
a formula , the formula is decomposed into its subformulas, and the value v() is
obtained by using the multioperations previously assigned to the connectives. Starting
from the atomic formulas (which get a single truth-value by means of v), the value
of the complex formulas must be chosen from the possible values returned by the
multioperators of the Nmatrix. Thus, each connective of the formula being evaluated
must be taken into account.
By its turn, Fidel structuresas well as the valuation semantics for LFIs studied in
Chaps. 2 and 3can evaluate certain complex formulas as a whole, and the resulting
value does not necessarily depend on the value of its immediate subformulas, but
on the value of other related formulas. In the specific case of axiom (cl), some
restrictions are required in such semantics in order to get the truth-value, say a, of
( ): namely, that a be less or equal than the truth-value of the formula . In
contrast, when evaluating the formula ( ) within an Nmatrix, from a given
value for (among a nonempty set of possible values), a possible value for is
chosen; after this, a possible value for is obtained, and finally a possible value
for ( ) is chosen from among a set of possibilities. This dependence on the
operators associated to the syntax of the formula limits the degree of non-determinism
required by the evaluation process. In fact, as shown by the example above, given
a valuation vS over an hypothetical Nmatrix for mbCcl, if is a formula different
from such that vS assigns the same values to and , then the value of
should be chosen from the set of possible values of , being so affected by the
restrictions imposed by axiom (cl). This would limit the possibilities of choice for
the value of the formula ( ), as it was shown above: clearly, mbCcl-valuations
have a greater degree of non-determinism. Behind this limitations lies, probably, the
necessity of considering infinite Nmatrices for characterizing mbCcl or some of its
extensions, as shown above in Theorem 6.3.5.
It should be observed, however, that this condition (the nesting of non-deterministic
operators) can, probably, be necessary, but never sufficient to prevent the character-
ization of a logic by means of a finite Nmatrix: as shown in Corollary 6.5.13, the
6.6 Axiom (cl) and Uncharacterizability by Finite Nmatrices 275

logic mbCci, which includes axiom (ci) (in which the non-deterministic operators
and are nested), can be characterized by a single 3-valued Nmatrix.

This section concludes by analyzing a possible axiomatic extension of mbCcl

which can be characterized by a pair of 3-valued swap structures over A2 , and so by
a pair of 3-valued Nmatrices.
In order to obtain such characterization, it is clear that some additional require-
ments must be added to mbCcl-valuations (recall Definition 3.3.26) in order to limit
its freedom of choice. Of course that requirements must be able to be expressed as
propositional axioms, which leads to the following system (recalling that is
an abbreviation for ( ) ( )):

Definition 6.6.2 Let mbCclND be the logic obtained from mbCcl by adding the
following axiom schemas:
 
( ) ( ) ( ) (AxND1)
 
( ) ( ) ( ) (AxND2)

An mbCclND-valuation is an mbCcl-valuation v (recall Definition 3.3.26) which

also satisfies the following clauses:
(vND1) v( ) = v( ) = 1 = v() = v()
(vND2) v( ) = v() = 1 and v() = v() = v(( )) = 0.
Let V mbCclND be the set of mbCcl-valuations. 

As usual, the consequence relation in the calculus mbCclND will be denoted by

mbCclND , while the semantical consequence relation with respect to valuations in
V mbCclND will be denoted by mbCclND .

Theorem 6.6.3 (Soundness of mbCclND w.r.t. valuations) For every  {}

L :
 mbCclND =  mbCclND .

Proof Clearly, it is enough to prove that any instances of axioms (AxND1) and
(AxND2) are valid with respect to mbCclND-valuations. The verification of the
details of the proof is left to the reader. 

Theorem 6.6.4 (Completeness of mbCclND w.r.t. valuations) For every  {}

L :
 mbCclND =  mbCclND .
276 6 Semantics of Non-deterministic Character for LFIs

Proof Suppose that  mbCclND , and consider a set  which is maximal non-
trivial with respect to in mbCclND such that  . By adapting the proof of
Theorem 3.3.27, the mapping v : L {0, 1} such that v() = 1 iff , for all
L , is a valuation for mbCcl. Now, suppose that v( ) = v( ) = 1.
Then, ( )  and (  ) , whence ( ) ( ) . Given
that ( ) ( ) ( ) is in  (since this formula is an
instance of an axiom of mbCclND and  is a closed theory of mbCclND), the
formula ( ) belongs to , using again that  is a closed theory. Then,
clearly v() = v() and so v satisfies clause (vND1). Analogously, it can be
proven that v satisfies clause (vND2), and so v is an mbCclND-valuation such that
v[] {1} but v() = 0. Therefore,  mbCclND . 

Two swap structures for mbCclND over A2 can be now defined. They are derived
from the one for mbCciw displayed in Table 6.2, in which and are respectively
defined as follows:
(Ra ) t = {t}, t t = {T } and t T = {t, T }; and
def def def

(Rb ) t = {T }, t t = {t, T } and t T = {T }.

def def def

These multialgebras induce respectively the 3-valued Nmatrices Ba and Bb having

the set DBciw
A2
= {T, t} of designated values. Thus, Ba and Bb are obtained from the
multialgebra for mbCciw displayed in Table 6.2 by changing the tables for and
as follows, respectively (Table 6.5):

Table 6.5
6.6 Axiom (cl) and Uncharacterizability by Finite Nmatrices 277

Let KmbCclND = {Ma , Mb }. The next results will show that these two Nmatrices
characterize mbCclND (see Corollary 6.6.7 below).
Proposition 6.6.5
(1) Let v be an mbCclND-valuation in the sense of Definition 6.6.2. Then, v induces
a mapping vS : L Bciw A2 given by vS () = (v(), v()) such that vS is either a
valuation over the Nmatrix Ma or a valuation over the Nmatrix Mb , and vS ()
DBciw
A
iff v() = 1, for every formula .
2
(2) Let vS : L Bciw
A2 be a valuation over the Nmatrix Ma . Then, it induces a
mapping v : L {0, 1} given by v() = 1 (vS ()), such that vS () DBciw
A2
iff
v() = 1, for every formula .
(3) Let vS : L Bciw
A2 be a valuation over the Nmatrix Mb . Then, it induces a
mapping v : L {0, 1} given by v() = 1 (vS ()), such that vS () DBciw
A2
iff
v() = 1, for every formula .
Proof (1) Let v : L {0, 1} be an mbCclND-valuation, and consider the mapping
vS : L Bciw A2 such that vS () = (v(), v()), for every formula . Observe that,
given that v satisfies clause (vNeg), any pair (v(), v()) is indeed a snapshot in
A2 and then the mapping vS is well defined. Since v satisfies clause (vND1), there
Bciw
are three cases to analyze:
Case 1: There is such that v() = v() = 1 and, for every , v() = v() = 1
implies v() = 1.
It will be proven that vS is a valuation over the Nmatrix Ma . By the hypothesis on v, if
vS () = t then vS () = t. Since t = {t} in Ma then vS () vS () whenever
vS () = t. If vS () {T, F} then clearly also vS () vS (), by the properties
of v. Suppose now that vS () = vS () = t. Then v() = v() = v() = v() =
v() = 1. Since v satisfies clause (vND2), it follows that v(( )) = 0. Then,
vS ( ) = T vS () vS () whenever vS () = vS () = t, by definition of Ma .
In any other case, it is clear that vS ( ) vS () vS (). It is also clear that
vS () vS () for every , since v() = (v() v()). On the other hand, by
the properties of v, it follows that vS (#) vS ()#vS (), for each # {, }. This
shows that vS is a valuation over the Nmatrix Ma such that, by the very definition,
vS () DBciwA2
iff v() = 1, for every formula .
Case 2: There is such that v() = v() = 1 and, for every , v() = v() = 1
implies v() = 0.
It will be proven that vS is a valuation over the Nmatrix Mb . Because of the
hypothesis on v, if vS () = t then vS () = T . From t = {T } in Mb it fol-
lows that vS () vS (), if vS () = t. As observed above, if vS () {T, F} then
also vS () vS (). Now, assume that vS () = t and vS () = T . Then v() =
v() = v() = 1 and v() = v() = 0. Using clause (vND2), which is sat-
isfied by v, it follows that v(( )) = 0. Thus, vS ( ) = T vS () vS ()
whenever vS () = t and vS () = T , by definition of Mb . In any other case, clearly
vS ( ) vS () vS (). As proved above, vS () vS () for every . Finally,
vS (#) vS ()#vS (), for each # {, }, by the properties of v. From this, vS
is a valuation over the Nmatrix Mb such that vS () DBciw A2
iff v() = 1, for every
formula .
278 6 Semantics of Non-deterministic Character for LFIs

Case 3: For every , v() = 1 implies v() = 0.

In this case v is a classical valuation with values in {T, F} and so it is both a valuation
over Ma and a valuation over Mb .
(2) Let vS : L Bciw A2 be a valuation over the Nmatrix Ma , and consider the map-
ping v : L {0, 1} such that v() = 1 (vS ()). From vS () vS () it follows
that v() 1 [vS ()] = {2 (vS ())}, hence v() = 2 (vS ()). From this,
v() 1 [vS ()] = {(v() v())}, whence v() = (v() v()).
Clearly v(#) v()#v() for every # {, , }, because of the definition
of the binary multioperations in the swap structure. Since vS () Bciw A2 it fol-
lows that v() v() = 1. It is immediate to see from the facts above that
v() v() v() = 0, thus v is an mbC-valuation (recall Definition 2.2.1).
Now, suppose that v(( )) = 1. If v( ) = 1 then vS () = t and
vS ( ) = t vS () vS (). Given that vS () vS () = t = {t}, it fol-
lows that vS () = t. This means that t = vS ( ) t t = {T }, which is
impossible. Then v( ) = 0, whence v() = v( ) = 1. Thus, v is an
mbCcl-valuation. Observe that, if v( ) = 1 then vS () = t. From vS ()
vS () it follows that vS () = t, thus v() = 1. Suppose now that
v( ) = v( ) = 1. By the last observation, v() = v() = 1 and
then v( ) = 1. This shows that v satisfies clause (vND1). Finally, if
v( ) = v() = 1 and v() = v() then v() = 1, by the observa-
tion above. This means that vS () = vS () = t. Since vS ( ) vS () vS () =
t t = {T } it is inferred that vS ( ) = T . From this, v(( )) = 0. Then, v
satisfies clause (vND2) and so it is an mbCclND-valuation. By the very definition,
vS () DBciw
A2
iff v() = 1, for every formula .
(3) Analogous to the proof of item 2, but now observing that, if v( ) = 1 then
v() = 0. The details of the proof are left as an exercise to the reader. 
Theorem 6.6.6 Let  {} be a set of formulas in L . Then:

 mbCclND if and only if  |=KmbCclND .

Proof Suppose that  mbCclND . Let vS be a valuation over the Nmatrix Ma such
that vS [] DBciwA2
. Then, the mbCclND-valuation v defined as in item 2 of Propo-
sition 6.6.5 is such that v[] {1}. By hypothesis, v() = 1 whence vS () DBciw
A2
.
Then,  |=Ma . Analogously, it is proved that  |=Mb and so  |=KmbCclND .
Conversely, suppose that  |=KmbCclND and let v be an mbCclND-valuation such
that v[] {1}. Using item 1 of Proposition 6.6.5, there is mapping vS which is
either a valuation over the Nmatrix Ma or a valuation over the Nmatrix Mb , such
that vS () DBciw
A2
iff v() = 1, for every formula . Then vS [] DBciw
A2
and so,
by hypothesis, vS () DBciw A
. Therefore v() = 1, showing that  mbCclND . 
2

Corollary 6.6.7 (Soundness and completeness of mbCclND w.r.t. 3-valued Nma-

trices) Let  {} be a set of formulas in L . Then:

 mbCclND if and only if  |=KmbCclND .

6.7 Some Remarks on Fidel Structures and Swap Structures 279

6.7 Some Remarks on Fidel Structures and Swap

Structures

In the previous sections, several LFIs were characterized by a semantics of Fidel

structures, and also by a semantics of swap structures. From an algebraic point of
view, a swap structure is a multialgebra which can be naturally expanded to a non-
deterministic matrix. An interesting point is that swap structures over a given logic
are defined in domains formed by triples (or pairs, depending on the characteristics
of the logic) of elements of an algebra of a given kind (in the case of LFIs, Boolean
algebras). As observed at the end of Remark 6.4.4, the Nmatrices introduced by
Avron were defined by means of snapshots, following the same intuitions as ours: in
order to analyze mbC (and its extensions), the truth-values (true or false) of ,
and are relevant, for any formula .
A closer antecedent to swap structures and snapshots, which in fact inspired its
definition, is Fidel and Vakarelovs notion of twist structures (recall Sect. 5.1.3 of
Chap. 5). Moreover, in his Ph.D. thesis ([23]), Fidel supports the idea that, by using
tuples of a given class of algebras, it is possible to study the structure and meaning of
the algebraic models of certain non-classical logics. His twist structures for Nelsons
logic are just an example of this general idea: in Chapter IV of [23], he introduces
an original semantics for the logic of Ockham algebras P3,1 (see [24]) in terms
of triples of elements of distributive lattices. Given a triple (a, b, c), a represents
the truth-value of a given formula , while b represents the truth-value of , and
c represents the truth-value of . On the other hand, in [25] Fidel proposes a
semantical characterization of da Costas logics C1 in terms of triples over Boolean
algebras such that any triple (a, b, c) represents the truth-value a of a given formula
, the truth-value b of , and the truth-value c of ( ). It seems clear that
these triples play the role of snapshots, in the sense of the swap structures we propose
here. These examples suggest the possibility of defining a general notion of n-swap
structures, for n 2, but this task will not be developed here.
From the results obtained in the previous sections of this chapter, it can be seen that
F-structures and swap structures constitute an appealing algebraic approach to (non-
truth-functional) logics like mbC and several useful extensions of it. In most cases,
it can be shown that both paradigms are equivalent. However, as it was proved in this
section, some axioms (specifically, axiom (cl)) cannot be treated by swap structures
(at least, in the way proposed here). This suggest that (finite) F-structures have greater
expressive power than (finite) swap structures (and so than finite Nmatrices).
An algebraic study of F-structures (seen as special classes of algebraic-relational
structures) as well as of swap structures (seen as multialgebras) is still to be developed.
Such study will open new horizons in the area of algebraization of non-classical
logics.
280 6 Semantics of Non-deterministic Character for LFIs

6.8 The Possible-Translations Semantics

The possible-translations semantics (PTSs) were introduced by Carnielli in 1990

(see [26]) in an attempt to offer a more generally acceptable philosophical interpre-
tation for some non-classical logics, and specially for paraconsistent logics. More
recent expositions on PTSs are given by Carnielli in [27], by Carnielli, Coniglio and
Marcos in [11] and by Marcos in [28, 29].
The basic intuition behind the PTSs is the notion of translations as morphisms
between logics (that is, maps preserving consequence relations). In an analogy to
what happens with natural languages, translations are thought as different world
views, and the concept of possible-translations semantics offers a way to interpret
a given logic L (with an unknown semantic characterization) as the combination of
all possible world views, formalized by an appropriate set of translations of the
formulas of L into a class of simpler logics with known semantic characterization.
By selecting a suitable collection of translations, and a set of constraints applied to
such translations, the original logic L acquires a semantic meaning supported by the
simpler logics (typically, but not necessarily, a collection of three-valued logics).

6.8.1 Possible-Translations Semantics for Some LFIs

As an illustrative example, a semantical characterization by PTSs (taken from [29])

will be given in this subsection for some of the LFIs studied in the previous chapters,
namely mbC, mCi, bC and Ci. It intends to show how the PTSs works well for several
logics which cannot be characterized by finite matrices, obtaining from there (in some
cases) a decision procedure for such logics. By its turn, an original characterization of
Cila by means of a PTS structure over a 3-valued logic will be obtained in Sect. 6.8.2.
Let PTS = {, , , 1 , 2 , 3 , 1 , 2 , 3 } be a signature, and let MPTS =
A, D be the matrix over PTS such that A = {T, t, F}, D = {T, t}, and the opera-
tions are defined by the truth-tables below.

Now, consider the following clauses governing any mapping f : L L P T S :

6.8 The Possible-Translations Semantics 281

(tr 0) f ( p) = p for p V ar ;
(tr 1) f (#) = ( f ()# f ()), for # {, , };
(tr 2) f () {1 f (), 2 f ()};
(tr 3) f () {1 f (), 3 f ()};
(tr 4) f (n+1 ) = 1 f (n ), for n N;
(tr 5) f () {2 f (), 3 f (), 2 f (), 3 f ()};
(tr 6) f () {1 f (), 1 f ()};
(tr 7) if f () = 1 f () then f () = 1 f ().
As usual, n (used in (tr 4)) denotes n applications of over the formula ; in
particular, 0 = . The following results were obtained in [29]:

Theorem 6.8.1 Let { f i }iI1 be the family of mappings f i : L L P T S satisfying

clauses (tr 0), (tr 1), (tr 2) and (tr 5) above. Let Li = M P T S for every i I1 . Then,
def

PTS1 = {Li }iI1 , { f i }iI1  is a possible-translations semantics for the logic mbC.
That is, for every set of formulas  {} of mbC, it holds:

 mbC iff f i [] |=M f i (), for every i I1 .

Theorem 6.8.2 Let { f i }iI2 be the family of mappings f i : L L P T S satisfying

clauses (tr 0), (tr 1), (tr 2), (tr 4) and (tr 6) above. Let Li = M P T S for every i I2 .
def

Then, PTS2 = {Li }iI2 , { f i }iI2  is a possible-translations semantics for the logic
mCi. That is, for every set of formulas  {} of mCi, it holds:

 mCi iff f i [] |=M f i (), for every i I2 .

Theorem 6.8.3 Let { f i }iI3 be the family of mappings f i : L L P T S satisfying

clauses (tr 0), (tr 1), (tr 3) and (tr 5) above. Let Li = M P T S for every i I3 . Then,
def

PTS3 = {Li }iI3 , { f i }iI3  is a possible-translations semantics for the logic bC.
That is, for every set of formulas  {} of bC, it holds:

 bC iff f i [] |=M f i (), for every i I3 .

Theorem 6.8.4 Let { f i }iI4 be the family of mappings f i : L L P T S satisfying

clauses (tr 0), (tr 1), (tr 3), (tr 6) and (tr 7) above. Let Li = M P T S for every i I4 .
def

Then, PTS4 = {Li }iI4 , { f i }iI4  is a possible-translations semantics for the logic
Ci. That is, for every set of formulas  {} of Ci, it holds:

 Ci iff f i [] |=M f i (), for every i I4 .

Let 1 j 4. It is worth noting that the set { f i () : i I j } is finite, since

the length of is finite. In view of these circumstances, if  {} is finite, it is
possible to determine in a finite number of steps whether f i [] |=M f i () for every
i I j . This means that the PTS defined above constitutes a decision procedure for
the corresponding logic.
282 6 Semantics of Non-deterministic Character for LFIs

6.8.2 A 3-Valued Possible-Translations Semantics for Cila

In this subsection, a semantical characterization of the logic Cilaover a 3-valued

PTS will be obtained (see Theorem 6.8.8), by adapting a similar result relative to the
logic C1 given by Marcos in [28]. This result shows that finite-valued PTSs are more
powerful that finite-valued Nmatrices, taking into account Theorem 6.3.5. From this
PTS characterization, a decision procedure for Cila is easily obtained, as it was done
in the previous subsection for some LFIs.
Consider the signature

P T S = {{i }i=1 , {i }i=1 , {i }i=1 , {i }i=1 , {i }i=1 }

 Cila 3 3 3 2 2

and let W3+ = A, D be the matrix over  Cila

P T S such that A = {T, t, F}, D = {T, t},
and the operations are defined by the truth-tables below.

As observed in [28], the reduct W3 of W3+ which omit the consistency operators
1 and 2 coincides with da Costa and DOttavianos logic J3 (recall Sect. 4.4.3 of
Chap. 4). From this, W3+ also coincides with J3 (and so with LFI1 and MPT, recall
Remark 4.4.12), since both consistency operators are definable in J3.
6.8 The Possible-Translations Semantics 283

Consider now a mapping f : L L Cila

PT S
defined recursively by following the
clauses below.
(T r 0) For p V ar :
(T r 0.1) f ( p) = p;
(T r 0.2) f ( p) = 2 p.
(T r 1) For ( # ), where # {, , }:
(T r 1.1) f ( # ) = ( f () 3 f ()) if ( # ) = ( ); otherwise:
(T r 1.2) f ( # ) = ( f () #1 f ()) if f () = 2 f () and f () = 1 f ();
(T r 1.3) f ( # ) = ( f () #2 f ()) if f () = 1 f () and f () = 2 f ();
(T r 1.4) f ( # ) = ( f () #3 f ()) otherwise.
(T r 2) For ( # ), where # {, , }:
(T r 2.1) f (( # )) = 1 f ( ) if ( # ) = ( ); otherwise:
(T r 2.2) f (( # )) = 1 f ( # ) if f () = 1 f () and f () = 1 f ();
(T r 2.3) f (( # )) 1 f ( # ), 2 f ( # ) otherwise.
(T r 3) For :
(T r 3.1) f () = 1 f () if f () = 1 f ();
(T r 3.2) f () 1 f (), 2 f () otherwise.
(T r 4) For :
(T r 4.1) f () = 1 f ().
(T r 5) For :
(T r 5.1) f () = 1 f () if f () = 1 f ();
(T r 5.2) f () = 2 f () if f () = 2 f ().
Observe that, by Definition 2.1.11 of complexity of a formula in L , l( # ) >
l() and l( # ) > l() for every and . This guarantees that the value f ()
is well-defined by recursion on the complexity l() of .
Definition 6.8.5 Let { f i }iI be the family of mappings f i : L L Cila
PT S
satisfying
clauses (T r 0)-(T r 5) above. Let Li = W3+ for every i I . Define the possible-
def

translations semantics structure PTSCila as follows:

PTSCila = {Li }iI , { f i }iI  .

def


Proposition 6.8.6 For every mapping f : L L Cila PT S
satisfying clauses (T r 0)-
(T r 5) above, and for every valuation h over the matrix W3+ , there exists a Cila-
valuation v (that is, a Cil-valuation satisfying additionally clauses (ca ), (ca ) and
(ca ) of Definition 3.6.1 of Chap. 3) such that, for every formula of Cila: v() = 1
iff h( f ()) D.
Proof Let f : L L Cila
PT S
be a mapping satisfying clauses (T r 0)-(T r 5) above,
and let h be valuation over the matrix W3+ . Consider the mapping v : L {0, 1}
defined as follows:
v() = 1 iff h( f ()) D .

In order to complete the proof, it suffices to prove that v is a Cila-valuation.

284 6 Semantics of Non-deterministic Character for LFIs

(vAnd): By (Tr1), f ( ) = f () i f () for some 1 i 3 and so h( f ( ))

= h( f () i f ()) = h( f ()) i h( f ()), since h is a  Cila
P T S -homomorphism.
Then, v( ) = 1 iff h( f ( )) D iff h( f ()) i h( f ()) D iff
h( f ()) D and h( f ()) D (by definition of i , for 1 i 3), iff v() = 1
and v() = 1.
(vOr): The proof is analogous, by observing that a i b D iff either a D or
b D, for a, b A and 1 i 3.
(vImp): The proof is analogous, by observing that a i b D iff either a = F or
b D, for a, b A and 1 i 3.
(vNeg): By (Tr2), (Tr3) and (Tr4), f () = i f () for some i = 1, 2. Hence,
h( f ()) = i h( f ()). Now, suppose that v() = 0. Then h( f ()) = F and so
i h( f ()) D. Thus, v() = 1.
(vCon): Suppose that v() = v() = 1. Then, h( f (()), h( f (()) D. By
definition of f , there exists i {1, 2} such that f () = i f () and f () =
i f (). From this, h( f (()) = i h( f ()) D and h( f (()) = i h( f ())
D for some i {1, 2}. If i = 1 then h( f ()) = F, by the truth-table of 1 . If i = 2
then also h( f ()) = F, by the truth-tables of 2 and 2 . In both cases, v() = 0
and so v satisfies clause (vCon).
(vCi): By (Tr4), f () = 1 f (), whence h( f ()) = 1 h( f ()). Suppose
that v() = 1. Then, 1 h( f ()) = T and h( f ()) = F, by the truth-table of
1 . By definition of f and h, there is i {1, 2} such that h( f ()) = i h( f ())
and h( f ()) = i h( f ()). By definition of 1 and 2 , and taking into account
that h( f ()) = F, it follows that i = 2, h( f ()) = t and h( f ()) = 2 t = t.
This means that v() = v() = 1 and so v( ) = 1, since v satisfies (vAnd).
Therefore, v satisfies (vCi).
(vCl): Suppose that v(( )) = 1. By definition of v, f and h, it follows that
h( f (( ))) = 1 h( f ( )) D, hence h( f ( )) = h( f ()) 3
h( f ()) = F. This implies that either h( f ()) = F or h( f ()) = F. By defini-
tion of f and h, there is i {1, 2} such that h( f ()) = i h( f ()) and h( f ()) =
i h( f ()). There are two cases to analyze:
Case 1: i = 1. Then h( f ()) = 1 h( f ()) = T , whence v() = 1.
Case 2: i = 2. If h( f ()) = F then h( f ()) = 2 h( f ()) = T for i = 1, 2 and so
v() = 1. On the other hand, if h( f ()) = 2 h( f ()) = F then h( f ()) = T .
From this, h( f ()) = 2 h( f ()) = T and so v() = 1.
From the previous analysis, it follows that v satisfies clause (vCl).
(vCf ): Suppose that v() = 1. Then, h( f ()) D. There are two cases to
analyze:
Case 1: f () = 1 f (). Then h( f ()) = 1 1 h( f ()) D, whence
h( f ()) D. From this, v() = 1.
 
Case 2: f () = 2 f (). Then h( f ()) 1 2 h( f ()), 2 2 h( f ()) .
If h( f ()) = 1 2 h( f ()) D then h( f ()) = T . If h( f ()) =
2 2 h( f ()) D then h( f ()) D. In both cases, v() = 1.
6.8 The Possible-Translations Semantics 285

From the previous analysis, it follows that v satisfies clause (vCf ).

(vCa1 )(vCa3 ): Assume that v() = v() = 1. If ( # ) = ( ) then
f (( )) = 1 f ( ) and so f (( )) = 1 f ( ).
Thus, h( f (( ))) = 1 h( f ( )) = T . From this, v( # ) = 1. Now,
suppose that ( # )
= ( ). There are four cases to analyze:
Case 1: f () = 1 f (), f () = 1 f (), f () = 1 f () and f () =
1 f (). From this, f (( # )) = 1 f ( # ) and so f (( # )) = 1 f ( # ).
This implies that h( f (( # ))) = 1 h( f ( # )) = T .
Case 2: f () = 1 f (), f () = 1 f (), f () = 2 f () and  f ()
 =
2 f (). Since h( f ()) = 2 h( f ()) D it follows that h( f ()) T, F . On
the other hand, f ( # ) = f () #2 f () and so h( f ( # )) = h(f ()) #2 h( f ()).
From the truth-tables above it follows that h( f ( # )) T, F . This means that
h( f (( # ))) = T .
Case 3: f () = 2 f (), f () = 2 f (), f () = 1 f () and f() = 
1 f (). Since h( f ()) = 2 h( f ()) D it follows that h( f ()) T, F .
Besides this, f ( # ) = f () #1 f () and so h( f ( # ))
 = h( f ()) #1 h( f ()).
From the truth-tables above it follows that h( f ( # )) T, F . This implies that
h( f (( # ))) = T .
Case 4: f () = 2 f (), f () = 2 f (), f () = 2 f () and f() = 
2 f (). From the previous analysis it follows that h( f ()), h( f ()) T, F .
Besides this, f ( # ) = f () #3 f (). Hence, h( f ( # ))
 = h(
 f ()) #3 h( f ()).
From the truth-tables above it follows that h( f ( # )) T, F . As a consequence
of this, h( f (( # ))) = T .
From the analysis of the four cases it follows that v(( # )) = 1, and so v
satisfies clauses (vCa1 )-(vCa3 ).
This means that v is a Cila-valuation, which concludes the proof. 

Proposition 6.8.7 For every Cila-valuation v there exists a mapping f : L

L Cila
PT S
satisfying clauses (T r 0)(T r 5) above, and a valuation h over the matrix W3+
such that, for every formula of Cila: v() = 1 iff h( f ()) D.

Proof Given a Cila-valuation v, consider the valuation h over the matrix W3+ defined
as follows, for every propositional variable p:

T if v( p) = 1 and v( p) = 0
h( p) = t if v( p) = 1 and v( p) = 1

F if v( p) = 0 (and so v( p) = 1).

Since h is a homomorphism, it is enough to define h for the propositional variables,

and extend it homomorphically to the algebra of formulas L Cila PT S
. Now, a mapping
f : L L Cila
PT S
satisfying clauses (T r 0)(T r 5) above will be recursively defined
(w.r.t. the complexity n of the formulas), with the following property: v() = 1 iff
h( f ()) D, for every formula .
286 6 Semantics of Non-deterministic Character for LFIs

Base: Suppose that n = 1. Then is a propositional variable and so f ( p) = p

and f ( p) = 2 p, by (Tr0). Observe that the formulas with complexity 2 are of
the form p, for p V ar . Since h is a homomorphism, h( f ( p)) = 2 h( p), for
every p V ar . From this, it is immediate to see that v() = 1 iff h( f ()) D, for
every formula with complexity 2.
Inductive step: Let n 2 and suppose that, for every formula such that l()
n, f () was defined, and v() = 1 iff h( f ()) D (induction hypothesis, IH).
Observe that, if l() < n then, by (IH),

(A) v()
= v() iff either h( f ()) {T, F} or f () = 1 f ().

Let such that l() = n + 1. The following cases must be analyzed:

1. = ( # ), where # {, , }. Then, f () is defined by using clause (Tr1).
Thus f () = f () #i f () and so h( f ()) = h( f ()) #i h( f ()), for some 1
i 3. It is easy to see that v() = 1 iff h( f ()) D. Take, for instance, the case
when # = . Then h( f ()) = h( f ()) i h( f ()), for some 1 i 3. Thus,
v() = 1 iff, by (vImp), either v() = 0 or v() = 1 iff, by (IH), either h( f ()) = F
or h( f ()) D iff h( f ()) = h( f ()) i h( f ()) D. The proof for # = and
# = is similar.
2. = , with / V ar .
2.1. = ( ). Then f () = 1 f (), by (Tr2.1), and so h( f ()) =
1 h( f ()). Taking into account the properties of Cila-valuations, v() = v() = 1
iff v() = 0 iff, by (IH), iff h( f ()) = F iff, by the truth-table of 1 , h( f ()) =
1 h( f ()) D.
2.2. =( # ), where # {, , }, but ( # )
=( ).
2.2.1. f () = 1 f () and f () = 1 f (). By (Tr2.2), f () = 1 f (). By (A),
v()
= v() and v()
= v(). By the properties of Cila-valuations, v() =
v() = 1 whence v() = 1. From this, v()
= v(). Thus, v() = 1 iff v() =
0 iff, by (IH), h( f ()) = F iff h( f ()) = 1 h( f ()) D.
2.2.2. f () = 1 f () and f () = 2 f (). Then v()
= v(), by (A). By
(Tr1.3), f () = ( f () #2 f ()). Consider the following subcases:
2.2.2.1. v()
= v(). Then, define f () = 1 f (). By a proof analogous to that
for the case 2.2.1, it follows that v() = 1 iff h( f ()) D.
2.2.2.2. v() = v() = 1. By (A), h( f ()) = t. Since f () = ( f () #2 f ())
then h( f ()) = (h( f ()) #2 h( f ())) and so, by the truth-tables of #2 , h( f ())
{F, t}. There are two subcases:
2.2.2.2.1. v() = v() = 1. Then h( f ()) = t, by (IH) and the observation above.
Define f () = 2 f (). Then, v() = 1 and h( f ()) = 2 h( f ()) = 2 t = t
D. That is, v() = 1 iff h( f ()) D.
2.2.2.2.2. v()
= v(). Define f () = 1 f (). Then, v() = 1 iff v() = 0 iff,
by (IH), h( f ()) = F iff h( f ()) = 1 h( f ()) D.
2.2.3. f () = 2 f () and f () = 1 f (). Then v()
= v(), by (A). By
(Tr1.2), f () = ( f () #1 f ()). Consider the following subcases:
6.8 The Possible-Translations Semantics 287

2.2.3.1. v()
= v(). Then, define f () = 1 f (). By a proof analogous to that
for the case 2.2.1, it follows that v() = 1 iff h( f ()) D.
2.2.3.2. v() = v() = 1. By (A), h( f ()) = t. Since f () = ( f () #1 f ())
then h( f ()) = (h( f ()) #1 h( f ())) and so, by the truth-tables of #1 , h( f ())
{F, t}. There are two subcases:
2.2.3.2.1. v() = v() = 1. Define f () = 2 f (). The proof is identical to that
for subcase 2.2.2.2.1.
2.2.3.2.2. v()
= v(). Define f () = 1 f (). The proof is identical to that for
subcase 2.2.2.2.2.
2.2.4. f () = 2 f () and f () = 2 f (). By (Tr1.4), f () = ( f () #3 f ()).
Consider the following subcases:
2.2.4.1. v()
= v() and v()
= v(). Then, define f () = 1 f (). By a proof
analogous to that for the case 2.2.1, it follows that v() = 1 iff h( f ()) D.
2.2.4.2. v()
= v() and v() = v() = 1. By (A), h( f ()) = t. Since f () =
( f () #3 f ()) then h( f ()) = (h( f ()) #3 h( f ())) and so, by the truth-tables of
#3 , h( f ()) {F, t}. Consider the following subcases:
2.2.4.2.1. v() = v() = 1. Define f () = 2 f (). The proof is identical to that
for subcase 2.2.2.2.1.
2.2.4.2.2. v()
= v(). Define f () = 1 f (). The proof is identical to that for
subcase 2.2.2.2.2.
2.2.4.3. v() = v() = 1 and v()
= v(). As in the case 2.2.3.2 (but using
the truth-tables of #3 ) it is proved that h( f ()) {F, t}. Consider the following
subcases:
2.2.4.3.1. v() = v() = 1. Define f () = 2 f (). The proof is identical to that
for subcase 2.2.2.2.1.
2.2.4.3.2. v()
= v(). Define f () = 1 f (). The proof is identical to that for
subcase 2.2.2.2.2.
2.2.4.4. v() = v() = 1 and v() = v() = 1. By (A), h( f ()) = h( f ()) =
t. Since f () = ( f () #3 f ()) then h( f ()) = (h( f ()) #3 h( f ())) = t. By
(IH), v() = 1. Consider the following subcases:
2.2.4.4.1. v() = v() = 1. Define f () = 2 f (). As in subcase 2.2.2.2.1 it is
proved that v() = 1 iff h( f ()) D.
2.2.4.4.2. v()
= v(). Then, v() = 0. Define f () = 1 f (). Consequently,
h( f ()) = 1 h( f ()) = 1 t = F. From this, v() = 1 iff h( f ()) D.
2.3. = . There are two cases to analyze:
2.3.1. f () = 1 f (). In this case, by (Tr3.1), f () = 1 f (). By (A), v()
=
v(). If v() = v() = 1 then, by (vCf ), v() = 1 = v(), a contradiction. Then
v()
= v(). From this, v() = 1 iff v() = 0 iff, by (IH), h( f ()) = F iff
h( f ()) = 1 h( f ()) D.
2.3.2. f () = 2 f (). There are two subcases to analyze:
2.3.2.1. v() = v() = 1. For this first subcase, define f () = 2 f (). By (IH),
h( f ()) = 2 h( f ()) D and so, by the truth-table of 2 , h( f ()) {t, F}. If
h( f ()) = F then, by (IH), v() = 0 and so, by (vCf ), v() = 0, a contradiction.
Therefore h( f ()) = t and then h( f ()) = 2 t = t. From this, h( f ()) = 2 t =
t. This shows that v() = 1 iff h( f ()) D.
288 6 Semantics of Non-deterministic Character for LFIs

2.3.2.2. v()
= v(). For this second subcase, define f () = 1 f (). Thus,
v() = 1 iff v() = 0 iff, by (IH), h( f ()) = F iff h( f ()) = 1 h( f ()) D.
2.4. = . Then, by (Tr4.1), f () = 1 f (). By the properties of Cila-valuations,
it is clear that v()
= v(). As in the case 2.3.2.2 it is proved that v() = 1 iff
h( f ()) D.
3. = . There are two cases to analyze:
3.1. f () = 1 f (). Then, by (Tr5.1), f () = 1 f (). By (A), v()
= v().
From this, and by the properties of Cila-valuations, v() = 1. On the other hand,
h( f ()) = 1 h( f ()) = T , by definition of 1 . It thus follows that v() = 1 iff
h( f ()) D.
3.2. f () = 2 f (). Thus f () = 2 f (), by (Tr5.2). Hence v() = 1 iff, by
the properties of Cila-valuations, v()
= v(), iff, by (IH) and the truth-table of
2 , h( f ()) {T, F} iff, by the truth-table of 2 , h( f ()) = 2 h( f ()) D.
This concludes the proof. 

Theorem 6.8.8 (Soundness and completeness of Cila w.r.t. PTSCila ) The structure
PTSCila (see Definition 6.8.5) is a possible-translations semantics for the logic Cila.
That is, for every set of formulas  {} of Cila, it holds:

 Cila iff f i [] |=W3+ f i (), for every i I.

Proof Let  {} be a set of formulas of Cila.

(Soundness): Suppose that  Cila . Let i I and consider a valuation h over the
matrix W3+ such that h( f i ()) D for every . By Proposition 6.8.6, there exists
a Cila-valuation v such that, for every formula of Cila: v() = 1 iff h( f ()) D.
From this, v() = 1 for every . Given that  Cila then, by soundness of
Cila w.r.t. Cila-valuations, v() = 1 and so, by construction of v, h( f i ()) D.
This shows that f i [] |=W3+ f i (), for every i I .
(Completeness): Suppose that f i [] |=W3+ f i (), for every i I . Let v be a
Cila-valuation such that v[] {1}. By Proposition 6.8.7, there exists a mapping
f : L L Cila
PT S
satisfying clauses (T r 0)-(T r 5) above, and a valuation h over the
+
matrix W3 such that, for every formula of Cila: v() = 1 iff h( f ()) D. By Def-
inition 6.8.5, f = f i for some i I . Now, by construction of f i and h, h( f i ()) D
for every  whence h( f i ()) D, by hypothesis. From this, v() = 1. By com-
pleteness of Cila w.r.t. Cila-valuations, it follows that  Cila . 

6.8.3 Some Remarks on Possible-Translations Semantics

The PTSs turned out to be well adapted to LFIs, as they offer a quite natural account
of understanding the phenomenon of a sentence and its negation being both true.
Such an understanding cannot be done, for instance, by means of finitely many-
valued matrices, since several paradigmatic LFIs cannot be characterized by means
of finite matrices (even with an arbitrary number of truth-values), as it was shown in
6.8 The Possible-Translations Semantics 289

Chap. 4. Additionally, in [30] it was shown that possible-translations semantics can

represent Nmatrices, and so the latter is a particular case of the former. Moreover,
from the previous result about uncharacterizability of Cila by means of a finite
Nmatrix, it is clear that PTSs are strictly more powerful than Nmatrices and swap
structures, offering so a wider degree of controlled non-determinism, as required by
some logics.
The first historical systems of da Costa have been introduced only proof-
theoretically, and only some years later bivalued semantics have been proposed for
their interpretation. The PTSs, as mentioned, appeared later, as an alternative seman-
tics.8 As it was proved in [28] (for da Costas logic C1 ) and in Propositions 6.8.6
and 6.8.7 above (for the logic Cila), bivalued semantics and possible-translations
semantics are in fact equivalent for these systems. However, PTSs have a much
more algebraic and operational flavor than bivalued semantics, which interpret the
non-deterministic character of these logics in a somewhat crude way. The study of
PTSs defined over matrix logicsas the ones presented in this sectionunder the
perspective of algebraic logic (for instance, as controlled restrictions to Nmatrix
semantics) deserves future research.
A particularization of the PTSs called society semantics was also proposed at
a later stage (see [31, 32]) as a way to combine copies of a many-valued logic in
order to obtain a higher-degree many-valued logic. Regarding valuations as agents,
the society semantics can be seen as a way to combine divergent world-views (of
agents) into a tolerant logic society.
The intuition around PTSs has been an inspiring idea to treat other theoretical
constructions. The so-called exogenous semantics approach to probabilistic logic in
[33] introduces a branching time logic for reasoning about probabilistic programs.
The exogenous semantics can be seen as a variant of PTSs for paraconsistent logic.
PTSs also motivated the idea of a sociology of mathematical objects, in analogy
to the study of logics through its relations with other logics (see [34]). The process
of combination of logics has two main aspects: (i) splitting of logics: an analyti-
cal process to decompose logics, and (ii) splicing of logics: a synthetic process to
combine (or compose) logics.
The PTSs can be seen as an instance of the splitting process: a given logic system
is decomposed into other (simpler) systems, providing, for instance a conservative
translation of the logic in analysis into a product (or weak product) of simpler or
better known logics. These aspects are reflected in the attempts to viewing logic
systems as a category with suitable properties, as in [34].
The book [35] is entirely dedicated to a mathematical theory of combining and
decomposing logics, from propositional and first-order based logics to higher-order
based logics as well as to non-truth functional logics (including mechanisms for com-
bining semantic structures and deductive systems). The PTSs occupy an important
position in the book, as a method for splitting and splicing logics (see also [36]).
Although from the beginning designed for the LFIs, the PTSs also apply to other
logics, as to ukasiewicz logics and further. As generalizations of the concept of

8 ThePTSs have been originally called non-deterministic semantics, but later the name was
changed.
290 6 Semantics of Non-deterministic Character for LFIs

truth-tabularity and non-deterministic matrices, PTSs extend these approaches in

terms of the preservation of certain meta-theoretical properties such as decidability
and compactness. All this witness the wide scope of possible-translations semantics
as an apt semantic tool, and their higher degree of generality.

References

1. Mortensen, Chris. 1980. Every quotient algebra for C1 is trivial. Notre Dame Journal of Formal
Logic 21(4): 694700.
2. Blok, Willem J., and Don Pigozzi. 1989. Algebraizable logics. Memoirs of the American
Mathematical Society, vol. 77(396). American Mathematical Society, Providence, RI, USA.
3. Czelakowski, Janusz. 2001. Protoalgebraic Logics, vol. 10. Trends in Logic Series. Dordrecht:
Kluwer Academic Publishers.
4. Font, Josep Maria, Ramn Jansana, and Don Pigozzi. 2003. A survey of abstract algebraic
logic. Studia Logica 74(12): 1397.
5. Font, Josep Maria, and Ramn Jansana. 2009. A General Algebraic Semantics for Sentential
Logics. Vol. 7 of Lecture Notes in Logic, 2nd edn. Ithaca, NY, USA: Association for Symbolic
Logic.
6. Font, Josep Maria. 2016. Abstract algebraic logic: An introductory textbook. Mathematical
Logic and Foundations Series, vol. 60. London: College Publications.
7. Lewin, Renato A., Irene Mikenberg, and Maria G. Schwarze. 1991. C1 is not algebraizable.
Notre Dame Journal of Formal Logic 32(4): 609611.
8. Carnielli, Walter A., and Joo Marcos. A taxonomy of C-systems. In [Walter A. Carnielli,
Marcelo E. Coniglio, and Itala M.L. DOttaviano, editors. Paraconsistency: The Logical
Way to the Inconsistent. Proceedings of the 2nd World Congress on Paraconsistency (WCP
2000),volume 228 of Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, New
York, 2002.], pp. 194.
9. Carnielli, Walter A., and Marcelo E. Coniglio. 2014. Swap structures for LFIs. CLE e-Prints,
14(1). http://www.cle.unicamp.br/e-prints/vol_14,n_1,2014.html.
10. Odintsov, Sergei P. 2008. Constructive Negations and Paraconsistency. Trends in Logic, vol.
26. Springer.
11. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Inconsis-
tency. In Handbook of Philosophical Logic (2nd edn.), vol. 14, ed. Dov M. Gabbay, and Franz
Guenthner, 193. Springer. doi:10.1007/978-1-4020-6324-4_1.
12. Quiroga, Vernica. 2013. An alternative definition of F-structures for the logic C1 . Bulletin of
the Section of Logic 42(34): 119134.
13. Rescher, Nicholas. 1962. Quasi-truth-functional systems of propositional logic. The Journal
of Symbolic Logic 27(1): 110.
14. Kearns, John T. 1981. Modal semantics without possible worlds. The Journal of Symbolic Logic
46(1): 7786.
15. Ivlev, Yury V. 1988. A semantics for modal calculi. Bulletin of the Section of Logic 17(3/4):
114121.
16. Coniglio, Marcelo E., Luis Farias del Cerro, and Newton M. Peron. 2015. Finite non-
deterministic semantics for some modal systems. Journal of Applied Non-Classical Logic,
25(1): 2045. doi:10.1080/11663081.2015.1011543.
17. Avron, Arnon, and Iddo Lev. 2001. Canonical propositional Gentzen-type systems. In Proceed-
ings of the 1st International Joint Conference on Automated Reasoning (IJCAR 2001). LNAI,
vol. 2083, ed. Rajeev Gore, Alexander Leitsch, and Tobias Nipkow, 529544. Springer.
18. Avron, Arnon, and Iddo Lev. 2005. Non-deterministic multi-valued structures. Journal of Logic
and Computation 15(3): 241261.
References 291

19. Marty, Frdric. 1934. Sur une gnralisation de la notion de groupe (in French). In Huitime
Congrs de Mathmaticiens Scandinaves, 4549, Stockholm.
20. Avron, Arnon. 2005. Non-deterministic matrices and modular semantics of rules. In Logica
universalis, ed. Jean-Y. Bziau, 149167. Basel: Birkhuser Verlag.
21. Avron, Arnon. 2005. Non-deterministic semantics for paraconsistent C-systems. In Symbolic
and quantitative approaches to reasoning with uncertainty. Proceedings of the 8th European
conference ECSQARU 2005, Barcelona, Spain, July 6-8, 2005. Lecture Notes in Computer
Science, vol. 3571, ed. Llus Godo, 625637. Springer.
22. Avron, Arnon. 2007. Non-deterministic semantics for logics with a consistency operator. Jour-
nal of Approximate Reasoning 45: 271287.
23. Fidel, Manuel M. 2003. Nuevos enfoques en Lgica Algebraica (New approaches to Algebraic
Logic, in Spanish). Ph.D. thesis, Universidad Nacional del Sur, Bahia Blanca, Argentina.
24. Blyth, Thomas S., and Jules C. Verlet. 1994. Ockham Algebras. Oxford: Oxford Science Pub-
lications, Oxford University Press.
25. Fidel, Manuel M. 2009. Una nueva semntica de tipo algebraico para las lgicas paraconsis-
tentes Cn de da Costa (A new algebraic-like semantics for da Costas paraconsistent logics
Cn , in Spanish. Unpublished draft. Presented at the Reunin Anual de la UMA, Mar del Plata,
Argentina.
26. Carnielli, Walter A. 1990. Many-valued logics and plausible reasoning. In Proceedings of the
Twentieth International Symposium on Multiple-Valued Logic, ed. George Epstein, 328335.
Charlotte, NC, USA: The IEEE Computer Society Press.
27. Carnielli, Walter A. 2000. Possible-translations semantics for paraconsistent logics. In Fron-
tiers in Paraconsistent Logic: Proc, ed. Diderik Batens, Chris Montersen, Graham Priest, and
Jean-Paul Van Bendegem, 149163. The I World Congress on Paraconsistency. Hertfordshire:
Research Studies.
28. Marcos, Joo. 1999. Semnticas de tradues possveis (Possible-translations semantics, in
Portuguese). Masters thesis, IFCH, State University of Campinas. http://www.bibliotecadigital.
unicamp.br/document/?code=vtls000224326&opt=4&lg=en_US.
29. Marcos, Joo. 2008. Possible-translations semantics for some weak classically-based paracon-
sistent logics. Journal of Applied Non-Classical Logics 18(1): 728.
30. Carnielli, Walter A., and Marcelo E. Coniglio. 2005. Splitting logics. In We Will Show Them:
Essays in Honour of Dov Gabbay, vol. 1, ed. Sergei Artemov, Howard Barringer, Artur dAvila
Garcez, Luis C. Lamb, and John Woods, 389414. Kings College Publications.
31. Carnielli, Walter A., and Mamede Lima-Marques. 1999. Society semantics and multiple-valued
logics. In Advances in Contemporary Logic and Computer Science. Proceedings of the XI
Brazilian Conference on Mathematical Logic, May 1996, Salvador, Bahia, Brazil. Contem-
porary Mathematics, vol. 235, ed. Walter A. Carnielli and Itala M.L. DOttaviano, 3352.
American Mathematical Society.
32. Fernndez, Vctor L., and Marcelo E. Coniglio. 2003. Combining valuations with society
semantics. Journal of Applied Non-Classical Logics 13(1): 2146. doi:10.3166/jancl.13.21-
46.
33. Baltazar, Pedro, Paulo Mateus, Rajagopal Nagarajan, and Nikolaos Papanikolaou. 2007. Exoge-
nous probabilistic computation tree logic. Electronic Notes in Theoretical Computer Science
190(3): 95110.
34. Mariano, Hugo L., and Caio Mendes. 2011. Towards a good notion of category of logics. In
TACL 2011The Fifth International Confererence on Topology, Algebra and Categories in
Logic, 2011, Marseille-France, 207210. Preprint available at arXiv:1404.3780.
35. Carnielli, Walter A., Marcelo E. Coniglio, Dov M. Gabbay, Paula Gouveia, and Cristina Ser-
nadas. 2008. Analysis and Synthesis of Logics: How to Cut and Paste Reasoning Systems.
Applied Logic Series, vol. 35. Springer. doi:10.1007/978-1-4020-6782-2.
36. Carnielli, Walter A., and Marcelo E. Coniglio. 2014. Combining logics. In The Stanford Ency-
clopedia of Philosophy, ed. Edward N. Zalta. Spring 2014 edn.
Chapter 7
First-Order LFIs

In the previous chapters, LFIs have been approached exclusively from the propo-
sitional viewpoint. This is justified by the fact that the main notions and issues of
paraconsistency in general, and LFIs, in particular, occur at the propositional level,
related to their main connectives, namely, paraconsistent negation, consistency and
inconsistency operators.
However, envisaged applications of LFIs to several contexts (linguistics, databases,
set theory, formal epistemology and so on) demand more than propositional lan-
guages, and richer languages may be required. The first, natural option, is to consider
first-order languages.
There are several proposals in the literature concerning the development of first-
order LFIs. Originally, da Costa in [1] (see also [2]) introduced the first-order ver-
sion Cn for each calculus Cn . For a fixed n, the calculus Cn is obtained from Cn by
adding the first-order axioms and inference rules of classical first-order logic, plus
the requirement of the propagation of consistency through quantifiers. This makes
sense in his framework, since this requirement generalizes the propagation of con-
sistency for conjunction and disjunction, taking into account that the universal and
the existential quantifiers can be regarded as arbitrary conjunctions and disjunctions,
respectively. He also investigated the extension of these systems with the (standard)
equality predicate, introducing a hierarchy of calculi known as Cn= . Later on, Alves
obtained several basic results of model theory for such calculi (see [3]).
The semantics for the calculi Cn and Cn= is, as much as the calculus which will
be analyzed in this chapter, 2-valued. In this way the usual Tarskian first-order struc-
tures are equipped with a paraconsistent valuation (similar to the ones considered
in Chaps. 2 and 3). The basic model theory of the first-order version of the 3-valued
paraconsistent logic J3 (recall Sect. 4.4.3 of Chap. 4) was developed by DOttaviano
in [47]. An alternative approach to 3-valued first-order LFIs will be described in
Sect. 7.9.
By their turn, Avron and Zamansky proposed in [8] a semantics based on non-
deterministic matrices, or Nmatrices, for several first-order LFIs. As mentioned in

Springer International Publishing Switzerland 2016 293

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_7
294 7 First-Order LFIs

Sect. 6.4 of Chap. 6, the truth-values as well as the operations of the Nmatrices are
obtained from an analysis of the valuation semantics for LFIs examined in Chaps. 2
and 3, and so this approach is conceptually very close to the one to be presented here.
This chapter is based on the definitions and results presented in the papers [9, 10],
including some new results and improvements. An application of this framework to
paraconsistent set theory will be developed in Chap. 8.

7.1 The Logic QmbC

This section introduces the logic QmbC, the first-order LFI to be investigated in
detail in this chapter.

Definition 7.1.1 Assume the propositional signature  for mbC, as well as the sym-
bols (universal quantifier) and (existential quantifier), together with punctuation
marks (commas and parentheses). Let V ari = {v1 , v2 , . . .} be a denumerable set of
individual variables. A first-order signature  for LFIs is composed of the following
elements:
a set C of individual constants;
for each n 1, a set of function symbols of arity n;
for each n 1, a set of predicate symbols of arity n. 

By convenience, given a first-order signature , it will be assumed that it has

at least one predicate symbol. The set of terms and of formulas of  (which are
recursively defined, as usual) will be denoted by T and L  , respectively. The set
L  is the first-order language generated by signature .
As usual when dealing with first-order languages, it is useful to consider two
notions of complexity: complexity of terms and complexity of formulas. The for-
mer will be defined as usual. The latter, however, will be slightly modified from the
usual definition, by extending the notion given for propositional languages in Defin-
ition 2.1.11. The reasons for this change (involving the complexity of the quantified
formulas) will be clear on Sect. 7.9.

Definition 7.1.2 (Terms Complexity) Let  be a first-order signature. The complex-

ity of a given term t T , denoted by l(t), is recursively defined as follows:
1. If t V ari C, then l(t) = 1;
2. If t = f (t1 , . . . , tn ), for 
f a function symbol of arity n and t1 , . . . , tn T (where
n > 0), then l(t) = 1 + nj=1 l(t j ). 

Definition 7.1.3 (Formula Complexity) Let  be a first-order signature. The com-

plexity of a given formula L  , denoted by l(), is recursively defined as follows:
1. If is atomic, that is, = P(t1 , . . . , tn ), for P a predicate symbol of arity n and

t1 , . . . , tn T (where n > 0), then l() = 1;

2. If = , then l() = l() + 1;
7.1 The Logic QmbC 295

3. If = , then l()
= l() + 2;

4. If = #, where # {, , }, then l()
= l()
+ l() + 1;

5. If = Qx, where Q {, }, then l()
= l() + 2. 

The notions of subformula, scope of an occurrence of a quantifier in a formula, free

and bound occurrences of a variable in a formula, and of a term free for a variable in
a formula, are the usual ones (the reader is referred to [11, 12] for precise definitions
of these notions).
Given , the set of atomic formulas and the set of sentences (or closed sentences
i.e., formulas without free variables) over  are denoted by At and SL  , respectively.
Given a formula , the formula obtained from by substituting every free occur-
rence of variable x by the term t will be denoted by [x/t]. The following concept
will be useful in our context:

Definition 7.1.4 Let and be formulas. If can be obtained from by means

of addition or deletion of void quantifiers, or by renaming bound variables (keeping
the same free variables in the same places), we say that and are variants of each
other. 

The system mbC, the basic LFI studied in Chap. 2, will be now extended to
first-order languages, producing a logic called QmbC which is defined as follows:

Definition 7.1.5 Let  be a first-order signature. The logic QmbC (over ) is

defined by the Hilbert calculus obtained by extending mbC (recall Definition 2.1.12)
expressed in the language L  by adding the following:
Axiom Schemas

[x/t] x, if t is a term free for x in (Ax11)

x [x/t], if t is a term free for x in (Ax12)
, whenever is a variant of (Ax13)

Inference rules:

if x is not free in (-In)
x

if x is not free in (-In)
x

The consequence relation of QmbC will be denoted by QmbC . Thus, if 

{} L  then  QmbC will denote that there exists a derivation in QmbC of
from .
296 7 First-Order LFIs

7.2 Basic Properties of QmbC

Some useful theorems and meta-theorems of QmbC will be established in this

section. A fixed first-order signature  will be assumed henceforward.

Theorem 7.2.1
1. For every formula : QmbC .
2. If  QmbC and  QmbC then , QmbC .
3. If  QmbC and  QmbC then  QmbC .
4. If  QmbC then  QmbC x .
5. If x is not free in then x( ) QmbC x .
6. For every formulas , , : QmbC ( ) (( ) ( )).
7. For every formulas , , : QmbC ( ( )) ( ( )).
8. For every formulas , , : QmbC ( ) (( ) ( )).
9. If  QmbC then , QmbC .
10. , QmbC .
11. If QmbC then QmbC ( ) ( ), for every formula .

Proof Item 1: As it is well known, any Hilbert calculus which contains axiom
schemas (Ax1), (Ax2) and (MP) can derive the schema . For the sake of
completeness, a classical derivation of this theorem (taken from [12]) is here dis-
played:
1. ( (( ) )) (( ( )) ( )) (Ax2)
2. (( ) ) (Ax1)
3. ( ( )) ( ) (MP 1,2)
4. ( ) (Ax1)
5. (MP 3,4).
Item 2: By hypothesis, it holds that: , QmbC and , QmbC
. But QmbC ( ) (( ) (( ) )), by (Ax8), and so
, QmbC ( ) , by using MP twice. By applying MP once again, it
follows that , QmbC .
Item 3: This is a consequence of Item 2, by observing that QmbC ,
by (Ax10).
Item 4: Consider the (meta)derivation below:

 QmbC Hypothesis
 
 QmbC x Ax1
 QmbC x MP
 QmbC x x -In
 QmbC x x Item 1
 QmbC x Item 2
7.2 Basic Properties of QmbC 297

Item 5: Consider the derivation in QmbC below:

1. x( ) (premise)
2. x( ) ( ) (Ax12)
3. (MP 1,2)
4. x (-In 3).
Item 6: Recall the valuation semantics for mbC given in Definition 2.2. It is thus
easy to see that ( ) (( ) ( )) is a valid formula. By the
completeness of mbC w.r.t. valuations established in Theorem 2.2.8, that formula is
derivable in mbC, for every , , . Since QmbC extends mbC, it follows that the
schema ( ) (( ) ( )) is derivable in QmbC .
Items 7 and 8: The proof is identical to that of Item 6.
Item 9: Consider the (meta)derivation below:

 QmbC Hypothesis
, QmbC Monotonicity
, QmbC MP

Item 10: This follows from Item 6 and Item 9 (used two times).
Item 11: This is a consequence of item 8 and MP. 
Recall from Sect. 2.4 of Chap. 2 that a strong (classical) negation can be defined in
mbC as = , where = ( ( )) is a bottom formula,1 for
any formula . In the case of first-order languages, the formula must be a sentence.
For simplicity, we designate a privileged one, and the subscript will be omitted in
and from now on, as it was done with mbC and its propositional extensions.
Proposition 7.2.2 (Strong negation) The strong negation satisfies the following
properties in mbC (and, therefore, also in QmbC ):
(i) QmbC ( ) for every and ;
(ii) QmbC
(iii) QmbC and QmbC
(iv) If ( QmbC ) and (, QmbC ) then (,  QmbC )
(v) QmbC ( ) ( ) and so QmbC
(vi) QmbC ( ) ( ) and so QmbC
(vii) QmbC ( ) ( ) and so QmbC
(viii) QmbC ( ( )) ( ( ))
(ix) QmbC ( ( )) ( ( ))
(x) QmbC ( ( )) ( ( ))
(xi) QmbC ( ) ( ) and so QmbC
(xii) QmbC ( ) ( )
(xiii) QmbC
(xiv) QmbC x x

1 That is: QmbC for every .

298 7 First-Order LFIs

(xv) QmbC x x
(xvi) QmbC ( ( x)) ( ( x))
(xvii) If QmbC then QmbC x x
(xviii) QmbC x( ) (x ), if x does not occur free in .

Proof Items (i)(iii), (v)(vii) and (xi)(xiii): It follows from the completeness of
mbC with respect to valuations (see Theorem 2.2.8), by observing that v() = 0,
and v() = 1 iff v() = 0, for every formula and every valuation v for mbC. An
argument similar to the proof of Theorem 7.2.1 Item 6 can thus be employed.
Item (iv): follows from Item (ii) and from Item 2 of Theorem 7.2.1.
Item (viii): follows from Item (v) and Item 11 from Theorem 7.2.1
Item (ix): follows from Item (vi) and Item 11 from Theorem 7.2.1
Item (x): follows from Item (vii) and Item 11 from Theorem 7.2.1
Item (xiv): Since QmbC x , by (Ax12), then QmbC x,
by Item (vii). By rule (-In), QmbC x x whence QmbC x
x, using again Item (vii).
Item (xv): From (Ax11) it follows that QmbC x and then, by Item (xi),
QmbC x . The result is obtained then by (-In).
Item (xvi): From Item (xv) and Item 11 of Theorem 7.2.1, QmbC  ( x) 
(
 x). Using again
 Item 11 of Theorem 7.2.1,  QmbC ( x)
( x) .
Item (xvii): Using (Ax12), QmbC x and then, by hypothesis and Item 10
from Theorem 7.2.1, QmbC x . By rule (-In), QmbC x x.
Item (xviii): From (Ax12), QmbC x( ) ( ) and then, by Item 7
from Theorem 7.2.1 and MP, QmbC (x( ) ). By applying
(-In), QmbC x (x( ) ). The result follows again by Item 7 from
Theorem 7.2.1 and MP. 

A most useful property of QmbC, the Deduction Meta-Theorem (DMT), will be

proved in the sequel. As it happens with classical first-order logic, this property does
not hold in the general case, but does hold under certain assumptions concerning free
variables of the formula being discharged from the assumptions. The proof will be
done by adapting the corresponding one for classical logic presented in [12].

Definition 7.2.3 Let d = 1 , . . . , n be a derivation in QmbC from a set of hypoth-

esis , and let . We say that i depends upon in d if:
i = ; or
i is obtained from j and k (with j, k < i) by (MP), where j or k depend
upon in d; or
i is obtained from j (with j < i) by (-In), where j depends upon in d; or
i is obtained from j (with j < i) by (-In), where j depends upon in d. 

Clearly, the notion above can be adjusted to any Hilbert calculus, by considering its
specific inference rules. The next result also holds in any Hilbert calculus (see [12]).
7.2 Basic Properties of QmbC 299

Lemma 7.2.4 If does not depend upon in the derivation of from  {},
then  QmbC .
Theorem 7.2.5 (Deduction Meta-Theorem (DMT) for QmbC ) Suppose that there
exists in QmbC a derivation of from  {}, such that no application of the rules
(-In) and (-In) to formulas that depend upon have as their quantified variables
free variables of . Then  QmbC .
Proof Let d = 1 , . . . , n be a derivation in QmbC of from  {}, satisfying
the conditions of the hypothesis of the theorem; then n = . By induction on n, it
will be proved that  QmbC i for every 1 i n. From this it will follow
that  QmbC , as required.
The proof is virtually identical to the one in [12] for first-order classical logic,
with the exception of the rules for quantification (which are different from the rules
of QmbC ), and so this case is the only one to be treated here. It is worth noting that
the part of the proof in [12] omitted here uses that is a theorem; this result,
however, was proved here in Item 1 of Theorem 7.2.1.
Suppose, thus, that  QmbC j for every 1 j < i, with i 2. By the
considerations above, there are only two cases that should be analyzed:
(1) There exists j < i such that j = and i = x (with x not free in
) is obtained from j by (-In). By induction hypothesis,  QmbC j and,
by the hypothesis on d, either j does not depend upon or x does not occur free in
. There are two subcases to be taken into account:
(1.1) j does not depend upon . By Lemma 7.2.4,  QmbC j , that is, 
QmbC . By applying rule (-In) it follows that  QmbC x , that
is,  QmbC i . From this,  QmbC i .
(1.2) x does not occur free in . As  QmbC j , that is,  QmbC
( ), then  QmbC ( ), by Item 7 of Theorem 7.2.1. Now, by
(-In) and Item 7 of Theorem 7.2.1 again,  QmbC (x ). That is,
 QmbC i .
(2) There exists j < i such that j = and i = x (with x not free
in ) is obtained from j by (-In). The proof of this case is quite similar to that of
Case 1, and so just the second subcase will be given:
(2.2) x does not occur free in . As  QmbC j , that is,  QmbC
( ), then  QmbC ( ) and so  QmbC ( ),
by Item (viii) of Theorem 7.2.2 and Item 7 of Theorem 7.2.1 respectively. As x does
not occur free in ( ), then  QmbC x ( ) follows by (-In).
Using again Theorem 7.2.1, Item 7,  QmbC
(x ). Thus  QmbC ( x) follows by Item
(x) from Theorem 7.2.2. From this,  QmbC ( x) follows by Item (xvi)
of Theorem 7.2.2, that is,  QmbC i . 
The general version of (DMT) is quite intricate, and so it could be complicated to
determine the conditions under which it can be applied in concrete situations. The
following particular case, which is simpler than the general case, is enough for most
applications.
300 7 First-Order LFIs

Corollary 7.2.6 ((DMT), simplified version) Suppose that there exists in QmbC a
derivation of from  {}, such that no application of the rules (-In) and (-In)
have, as their quantified variables, free variables of (in particular, this holds when
is a sentence). Then  QmbC .

Corollary 7.2.7
1. If x is not free in then x( )  x.
2.  x( ) ( x), if x is not free in .

Proof Item 1 can be obtained by (Ax12), MP and -In. Item 2 follows from Item 1
through (DMT). 

The Deduction Meta-Theorem is a rather useful tool which allows to simplify

derivations in QmbC . As a simple application, the following result, which will be
used in the sequel, is proved by using (DMT):

Theorem 7.2.8 If QmbC then QmbC and QmbC

, for every .

Proof For the first one, consider the following (meta) derivation in QmbC :

, QmbC MP
QmbC Hypothesis
, QmbC Transitivity

Therefore, by (DMT), QmbC .

For the second one, the following (meta)derivation suffices:

QmbC Hypothesis
, QmbC Monotonicity
, QmbC Reflexivity
, QmbC MP

Therefore, by (DMT), QmbC . 

Some key results involving quantifiers can now be obtained. These results are
essential in order to prove the Completeness theorem for QmbC, more precisely
when proving that non-trivial theories can be conservatively extended to non-trivial
Henkin theories (Theorem 7.5.3). Recall that denotes the strong negation.

Lemma 7.2.9 In QmbC the following hold:

(i) QmbC ( ) implies QmbC (x x)
(ii) QmbC x x
(iii) QmbC x x
7.2 Basic Properties of QmbC 301

(iv) QmbC x x
(v) QmbC (x ) x( ) if x does not occur free in
(vi) QmbC ( x ) x( ) if x does not occur free in .

Proof (i) Suppose that QmbC ( ). Then QmbC ( x), by (Ax11), and
so QmbC ( x), by the transitivity of . The result follows by applying rule
(-In).
(ii) Since QmbC , by Proposition 7.2.2(iii), the result follows from
Item (i).
(iii) By (Ax11), QmbC x. By Proposition 7.2.2(xi), QmbC x
. By (-In), QmbC x x. Finally, by Proposition 7.2.2(xi), QmbC
x x.
(iv) By Item (iii), QmbC x x. The result follows by Item (ii) and
by the transitivity of .
(v) By MP, (x ), x QmbC . But QmbC ( ) and QmbC
( ) x( ), by (Ax1) and (Ax11), respectively. Thus (x ), x
QmbC x( ) and so (x ) QmbC x x( ), by (DMT). On
the other hand, QmbC ( ), by Proposition 7.2.2(i), and so QmbC
x x( ), by Item (i). But QmbC x x, by Item (iii), there-
fore QmbC x x( ), by the transitivity of . From this, (x
), x QmbC x( ). Thus, by (DMT) and Proposition 7.2.2(iv), (x
) QmbC x( ). The result follows by (DMT), as x does not occur free in .
(vi) From ( x ), x( ) it follows that ( x ), ( ),
by (Ax12), and from this ( x ), , , by Proposition 7.2.2(xii), (Ax4)
and (Ax5). From this x , is obtained by MP. But = and
QmbC x , by (-In). That is, QmbC x . Combining this with
the inference above, from ( x ), x( ) it follows that x , x
and from this one obtains . Therefore, ( x ) QmbC x( ), by
(DMT) and the definition of . By Item (iv), ( x ) QmbC x( ). As
x does not occur free in , the result follows again by (DMT). 

Lemma 7.2.10 If x does not occur free in and , the following holds in QmbC :
1. If  QmbC ( ) then  QmbC (x )
2. If  QmbC ( ) then  QmbC (x )
3. If  QmbC ( ) then  QmbC ( x )
4. If  QmbC ( ) then  QmbC ( x ).

Proof 1. By Lemma 7.2.9(v), QmbC (x ) x( ) and so x

QmbC x( ). Hence, by Theorem 7.2.8, QmbC (x( ) )
((x ) ).
Suppose, thus, that  QmbC ( ) . Then  QmbC x( ) ,
by (-In). By the observation above,  QmbC (x ) .
2. Consider the following derivation in QmbC :
302 7 First-Order LFIs

1. (premise)
2. x (premise)
3. x (Ax12)
4. (MP 2,3)
5. (MP 1,4).
Thus, , x QmbC , and so, by (DMT), it follows that QmbC
x . The result follows by the transitivity of derivations.
3. By Lemma 7.2.9(vi), QmbC ( x ) x( ), and so x
QmbC x( ). By Theorem 7.2.8, it follows that QmbC (x( )
) (( x ) ).
Suppose, now, that  QmbC ( ) . By (-In) it follows that  QmbC
x( ) . Then  QmbC ( x ) , by the observation above.
4. Consider the following derivation in QmbC :
1. (premise)
2. (premise)
3. (MP 1,2)
4. x (Ax11)
5. x (MP 3,4).
Thus, , QmbC x. By (DMT), QmbC x. Therefore, if
 QmbC then  QmbC x, by transitivity of derivations. 

The results concerning the Hilbert calculus QmbC stated in this section are
sufficient for the enterprise of semantics, as well as for clarifying certain subtleties
therein.

7.3 Tarskian Paraconsistent Structures

The proposed semantics for QmbC consists of usual Tarskian first-order structures,
together with paraconsistent valuations. In the following sections the soundness and
completeness of QmbC with respect to such interpretations will be obtained.

Definition 7.3.1 (Structures) Let  be a first-order signature (see Definition 7.1.1).

A (first-order) structure over  is pair A = A, IA  such that A is a nonempty set
(the domain of the structure) and IA is an interpretation mapping which assigns, to
each individual constant c C, an element IA (c) of A; to each function symbol f
of arity n, a function IA ( f ) : An A; and to each predicate symbol P of arity n, a
relation IA (P) An . 

A structure A over  induces an interpretation mapping ()A : C T A from

the set C T of closed terms (that is, terms without variables) of  to the set A. This
mapping is recursively defined as follows:
7.3 Tarskian Paraconsistent Structures 303

cA = IA (c) if c is an individual constant;

f (t1 , . . . , tn )A = IA ( f )(t1A , . . . , tnA ) if f is a function symbol of arity n and
t1 , . . . , tn C T .

From now on, the expressions f A and P A will be used instead of IA ( f ) and
IA (P), for a function symbol f and a predicate symbol P, respectively.
In order to interpret the quantifiers in a given structure A, it is helpful to give
a formal name to each element of the domain A of A by means of new individual
constants.

Definition 7.3.2 (Diagram languages and extended structures) Consider a structure

A over . The diagram language of A, denoted by L  (A), or simply L(A), is defined
over the signature  A obtained from  by adding a new individual constant a for
each element a of the domain A of A. We will denote by T (A), or simply by T (A),
the set of terms of the diagram language of A.
The structure A is naturally extended to a structure A = A, IA  over  A by
defining IA (
a) = a for every a A. 

The next step is to extended the mbC-valuations to QmbC-valuations, which

are able to interpret formulas in a given structure. For technical reasons, it will be
necessary to deal with some notions introduced in the following definition:

Definition 7.3.3 (Multiple substitution) Let A be a structure for a signature , and

x = x1 , . . . , xn a sequence of different variables. The set of formulas of L(A) whose
free variables occur in the sequence x is denoted by L(A)x , and we say that x is
a context for the formulas in L(A)x . The set L x of all the formulas of L  with
context x is defined analogously. Given a sequence a = a1 , . . . , an of elements in
A and L(A)x , we denote by [ x /
a ] the sentence of SL(A) , obtained from by
simultaneously substituting every free occurrence of variable xi by the constant a i , for
1 i n. In the same way, if t is a term over the signature  A whose variables occur
in the sequence x, then t[ x /
a ] is the closed term obtained from t by simultaneously
substituting every occurrence of variable xi by the constant a i , for 1 i n; in this
case, we say that x is a context for t. The set of all the terms of the signature  A of
L(A) with context x will be denoted by T (A)x . 

Remark 7.3.4 In the case n = 1, the notation introduced in Definition 7.3.3 is dif-
ferent to that introduced in Sect. 7.1 concerning substitutions of terms for variables.
In fact, according to the latter, [x/a]
denotes the substitution of constant a for vari-
able x. But, according to Definition 7.3.3, [x/a] (when considering x as a context)
denote the same formula. As it will be convenient to identify (informally) an element
b of A with the constant b of  A , this duality is not problematic. (Notice that this
duality already appears in the following definition.)

Definition 7.3.5 (QmbC-valuations)

  Let A be a structure over  with domain A. A
mapping v : SL(A) 0, 1 is a QmbC-valuation over A if it satisfies the following
clauses:
304 7 First-Order LFIs

 
(vPred) v(P(t1 , . . . , tn )) = 1  t1A , . . . , tnA  IA (P), for P(t1 , . . . , tn )
At A
(vOr) v( ) = 1 v() = 1 or v() = 1
(vAnd) v( ) = 1 v() = 1 and v() = 1
(vImp) v( ) = 1 v() = 0 or v() = 1
(vNeg) v() = 0 = v() = 1
(vCon) v() = 1 = v() = 0 or v() = 0
(vVar) v() = v() whenever is a variant of
(vEx) v(x) = 1 v([x/a]) = 1 for some a A
(vUni) v(x) = 1 v([x/a]) = 1 for every a A
(sNeg) For every contexts ( x ; z) and ( x ; y), for every sequence (  in A inter-
a ; b)
preting (x ; y), for every L(A)x;z and every t T (A)x;y such that t is free for z
 A 
in , if [z/t] L(A)x;y and b = (t[ x ; y/a ; b]) then:
v(([z/t])[ x ; y/ 
a ; b]) = v([ x ; z/
a ; b]) =
v(([z/t])[ x ; y/  = v([
a ; b]) x ; z/
a ; b])
(sCon) For every contexts ( x ; z) and ( x ; y), for every sequence (  in A inter-
a ; b)
preting (x ; y), for every L(A)x;z and every t T (A)x;y such that t is free for z
 A 
in , if [z/t] L(A)x;y and b = (t[ x ; y/a ; b]) then:
v(([z/t])[ x ; y/  = v([
a ; b]) x ; z/
a ; b]) =
v(([z/t])[ x ; y/  = v([
a ; b]) x ; z/
a ; b]).


Notice that, in particular,

v(P(a 1 , . . . , a n )) = 1  a1 , . . . , an  P A

for every n-ary predicate symbol P and every a1 , . . . , an A. On the other hand,

v() = 1 v() = 0

for every formula .

Definition 7.3.6 (Interpretations) An interpretation for QmbC (over the signature


) is a pair A, v, such that A is a structure over  and v : SL(A) 0, 1 is a
QmbC-valuation. 

By using the previous definition, the notion of satisfiability, as well as the seman-
tical consequence relation, are both defined for sentences of the extended language.

Definition 7.3.7 (Semantical consequence relation) An interpretation A, v for

QmbC over  satisfies a sentence SL(A) , which is denoted by A, v  , if
v() = 1. If  SL(A) , then A, v   means that A, v  for every . Finally,
given  {} SL(A) , is said to be a semantical consequence of , denoted by
  , if A, v   implies that A, v  , for every interpretation A, v. 
7.3 Tarskian Paraconsistent Structures 305

Remark 7.3.8 Clauses (sNeg) and (sCon) are of a purely technical character. They
establish that if two related formulas in the diagram language involving substitutions
get the same truth-value, then this equality must be preserved through the non-truth-
functional connectives and . For instance, let P be a symbol for a unary predicate
and f a symbol for a unary function. Let x be the empty context (and so a is also
empty); y = x (a single variable), b = a (a single element of A), = P(z) (where
 
z is a variable) and t = f (x). Let b = (t[x/a])A = f (a) A = f A (a). Then,

v((P(z)[z/t])[x/a]) = v(P(t)[x/a]) = v(P( f (x))[x/a]) = v(P( f (a)))

while  
= v P f A (a)
v(P(z)[z/b]) = v(P(b)) .
 
By (vPred), v(P( f (a)))
and v P f A (a) coincide. However, only clauses
 
(sNeg) and (sCon) can guarantee that v(# P( f (a))) = v # P f A (a) for #
{, }, as expected. This feature will be fundamental in order to prove the Substitu-
tion Lemma (Theorem 7.3.14) which, in turn, is crucial in the proof of the soundness
of QmbC with respect to interpretations.
It is important to note that in the absence of (sNeg) and (sCon), it is possible to
find interpretations falsifying axioms (Ax11) and (Ax12) (see Remark 7.4.2).
Corollary 7.3.9 The logic QmbC is a (strong) LFI.
Proof In order to prove items (i.a), (i.b) and (i.c) of Definition 2.1.10, the proof of
Theorem 2.3.2 can be easily adapted, by taking as p and q two different atomic closed
formulas. With respect to item (ii) of such definition, it follows from clause (vCon)
of Definition 7.3.5. 
The extension of the semantical notions introduced above to general formulas
(that is, to formulas containing free variables) can be done by appealing to some
concepts from Definition 7.3.3.
Definition 7.3.10 (Extended valuation) Let A be a structure over  ,  x a context
and a a sequence of elements in A interpreting x. If v : SL(A) 0, 1 is a QmbC-
 
valuation over A, its extension vax : L(A)x 0, 1 is defined as follows: vax () =
v([x /
a ]), for every L(A)x . 
Remark 7.3.11

(1) Clearly, if L(A)x and y = ( x ; z ) with z = z 1 , . . . , z m then vay;b () = vax ()
for every sequence b = b1 , . . . , bm in A interpreting z . In particular, v() = vax ()
for every x and a , whenever SL(A) .
(2) The clauses for QmbC-valuations (see Definition 7.3.5) can be reintroduced
in terms of extended valuations. The clauses for connectives and quantifiers are
essentially the same: it is enough to carry on the context x and the sequence a
interpreting it. For instance, clause (vUni) changes to
306 7 First-Order LFIs

(vUni) vax (x) = 1 vax ([x/a]) = 1 for every a A,

observing that whether x occurs or not in x is irrelevant. In order to see this, note that
vax (x) = v((x)[ x /
a ]) = v((y([x/y]))[ x /
a ]), where y is a variable that does
not occur either in x or in , by clause (vVar) and the definition of substitution. But
(y([x/y]))[ x /
a ] = y([x/y][ a ]), and so vax (x) = v(y([x/y][
x / x /
a ])) =
1 iff v([x/y][ x /
a ][y/a]) = v([x/a][ x /a ]) = 1, for every a A. But the latter is
equivalent to saying that vax ([x/a]) = 1, for every a A.
On the other hand, the two clauses concerning substitution can be presented in
a simplified way. Thus, under the same notation and assumptions as for (sNeg) and
(sCon), the corresponding clauses for extended valuations are the following:
;b ;b
(sNeg) vax; a ;b
y ([z/t]) = vx;z () = vax; a ;b
y ([z/t]) = vx;z ()
 
(sCon) vax;
;b a ;b
y ([z/t]) = vx;z () = vax;
;b a ;b
y ([z/t]) = vx;z ().

Definition 7.3.12 (Extended semantical consequence relation) We say that an inter-

pretation A, v satisfies a formula L(A)x , denoted by A, v x , if vax () = 1
for every sequence a in A interpreting x. If  L(A)x , then A, v x  means that
A, v x , for every . Finally, given  {} L(A)x , we say that is a
semantical consequence of  in context x, denoted by  x , if A, v x  implies
that A, v x , for every interpretation A, v. 

Remark 7.3.13 Observe that the notions x and  coincide, as long as sentences of
L(A) are involved. Moreover,

 x ()  ()

where () = x1 xn and () = {() : }.

This section ends with the proof of a technical result which is fundamental in
order to state the soundness of QmbC with respect to the proposed semantics. In
order to lighten notation, and without loss of generality, extended valuations will be
used.

Theorem 7.3.14 (Substitution Lemma) Let t be a term free for the variable z in the
formula . Suppose that (x ; z) and (
x ; y) are contexts for and [z/t], respectively.
 A 
Let A, v be an interpretation for QmbC . If b = (t[ x ; y/
a ; b]) then:


vax;
;b a ;b
y ([z/t]) = vx;z ().

Proof The proof is identical with that for classical logic, by induction on the com-
plexity of L(A)x;z .
(a) = P(t1 , . . . , tk ), with P a symbol for predicate and t1 , . . . , tk terms in T (A)x;z .
Then, [z/t] = P(t1 [z/t], . . . , tk [z/t]). By the definition of extended valuation it
follows that
  
vax;
;b
y
([z/t]) = 1 iff  ((t1 [z/t])[
x ; y/  A, . . . , ((tk [z/t])[
a ; b]) x ; y/  A  IA(P).
a ; b])
7.3 Tarskian Paraconsistent Structures 307

By induction on the complexity of the term u T (A)x;z , it is easy to prove that

 
((u[z/t])[
x ; y/  A
a ; b]) = (u[ a ; b])A
x ; z/

 A 
for b = (t[
x ; y/
a ; b]) . From this,
 
((t1 [z/t])[
x ; y/  A
a ; b]) , . . . , ((tk [z/t])[
x ; y/  A
a ; b])  IA (P)

if and only if
 
(t1 [ a ; b])A , . . . , (tk [
x ; z/ a ; b])A  IA (P).
x ; z/

As
 
 (t1 [ a ; b])A , . . . , (tk [
x ; z/ a ; b])A  IA (P)
x ; z/ iff vax;z
;b
() = 1

it follows that

vax;
;b
y ([z/t]) = 1 iff vax;z
;b
() = 1.


That is, vax; ;b a ;b
y ([z/t]) = vx;z ().
(b) = (#), with # {, , }. Assuming that and satisfy the property (by
induction hypothesis), then also satisfies the property, as v is truth-functional for
these connectives.
(c) = x. If z does not occur free in , the result is obviously true. If z
occurs free in then, as t is free for z in , it follows that x does not occur in
t. Thus, [z/t] = (x)[z/t] = x([z/t]). By definition of extended valuation
and Remark 7.3.11(2), if y is a variable that does not occur in either x; y; z or
;b a ;b  = v(y([z/t]
, vax;y ([z/t]) = vx;y (x([z/t])) = v((x([z/t]))[ x ; y/
a ; b])
[x/y][ x ; y/ 
a ; b])). Then, by (vUni) and the equations above,


vax;
;b  =1
y ([z/t]) = v(y([z/t][x/y][
x ; y/
a ; b]))

if and only if

x ; y/
v([z/t][x/y][ 
a ; b][y/a]) = 1 for every a A.

But [z/t][x/y][ x ; y/ 

a ; b][y/a] = ([x/y])[z/t][ x ; y; y/  a], as we are sub-
a ; b;
stituting individual constants for variables, and so the simultaneous substitution coin-
;b
cides with the iterative substitution. Thus, vax; y ([z/t]) = 1 iff v(([x/y])[z/t]

[
x ; y; y/  a]) = v
a ; b; 
a ; b;a
(([x/y])[z/t]) = 1, for every a A.
x;y ;y
 A 
By the induction hypothesis applied to [x/y], and since b = (t[
x ; y/
a ; b]) =

 a]) (as y is new),
A
(t[
x ; y; y/
a ; b;


vax;
;b;a
y ;y (([x/y])[z/t]) = 1 iff vax;z;y
;b;a
([x/y]) = 1.
308 7 First-Order LFIs

On the other hand,

vax;z;y
;b;a
([x/y]) = 1 for every a iff vax;z
;b
(y[x/y]) = vax;z
;b
(x) = vax;z
;b
() = 1

and so

vax;
;b
y ([z/t]) = 1 iff vax;z
;b
() = 1.


That is, vax;
;b a ;b
y ([z/t]) = vx;z ().
(d) = x. This is a consequence of the fact that v(x) = v(x).
(e) = #, with # {, }. By induction hypothesis,


vax;
;b a ;b
y ([z/t]) = vx;z ()

and then, by clauses (vNeg) and (vCon) ,

 
vax;
;b a ;b a ;b a ;b
y ([z/t]) = vx;y (#[z/t]) = vx;z (#) = vx;z ().

The importance of clauses (vNeg) and (vCon) is clear, therefore, from the proof
of the above theorem.

7.4 Soundness Theorem for QmbC

The next step is to prove that the logic QmbC is sound and complete with respect
to its semantics of Tarskian paraconsistent interpretations. For simplicity, this will
be proved just for sentences (i.e., for formulas without free variables). That is, if
 {} is a set of sentences, then

 QmbC   .

It should be observed that, despite the fact that the premises (the set ) and the
conclusion (the formula ) are sentences, a given derivation of from  can involve
formulas with free variables, and so the use of extended valuations will be required.
On the other hand, there is no loss of generality by proving soundness and complete-
ness just for sentences, by virtue of Remark 7.3.13 (which obviously also holds in
the Hilbert calculus QmbC ).
In this section, the soundness of QmbC will be proved, namely:

Theorem 7.4.1 (Soundness of QmbC with respect to interpretations) For every set
of sentences  {}: if  QmbC then   .
7.4 Soundness Theorem for QmbC 309

Proof By induction on the length n of a derivation 1 , . . . , n of from  in QmbC,

we will prove that given a structure A, each QmbC valuation v over A, such that
A, v  , satisfies the following: vax (i ) = 1 for every sequence a in A interpreting
x and every i i n, where x is a context for every i (1 i n). In particular,
it will be proved that v() = 1, as desired.
It is clear that, in order to get the desired result, it is enough to prove the following:
(i) vax () = 1 for every a and every instance of an axiom scheme of QmbC
(ii) if vax (1 ) = 1 and vax (1 2 ) = 1 for every a then vax (2 ) = 1 for every a
(iii) if vax;y
;b
(1 2 ) = 1 for every ( a ; b), and if the variable y does not occur free
a
in 1 , then vx (1 y2 ) = 1 for every a
(iv) if vax;y
;b
(1 2 ) = 1 for every ( a ; b), and if the variable y does not occur free
in 1 , then vax (y1 2 ) = 1 for every a .
For proving (i), it is sufficient to analyze the axioms involving quantifiers; the others
are true because of the soundness theorem of mbC for valuations (see Theorem 2.2.2).
The same holds for item (ii) (concerning MP), which is obviously true. Thus con-
sider the following cases for item (i):
(i.1) = z [z/t] where t is free for z in . Let x be a context formed by all
the variables occurring free in z and let ( x ; y) be a context formed by the vari-
ables occurring free in [z/t]. Consider a sequence (  in A interpreting (
a ; b) x ; y).
a ;b a ;b ;b
If vx;y (z) = 0 then vx;y () = 1. If, on the other hand, vax;y (z) = v a
x (z) =1
then vax;z
;b
() = 1 for every b A (by the considerations above concerning simulta-
 A 
neous and iterated substitutions). In particular, va;b () = 1 for b = (t[
x;z x ; y/
a ; b]) .

By the Substitution Lemma (Theorem 7.3.14), vax;
;b a ;b
y ([z/t]) = vx;z (), as t is free for

z in . From this vax;;b
y ([z/t]) = 1, as required.
(i.2) = [z/t] z where t is free for z in . The proof is analogous to that of
item (i.1).
(i.3) = , where is a variant of . This is an obvious consequence of clause
(vVar) and the fact that [ x /
a ] is a variant of [x /
a ] whenever is a variant of .
Now, in order to prove (iii), suppose that vax;y ;b
(1 2 ) = 1 for every ( a ; b), where
a
the variable y does not occur free in 1 . Fix the sequence a . If vx (1 ) = 0 then
vax (1 y2 ) = 1. On the other hand, if vax (1 ) = vax;y;b
(1 ) = 1 then, by hypoth-
a ;b
esis, vx;y (2 ) = vx (2 [y/b]) = 1, for every b A. From this, vax (y2 ) = 1.
a

Item (iv) is proved in a similar way. 

Remark 7.4.2 As it was already observed in Remark 7.3.8, clauses (sNeg) and (sCon)
are fundamental in order to prove the soundness theorem for QmbC .
Consider, for instance, = P(z) and t = f (x, y), with P a symbol denoting a
unary predicate. Suppose that va;b
x;y (z) = v(zP(z)) = 1. Then,

= 1 for every e A.
v(P(e)) (1)
310 7 First-Order LFIs

In particular,  
v P f A (a, b) = 1. (2)

On the other hand,

va;b a;b
x;y ([z/t]) = vx;y (P( f (x, y))) = v(P( f (a,
b))). (3)

In order to guarantee that v(P( f (a, = 1, we need to ensure that

b)))
 
v P f A (a, b) = v(P( f (a,
b))). ()

But the latter is only obtained from the Substitution Lemma or, in this specific case, by
clause (sNeg). In other words, without (sNeg) it would be possible to find a valuation
v over a structure A such that va;b a;b
x;y (zP(z)) = 1 but vx;y (P( f (x, y))) = 0. That
is, it would be possible to falsify the instance

z P(z) P( f (x, y))

of axiom schema (Ax12). By a similar argument, it would be possible to falsify the

instance
z P(z) P( f (x, y))

of axiom schema (Ax12) without the presence of clause (sCon).

7.5 Completeness Theorem for QmbC

Given a first-order signature , any set of sentences in L  will be called a theory.

This section is devoted to prove the completeness of QmbC with respect to
interpretations. The proof will be analogous to that for classical logic: given a theory 
which does not deduce a given sentence (being, therefore, non-trivial), a canonical
interpretation will be constructed which satisfies  but does not satisfy , proving
the following:
 QmbC =   .

In order to do this, the original theory  will be conservatively extended to a

Henkin theory  in an extended signature, that is, to a theory containing a witness
for each existential sentence. Since  is a conservative extension of , it does not
derive . Thus, by using a classical and general result by Lindenbaum-os,  will
be extended to a maximal theory  which does not derive and is still a Henkin
theory. Using a canonical structure generated from , the characteristic map of

 will constitute a QmbC-valuation which, as required, satisfies  but does not
satisfy .
7.5 Completeness Theorem for QmbC 311

7.5.1 Henkin Theories

A Henkin theory is a theory designed to comply with the inference rules for quan-
tifiers. Specifically, a Henkin theory has associated to it a set of constants called
witnesses) which allows every existential sentence to be satisfied.
Definition 7.5.1 (Henkin theory) Consider a theory  SL and a nonempty set C
of constants of the signature  of L. Then,  is called a C-Henkin theory in QmbC
if it satisfies the following: for every sentence of the form x in SL , there exists a
constant c in C such that if   x then   [x/c]. 
The set C is called a set of witnesses of . The next step is to prove that any
theory can be conservatively extended to a C-Henkin theory, for some C.
Theorem 7.5.2 (Theorem of Constants) Let  SL be a theory in QmbC over
a signature , and let CQmbC be the consequence relation of QmbC over the
signature C , which is obtained from  by adding a set C of new individual constants.
Then, for every SL ,

 QmbC iff  CQmbC .

That is, QmbC (over C ) is a conservative extension of QmbC (over ).

Proof The proof is analogous to that for classical first-order logic: given a derivation
of from  in QmbC over C , the constants of C occurring in are replaced
uniformly by new variables, obtaining a finite sequence  of formulas over . But
the instances over C of axioms of QmbC occurring in become instances over
 of axioms of QmbC, and the same holds for the instances of inference rules.
Thus  is in fact a derivation of from  in QmbC over . The converse is
obvious. 
Theorem 7.5.3 Every theory  SL in QmbC over a signature  can be conser-
vatively extended to a C-Henkin theory  H in QmbC over a signature C , as in
Theorem 7.5.2. That is,   H and if SL then  QmbC iff  H CQmbC .
Additionally, any extension of  H by sentences in the signature C is also a
C-Henkin theory.
Proof We begin by defining an increasing denumerable sequence of signatures
0 1 . . . such that each n+1 is obtained from n by adding new individ-
ual constants. The language L n generated by n will be denoted by L n , and so
L n L n+1 .
The definition of the signatures is as follows:
(i) 0 = ; then, L 0 = L 0 = L  .
(ii) 1 is obtained from 0 by adding the set of new individual constants
 
W1 = cx : x is a sentence of L 0 ;
312 7 First-Order LFIs

(iii) For n 1, n+1 is obtained from n by adding the set of new individual
constants
 
Wn+1 = cx : x is a sentence of L n L n1 .

Let C = n1 Wn be the set of new individual constants, let C = n0 n be the

signature obtained by adding the new constants, and let L C = L C .
Consider now the following sequence of sets of non-logical axioms over C :

AX 0 =
 
AX n+1 = x [x/cx ] : x SL n (for n 0).

Finally, let  H =  n1 AX n . Observe that  H SL C , and that it extends .

We will prove now that  H is a conservative extension of . Thus, let SL such
that  H CQmbC , and let be a derivation of from  H in QmbC over C . As
is finite, there exists a finite set 0H  H such that , 0H CQmbC . Let x
[x/cx ] in 0H , and let 1H = 0H {x [x/cx ]}. Given that 0H is a
set of sentences, we can apply (DMT) in order to obtain , 1H CQmbC (x
[x/cx ]) . Observe that the constant cx only appears in the conclusion
and so, by using the same technique employed in the proof of Theorem 7.5.2,
we can substitute that constant by a new variable, namely y. This means that
, 1H CQmbC (x [x/y]) . By Lemma 7.2.10(3), , 1H CQmbC
(x y([x/y])) . On the other hand, CQmbC x y([x/y]):
indeed, since = [x/y][y/x] then CQmbC y([x/y]), by axiom (Ax11),
and so CQmbC x y([x/y]), by rule (-In). From this, , 1H CQmbC .
By repeating this process, every element of 1H is eliminated in a finite number of
steps, proving that  CQmbC . By Theorem 7.5.2, we finally obtain that  QmbC .
This shows that  H is in fact a conservative extension of .

To finish the proof, consider an extension  H of  H (in particular, we can take

 H =  H ), formed by sentences of L C . Suppose that for some sentence x
  
L C , it holds that  H CQmbC x. As  H extends  H , then  H CQmbC x
 
[x/cx ] and so  H CQmbC [x/cx ]. This shows that  H (and, in particular,
 H ) is a C-Henkin theory in QmbC over C . 

7.5.2 Canonical Interpretations

As in the propositional case, the proof of completeness requires the use of maxi-
mal theories with respect to a sentence. For that reason the Lindenbaum and os
Theorem 2.2.6 will be used, restricted to sentences of QmbC .
Recall from Chap. 2 the notion of Tarskian logic (Definition 2.1.4). Given that
the logic QmbC is defined by means of a Hilbert calculus where the inference
rules are finitary, it is Tarskian and finitary. Therefore, the Lindenbaum-os Lemma
7.5 Completeness Theorem for QmbC 313

(Theorem 2.2.6) holds for it. In particular, it holds for QmbC restricted to sentences:
it is easy to see that the consequence relation of QmbC, when restricted to sentences,
is Tarskian and finitary. Then, if we also restrict the set L of Definition 2.2.3 to
sentences (that is, to SL ), the following holds:

Corollary 7.5.4 Let  {} SL be a set of sentences such that  QmbC . Then,

there exists a set of sentences  SL extending  which is maximally non-trivial
with respect to in QmbC (by restricting QmbC to sentences).

Theorem 7.5.5 (Canonical interpretation) Let  SL be a set of sentences over

a signature  containing at least one individual constant. Assume that  is a C-
Henkin theory in QmbC for a nonempty set C of individual constants of , and
that  is also maximally non-trivial with respect to in QmbC, for some sentence
. Then  induces
  a canonical structure A and a canonical QmbC-valuation
v : SL(A) 0, 1 over A such that, for every sentence SL :

A, v   QmbC .

Proof Let A = C T be the set of closed terms (that is, without variables) over the
signature . We define now a structure A = A, IA  over  as follows: IA (c) = c, if
c is an individual constant; if f is a function symbol, then IA ( f ) : An A is such
that IA ( f )(t1 , . . . , tn ) = f (t1 , . . . , tn ) (and so t A = t for every t C T ). Finally,
define the interpretation of the predicate symbols as follows:

t1 , . . . , tn  IA (P)  QmbC P(t1 , . . . , tn ).

Recall from Definition 7.3.2 the diagram language L(A) of A, its set T (A) of terms
and the extended structure  A such that IA  (t) = t. Here, t is a new constant intro-
duced for each closed term t C T (which is as an element of the domain of D).
Let C T (A) be the set of closed terms of the language  L(A),
 and consider a mapping
: C T (A) C T recursively defined as follows: t = t if t C T ; c = c if c
is an individual constant of ; and ( f (t1 , . . . , tn )) = f (t1 , . . . , tn ) if f is a func-

tion symbol of . It is clear that t = t A for every t C T (A). This mapping can
be naturally extended to sentences: let : SL(A) SL be recursively defined as fol-
lows: (P(t1 , . . . , tn )) = P(t1 , . . . , tn ) if P(t1 , . . . , tn ) is atomic; (#) = #( )
if # {, }; (#) = ( # ) if # {, , }; and (Qx ) = Qx( ) if
Q {, }. Clearly, is the sentence of L  obtained from by substituting every
occurrence of a constant t by the term t itself.  
Finally, we define the mapping v : SL(A) 0, 1 as follows:

v() = 1  QmbC .

By construction of v, it is clear that for every sentence SL :

A, v   QmbC .
314 7 First-Order LFIs

The proof will be completed by showing that v is in fact a QmbC-valuation (recall

Definition 7.3.5).
In order to prove (vPred), if P(t1 , . . . , tn ) is an atomic sentence of L(A), then

v(P(t1 , . . . , tn )) = 1  QmbC P(t1 , . . . , tn ).


But this happens iff t1 , . . . , tn  IA (P), by definition of IA (P). Given that t = t A
for every t C T (A), then
 
v(P(t1 , . . . , tn )) = 1 t1A , . . . , tnA  IA (P).

In order to see that v satisfies clauses (vOr), (vAnd), (vImp), (vNeg) and (vCon), the
proof is analogous to that for mbC (see Theorem 2.2.7), and using the definition
of .
The satisfaction of clause (vVar) follows from axiom (Ax13) and the definition of .
In order to prove (vEx), firstly observe that, if x SL then  QmbC x implies
that  QmbC [x/c] for some constant c of C (which is an element of C T ),
as  is a C-Henkin theory in QmbC . On the other hand, if  QmbC [x/t]
for some closed term t in C T then  QmbC x, in virtue of (Ax11) and
MP. Consider now a sentence in L(A) of the form x. Then v(x) = 1 iff
 QmbC (x) iff  QmbC x() , by definition of . From this, and by the
observation above, one infers that v(x) = 1 iff  QmbC (() )[x/t] for some
closed term t in C T . On the other hand, it is easy to prove by induction on the com-
plexity of that (() )[x/t] = ([x/t ]) , for every t C T . Thus, v(x) = 1 iff
 QmbC ([x/t]) for some t of C T . From this it follows that v(x) = 1 iff
v([x/t ]) = 1 for some element t of C T .

Concerning (vUni), as v satisfies the clauses for the propositional connectives then
v() = 1 iff v() = 0. On the other hand, v(x) = v(x), because of the
theorems QmbC x x and QmbC x x of QmbC (and
by the Soundness Theorem). From this, and using clause (vEx), it can be immediately
seen that v satisfies clause (vUni).

Finally, we will prove that the pair A, v satisfies the Substitution Lemma (see
Theorem 7.3.14) and so the mapping v satisfies the clauses (sNeg) and (sCon).
Facts: Let t be a term free for a variable z in a formula . Suppose that ( x ; z) and
(
x ; y) are contexts for and [z/t], respectively, and let b = (t[ x ; y/  .
a ; b])
Then:
(i) ((u[z/t])[x ; y/  = (u[
a ; b]) a ; b]) , for every term u T (A)x;z .
x ; z/
(ii) (([z/t])[x ; y/  = ([
a ; b])
a ; b]) .
x ; z/
Item (i) can be easily proved by induction on the complexity of u. (Notice that this

fact was already used in item (a) of the proof of Theorem 7.3.14, given that u A = u
for every term u).
7.5 Completeness Theorem for QmbC 315

Item (ii) is proved by induction on the complexity of . If is atomic, the result

follows immediately by item (i). The propagation of the induction hypothesis through
the connectives , , , and is obvious. The propagation of the induction
hypothesis through the quantifiers is a consequence of the fact that t is free for
z in . Therefore, x does not occur in t when = Qx, with Q {, }. From
this, ((Qx)[z/t])[ x ; y/  = Qx(([z/t])[
a ; b] x ; y/  and the result follows
a ; b]),
by induction hypothesis and the definition of . This concludes the proof of the
Facts.
Now, let be a formula of L(A), and t a term free for the variable z in such
x ; z) and (
that ( x ; y) are contexts for and [z/t], respectively. Then

;b 
vax;y ([z/t]) = v(([z/t])[
x ; y/
a ; b])

and so

vax;
;b  .
y ([z/t]) = 1  QmbC (([z/t])[
x ; y/
a ; b])

On the other hand, vax;z

;b
() = v([
x ; z/
a ; b]) therefore

vax;z
;b
() = 1  QmbC ([ a ; b]) .
x ; z/

Finally, by taking b = (t[

x ; y/  , it follows by Facts(ii) that
a ; b])


vax;
;b a ;b
y ([z/t]) = vx;z ()

as desired. From this, the mapping v satisfies clauses (sNeg) and (sCon).
This proves that the pair A, v is an interpretation with the required
properties. 

Theorem 7.5.6 (Completeness of QmbC with respect to interpretations) For every

set of sentences  {} over a signature , if   then  QmbC .

Proof Suppose that  {} SL such that  QmbC . By Theorem 7.5.3, there

exists a C-Henkin theory  H defined over a signature C which conservatively
extends , that is: for every sentence SL ,  QmbC iff  H CQmbC , recall-
ing that CQmbC denotes the consequence relation of the Hilbert calculus QmbC
over the signature C . By Corollary 7.5.4, there exists a set of sentences  H over
the signature C which extends  and is maximally non-trivial with respect to in
QmbC (as a calculus defined over C ). By Theorem 7.5.3,  H is also a C-Henkin
theory. By Theorem 7.5.5, we can define a canonical interpretation A, v over C
such that, for every sentence over C ,

A, v   H CQmbC .
316 7 First-Order LFIs

In particular, A, v   (as   H ) and A, v  (as  H CQmbC ). Finally, let A be

the reduct of A to the signature .2 That is, IA coincides with IA over  (and so just
forgets the interpretation of the individual constants in C). Let v be the restriction
of v to the set of sentences SL( A ) of the diagram language of A. Clearly, A, v 
iff A, v  , for every sentence SL( A ) . Therefore, A, v  is an interpretation
for QmbC over  such that A, v   but A, v  . This shows that   , as
required. 

7.6 Compactness and Lowenhim-Skolem Theorems

for QmbC

In this section, some fundamental theorems which form the starting point for classical
Model Theory will be also proved for QmbC, namely: the Compactness and the
Lowenhim-Skolem Theorems. The Lowenhim-Skolem Theorems state, as in the
familiar classical first-order case, that a first-order theory with an infinite model has
infinite models of every infinite cardinality. The Compactness Theorem for QmbC
states, as in the classical case, that a set of first-order sentences has a model if and only
if every finite subset of it has a model. This theorem, an important tool in classical
model theory, is also important here as it provides a useful method for constructing
models of any set of sentences that are non-trivial.
Firstly, some basic definitions and results will be stated.
Definition 7.6.1 Let  be a first-order signature for LFIs. Consider the sets
C = {c : c is an individual constant of }
F = { f : f is a function symbol of arity n of , for some n 1}
P = {P : P is a predicate symbol of arity n of , for some n 1}.
The cardinal of , denoted by ||||, is the cardinal of the set

CF P

where denotes the set of natural numbers. 

Definition 7.6.2 Let A = A, IA  be a structure. The cardinal of A, denoted by
||A||, is the cardinal of the set A. Given an interpretation A, v, its cardinal is, by
definition, the cardinal of the structure A. 
Definition 7.6.3 Let  be a set of sentences over a signature . We say that  is
satisfiable if there exists an interpretation A, v such that A, v   (recall Defini-
tion 7.3.7). Otherwise,  is said to be unsatisfiable. If A, v   then A, v is called
a model of . 

2 As usual, if  is a subsignature of another signature  , then any structure A over  can be seen
as a structure over , by forgetting the interpretation of the symbols in  that do not belong to
. Such a structure over  is called the reduct of A to  (see [13]).
7.6 Compactness and Lowenhim-Skolem Theorems for QmbC 317

By recalling that a strong negation can be defined in QmbC, it is immediate to

prove the following:

Proposition 7.6.4 Let  {} be a set of sentences over a signature . Then  

if and only if  {} is unsatisfiable.

Proof The proof is entirely analogous to that for classical logic, taking into account
the definition of the consequence relation  (see Definition 7.3.7) and the fact that
A, v  if and only if A, v  . 

Using the previous result, the following can be easily obtained:

Proposition 7.6.5 Let  be a signature. Then the following statements are equiva-
lent in QmbC :
(i) For every set of sentences  over  the following holds: every finite subset of 
is satisfiable if and only if  is satisfiable;
(ii) For every set of sentences  {} over  the following holds:   if and only
if 0  for some finite subset 0 of .

Proof (i) (ii): Assuming (i), suppose that   . Then  {} is unsatisfi-

able, by Proposition 7.6.4. By (i), there exists a finite subset 0 of  {} such
that 0 is unsatisfiable. Let 0 = 0 {}. Then 0 {} is also unsatisfiable
(since it contains 0 ). Using again Proposition 7.6.4, it follows that 0  , where
0 is a finite subset of . The converse is obvious.
(ii) (i): Assuming (i), suppose that  is a set of sentences over  which is unsat-
isfiable. Since clearly  is nonempty, there exists some sentence, say , belonging
to . Given that A, v  if and only if A, v  , for every interpretation A, v,
it follows that  {} is unsatisfiable and so   , by Proposition 7.6.4.
Using (ii) we infer that 0  for some finite subset 0 of . That being so, and
using Proposition 7.6.4 again, it follows that 0 {} is unsatisfiable. But then
0 {} is unsatisfiable, by the observation above, where 0 {} is a finite subset
of . The converse is immediate. 

From the last result, the compactness of QmbC can be stated as follows:

Definition 7.6.6 The logic QmbC is (semantically) compact if it satisfies state-

ment (i) (or, equivalently, statement (ii)) of Proposition 7.6.5. 

Theorem 7.6.7 (Compactness of QmbC) The logic QmbC is (semantically)

compact.

Proof It will now be proved that QmbC satisfies statement (ii) of Proposition 7.6.5.
Thus, suppose that  {} is a set of sentences such that   . By the Completeness
Theorem 7.5.6 it follows that  QmbC . Hence, since the syntactical consequence
relation QmbC is finitary, there is some finite subset 0 of  such that 0 QmbC
and so 0  , by the Soundness Theorem 7.4.1. The converse is immediate and so
QmbC satisfies statement (ii) of Proposition 7.6.5, being therefore compact. 
318 7 First-Order LFIs

Definition 7.6.8 Given a theory  over , that is, a set of sentences in L  ,  is said
to be non-trivial if   for some sentence over . 
Theorem 7.6.9 (Downward Lowenhim-Skolem Theorem for QmbC) Let  be a
signature. Every non-trivial theory  over  has a model of cardinal equal to ||||.

Proof Suppose that  is a non-trivial theory over . Then there is some sentence
over  such that   . By the proof of the Completeness Theorem 7.5.6, there
exists an interpretation A, v  for QmbC over  which is a model for , such that
the domain A of the structure A is the set of closed terms over the signature C . It
is not hard to prove that the cardinal of A is ||||. 

Lemma 7.6.10 Let  be a signature, and let  be a non-trivial theory over . If 

has a model of cardinal , then it has a model of cardinal  , for every cardinal 
greater or equal than .

Proof Let A, v be a model of  of cardinal such that A = A, IA . Let  be a

cardinal strictly greater than , and let A be a set of cardinality  such that A A .
Fix an element a of A, and define a structure A = A , IA  over  as follows. If
c is an individual constant of  then IA (c) = IA (c). If P is a predicate symbol
of  of arity n and a1 , . . . , an  (A )n then a1 , . . . , an  IA (P) if and only
if a1 , . . . , an  IA (P), where ai is ai , if ai A, or ai is a otherwise, for every
1 i n. If f is a function symbol of  of arity n and a1 , . . . , an  (A )n then
IA ( f )(a1 , . . . , an ) = IA ( f )(a1 , . . . , an ), where each ai is defined as above. Finally,
consider a QmbC-valuation v over A which extends v, such that v ([ a  ]) =
x /
  
v([ a ]) for every with context x, every a = a1 , . . . , an and every a = a1 , . . . , an
x /
such that each ai is defined from ai as above (it is easy to see that it is always possible
to define such a valuation from v). Then A , v  is a model of  of cardinal  . 
Theorem 7.6.11 (Upward Lowenhim-Skolem Theorem for QmbC) Let  be a
signature. Every non-trivial theory  over  has a model of cardinal , for every
cardinal greater or equal than ||||.

Proof It is a direct consequence of Theorem 7.6.9 and Lemma 7.6.10. 

7.7 QmbC with Equality

Despite being rich and versatile, first-order languages need to incoporate a (logical)
identity predicate in order to develop interesting applications such as paraconsistent
model theory or paraconsistent set theory. It is thus desirable that the quantified ver-
sion QmbC of mbC should be expanded with a binary predicate representing the
equality (or identity) relation, satisfying the usual axioms, which should be invari-
ably interpreted as the real equality relation. As such, the predicate will be viewed,
from a semantical point of view, as a logical symbol (akin to the connectives and
quantifiers), not belonging to the signatures.
7.7 QmbC with Equality 319

In formal terms, let  be a first-order signature for LFIs (recall Definition 7.1.1).
The set of formulas with equality over , denoted by L  , is the set of formulas
over the signature  obtained from  by adding a new symbol for a binary
predicate. As usual, (t t  ) will stand for the atomic formula (t, t  ) (where t and
t  are terms of the language). The set of sentences of L
 will be denoted by S . On the
other hand, the diagram language of A and the corresponding sets of sentences, when
including the equality symbol , will be denoted by L (A) and SL (A) respectively.
Despite considering the symbol , from the syntactical point of view, as a new
predicate symbol, it should be noticed that, form the semantical point of view, will
be always interpreted as the diagonal relation {(a, a) : a A} in every structure A
with domain A.
If is a formula and y is a variable free for the variable x in , then [x  y]
denotes any formula obtained from by replacing some, but not necessarily all, free
occurrences of x by y.

Definition 7.7.1 Let  be a first-order signature for LFIs. The logic QmbC (over
) is the extension of QmbC over L  obtained by adding to QmbC, besides all the
new instances of axioms and inference rules involving the equality predicate , the
following axiom schemas:

x(x x) (AxEq1)
 
xy (x y) ( [x  y]) , if y is a variable free for x in (AxEq2)

It is worth noting that axiom (AxEq2) depends on each and each specific
[x  y]. Since QmbC is an axiomatic extension of QmbC, it satisfies the deduction
meta-theorem DMT (see Theorem 7.2.5) as well as the simplified version of it (see
Corollary 7.2.6).
As in the case of classical logic, the broad scope of axiom (AxEq2) allows to
prove that is an equivalence relation which is congruential. In fact, let 
QmbC be
the consequence relation of the Hilbert calculus QmbC . Then:

Proposition 7.7.2 The following holds in QmbC :

(1)  QmbC xy((x y) (y x));
(2)  QmbC xyz((x y) ((y z) (x z))).
Thus, is an equivalence relation. Moreover, if (t t) and are formulas in
L x and yi is a variable free for xi in not occuring in the context x = x1 , . . . , xn
for every 1 i n, then
 n


QmbC (xi yi ) (t t[
x /y ])
i=1
320 7 First-Order LFIs

and
 n
  

QmbC (xi yi ) ( [
x /y ]) ([
x /y ] ) .
i=1

Proof It is similar to the proof for classical logic. For (1), consider the instance of
axiom (AxEq2) where (x) is (x x) and [x  y] is (y x). For (2), consider the
instance of axiom (AxEq2) in which x and y are substituted by y and z, respectively,
(y) is (x y) and [y  z] is (x z), and use (DMT). Finally, since is reflexive
(by axiom (AxEq1)), symmetric (by (1)) and transitive (by (2)), it is an equivalence
relation. The rest of the proof follows from axiom (AxEq2). 
Definition 7.7.3 The semantics for QmbC is given by interpretations A, v (recall
Definition 7.3.6) such that v : SL (A) 0, 1 , called a QmbC -valuation, is a
QmbC-valuation satisfying, additionally, the following clauses:
 
(vEq1) v(t1 t2 ) = 1 t1A = t2A , for every t1 , t2 C T (A) (the set of closed
terms of the language L(A))
(vEq2) v(a b) = 1 = v([x, y/a, = v(([x  y])[x, y/a,
b]) for every
b])
a, b A, if y is a variable free for x in . 
Since v(a a) = 1 for every a A, by (vEq1), then v(x(x x)) = 1, by (vUni).
However, it is possible to have v((t t)) = 1, that is, A, v  (t t), for
some interpretation A, v and some term t. In other words, we do not require that
v((t t)) = 1 be always the case.
From the clauses (vEq1) and (vEq2), it is clear that we can extend the Substitution
Lemma (Theorem 7.3.14) to QmbC , as it clearly holds for atomic formulas of the
form (t t  ) (and the proof is done by induction on the complexity of formulas).
Remark 7.7.4 At first sight, it would seem that the clause for valuations correspond-
ing to (AxEq1) should be simply v(x(x x)) = 1 or, equivalently, v(a a) =1
for every a A. However, in order to ensure the validity of the Substitution Lemma,
the stronger condition (vEq1) must be required. In fact, recall from Remark 7.4.2
that the validity of the Substitution Lemma is necessary to guarantee the soundness
of (Ax12). Consider again the terms t1 = f A (a, b) and t2 = f (a,
b) of Remark 7.4.2.
If we simply require for the QmbC -valuations the condition v(a a) = 1 for
 
every a A, there is no guarantee that v(t1 t2 ) = 1 despite t1A = t2A . This situa-
tion would violate the Substitution Lemma and so the Soundness theorem of QmbC
with respect to interpretations, as observed above.
It is easy to extend the previous results in order to prove the following soundness
and completeness theorem for QmbC . If we thus denote by  the semantical con-
sequence relation with respect to interpretations (see Definition 7.7.3), the following
holds:
Theorem 7.7.5 (Soundness and Completeness of QmbC with respect to interpre-

tations) For every set of sentences  {} S in a language with equality over a
signature :
7.7 QmbC with Equality 321

 
QmbC   .

Proof Soundness can be easily established from the considerations above. For com-
pleteness, we will adapt the proof of Theorem 7.5.6 as follows: assuming that
 {} SL is a set of sentences with equality over a signature  such that
 QmbC , let  be a set of sentences with equality over the signature C which
H

extends  and is maximal non-trivial with respect to in QmbC (as a calculus

defined over C ) and is also a C-Henkin theory. A canonical interpretation A, v
over C will be defined now such that, for every sentence over C ,

A, v   H C
QmbC .

Thus, define in the set C of constants the following relation: c  d iff  H C QmbC
(c d). Then  is an equivalence relation, by Proposition 7.7.2. Let  c = {d C :
c  d} for c C, and let A = { c : c C}. The structure A over C with domain
A is defined as follows:
(i) if c is an individual constant in C then IA (c) = d,  where d C is such that
 QmbC (c d);
H C

(ii) if f is a function symbol, then IA ( f ) : An A satisfies IA ( f )(

c1 , . . . ,
cn ) = 
c
where c C is such that  H QmbC C ( f (c1 , . . . , cn ) c);
(iii) if P is a predicate symbol, then

c1 , . . . ,
 cn  IA (P)  H QmbC C P(c1 , . . . , cn ).

The proof that IA is well-defined is similar to that for classical logic (see, for ins-
tance, [13]). Just to give an idea of how the proof works, consider clause (ii) defi-
ning the interpretation of function symbols. Thus, let f be a function symbol of arity
n 1, and let c1 , . . . , cn C. Let (x) be the formula ( f (c1 , . . . , cn ) x);
then C QmbC [x/ f (c1 , . . . , cn )], by axioms (Ax12), (AxEq1) and by MP.
Given that C QmbC [x/ f (c1 , . . . , cn )] x(x), by axiom (Ax11), it follows
that C
QmbC x(x). But 
H is a C-Henkin theory, hence  H C
QmbC [x/c]
for some c C. That is,  H C QmbC ( f (c1 , . . . , cn ) c), for some c C. By
Proposition 7.7.2, if d C is such that  H C QmbC ( f (c1 , . . . , cn ) d) then
c  d. Additionally, if d1 , . . . , dn C are such that ci  di for every 1 i n
then  H C QmbC ( f (c1 , . . . , cn ) f (d1 , . . . , dn )), whence  H C
QmbC
( f (d1 , . . . , dn ) c), using again Proposition 7.7.2. This shows that clause (ii) above
is well-defined.  
Now, let v : SL (A) 0, 1 be the mapping defined as follows:

v() = 1  H C
QmbC

where is the sentence over C obtained from by substituting every occurrence

of a constant 
c by the constant 
c. Thus, for every sentence over C ,
322 7 First-Order LFIs

A, v   H C
QmbC .

It is therefore proved that v is a QmbC -valuation over A. Finally, the reduct A, v  of
A, v to  is an interpretation for QmbC over  such that A, v   but A, v  ,
showing that   . 
We are now ready to adapt the definitions and results of the previous section to
the logic QmbC . Thus, the logic QmbC can be proved to be compact (by using a
notion of compactness similar to that of Definition 7.6.6), and the two Lowenhim-
Skolem Theorems hold for QmbC . From this, it is easy to prove the following:
Proposition 7.7.6 If a theory  of QmbC has arbitrarily large finite models, then
it has an infinite model.
Proof Given a theory  of QmbC over a signature  with arbitrarily large finite
models, consider a denumerable set C = {cn : n 0} of new individual constants.
Let C be the signature obtained from  by adding the set C of individual constants
and let  be the following theory over C :

 =  {(cn cm ) : n < m}.

Under the given hypothesis over , it is easy to prove that every finite subset of 
is satisfiable. By the Compactness Theorem for QmbC , the theory  has a model
A, v and so the domain A of A must be infinite. Let A , v  such that A is the
reduct of A to  and v is the corresponding restriction of v to SL(A ) . Since  SL
then A , v  is a model of  which is infinite. 

7.8 First-Order Characterization of Other Quantified LFIs

In the previous sections we have carefully studied QmbC, the first-order extension
of mbC, which constitutes the simplest propositional LFI defined in [14] and studied
here in Chap. 2. As seen in Chap. 3, there exist several propositional extensions of
mbC which can be considered, to which the concepts and techniques employed in
the previous sections could be readily applied in order to obtain the corresponding
first-order versions. For instance, recall from Chap. 3 the following extensions of
mbC:
(i) The logic mCi (recall Definition 3.5.12) is the extension of mbC obtained by
adding the following axiom schemas:

( ) (ci)
n+2
(for n 0)
n
(n )
7.8 First-Order Characterization of Other Quantified LFIs 323

(ii) The logic Ci is obtained from mCi by adding the axiom

(cf)

or, equivalently, by adding to mbC the axioms (ci) and (cf).

(iii) The system Cil is obtained from Ci by adding the axiom

( ) (cl)

(iv) The logic Cia is obtained by the addition of the following axiom schemas to
Ci:

( ) ( ) (ca )
( ) ( ) (ca )
( ) ( ) (ca )

(v) The logic Cila is obtained by the addition of the axiom schema (cl) to Cia or,
equivalently, of the axioms (ca ), (ca ) and (ca ) to Cil.
Observe that all the extensions of mbC presented above, which will be used in
Chap. 8 in order to define paraconsistent set theories based on LFIs, consist in the
addition of some axiom schemas. As proved in Chap. 3, the corresponding clauses
for the associated valuations can be straightforwardly associated to such axioms.
From this, the soundness and completeness theorem of mbC are extended to all
these propositional systems. Consequently, the first-order version of each of the
LFIs introduced above is straightforwardly defined by adding to QmbC the corre-
sponding axiom schemas from the list above. Thus, for instance, QCi is obtained
from QmbC by adding axiom schemas (ci) and (cf). Of course, the valuations of the
interpretation structures must satisfy the additional clauses for valuations required
in each case. Thus, a QCi-valuation over a structure A is a QmbC valuation (recall
Definition 7.3.5) satisfying, additionally, the following clauses:
(vCon) v() = 1 = v() = 1 and v() = 1
(vNeg) v() = 1 = v() = 1.

An interpretation for QCi is therefore, a pair A, v, such that A is a structure and v
is a QCi-valuation.
From this perspective, the proof of soundness and completeness theorems of
QmbC stated above can be easily extended to the new quantified LFIs. It is imme-
diate that (DMT) holds in all these systems, given that only axioms (and no rules)
were added to QmbC . It should be clear that all these systems can also be equipped
with an equality predicate , as it was done for QmbC in Sect. 7.7. The details of
these constructions are left to the reader.
324 7 First-Order LFIs

7.9 First-Order LFI1 and the Logic of Quasi-truth

Recall from Sect. 7.1 the basic concepts for first-order LFIs, in particular, the notions
of first-order signature  (Definition 7.1.1) as well as the first-order language L  for
LFIs generated form .
In this section, the quantified extension of LFI1 seen as the Hilbert calculus LFI1
over  (recall Definition 4.4.41) will be presented, under a different semantical per-
spective, and the corresponding theorems of soundness and completeness of LFI1
w.r.t. this semantics will be obtained. The same treatment will be given for QP1, a
quantified version of Settes 3-valued logic P1 described in Sect. 4.4.4.
The semantics to be analyzed in the following sections, which uses partial struc-
tures, is based on the theory of quasi-truth, introduced by Mikenberg et al. in [15]. The
notion of quasi-truth, also called pragmatic truth and partial truth, is a generalization
of Tarskis concept of truth in a structure. Under the perspective of quasi-truth, the
membership (or not) of a given tuple of the domain in such a relation is not always
defined, or is defined in an ambiguous way, and this ambiguity is represented by
partial structures. Within such structures, the relations are partial, in the following
sense. A partial relation R is a triple of sets R+ , R , Rb  such that R+ is the set
of tuples which effectively belong to R; R is the set of tuples which effectively
do not belong to R, and Rb is the set of tuples whose membership to R is (still)
undetermined.
By means of a model-theoretical approach based on such partial structures, the
conceptual framework of quasi-truth provides a way of accommodating the con-
ceptual incompleteness inherent in scientific theories. Bueno and da Costa defend
in [16], for instance, the view that if scientific theories are taken to be quasi-true,
and if the underlying logic is paraconsistent, it is perfectly rational for scientists and
mathematicians to entertain theories involving contradictions without triviality, and
that such a move provides a new way of thinking about the foundations of science.
Bohrs theory of the atom and classical electrodynamics, for instance, are well-known
examples discussed in the literature on scientific theories involving contradictions
(see a discussion about contradictions in science in Sect. 9.2). In [15] a rigorous for-
mal presentation of this idea was given, of which one of the most salient features
is the notion of partial structure. The formal aspects of the theory of quasi-truth
were generalized in [10], by means of a 3-valued model theory for a first-order LFI
called LPT1, based on the matrix logic MPT presented in Sect. 4.4.7, and which
coincides (up to language) with the quantified version of LFI1. Conceptual aspects
of the theory of quasi-truth will be additionally discussed in Sect. 9.3 of Chap. 9.

7.9.1 Semantics of Partial Structures

In this section the original notion of partial structures will be given, together with a
generalization of it based on the proposal in [10] for the logic LPT1, which produces
a semantical framework suitable to interpret the first-order extension of LFI1.
7.9 First-Order LFI1 and the Logic of Quasi-truth 325

Definition 7.9.1 (see [15]) Let  be a first-order signature.

(a) Let A be a nonempty set. A k-ary partial relation R defined over A is an ordered
triple R+ , R , Rb , where R+ , R , and Rb are pairwise disjoint sets, and R+
R Rb = Ak such that:
(i) R+ is the set of k-tuples which certainly belong to R;
(ii) R is the set of k-tuples which certainly do not belong to R;
(iii) Rb is the set of k-tuples whose membership to R is undetermined.
(b) A partial structure over the signature  is an ordered pair P = A, ()P  where
A is a nonempty set and ()P is a function defined over L  such that, for every
P P P
symbol R for a k-ary relation in , R P = R+ , R , Rb  is a k-ary partial relation.
In the case of symbols for functions and constants ()P is defined as usual (see
Definition 7.3.1). 
Concerning the clause (iii) of partial relations, there exist (at least) two interpreta-
tions: (1) to assume that it means lack of information (a truth-value gap) and so both
R(a1 , . . . , ak ) and R(a1 , . . . , ak ) are false for a Rb , thus arriving to a paracom-
plete logic; or (2) to assume that it means excess of information (a truth-value glut)
and so both R(a1 , . . . , ak ) and R(a1 , . . . , ak ) are true for a Rb , arriving thus to
a paraconsistent logic. In the present setting the second perspective is adopted, so
defining a 3-valued paraconsistent first-order logic.
Observe that, if Rb = , then R is a classical k-ary relation which can therefore
be identified with R+ .
The next step is to generalize the interpretation of predicates as triples to complex
formulas, in a way coherent with the truth-valued operators of LFI1. Moreover, the
presentation of LFI1 over the signature  (that is, by taking Halldns meaningful-
ness operator as primitive instead of ) called LFI1 (recall Definition 4.4.39) will
be regarded instead of LFI1. As proved in Theorem 4.4.45, LFI1 is the semantical
counterpart of LFI1 .
The following definition is adapted from [10] with suitable modifications:
Definition 7.9.2 Let  be a first-order signature.
(a) Let t be a term over , and let x = x1 , . . . , xk be a context for t (recall Defini-
tion 7.3.3), that is, a list of distinct variables containing all the variables occurring in
t. Let P be a partial structure over  and let a = a1 , . . . , ak be a list of elements of
the domain A of P such that ai interprets xi (for i = 1, . . . , k). The interpretation of
P
t in P, x and a , denoted by tx [ a ], is the element of A recursively defined as usual,
by adapting the definition given in Sect. 7.3 for closed terms.3
(b) Let be a formula of L  , x = x1 , . . . , xk be a context for , that is, a list of distinct
variables containing all the variables occurring free in , and let P be a partial struc-
 P
ture over . The LFI1  of  in Pand x, denoted by x , is the k-ary
-interpretation
P P P P
partial relation x = + , , b over A defined recursively as fol-
x x x
lows (from now on, the context x will be omitted, when there is no risk of confusion):

P
3 Namely, an additional clause must be added: if t is a variable xi then tx [
a ] is ai .
326 7 First-Order LFIs

(T1) (R(t1 , . . . , tm ))P = S+ , S , Sb  such that, for {+, , b},

def

 
P
a ], . . . , tmP [
S = a Ak : (t1 [ a ]) RP ;
 
P P P
(T2) ()P = , + , b ;
def

 
P P P
(T3) ()P = + , b , ;
def

 
P P P
(T4) ()P = b , + , ;
def

   
P P P P P P P P
(T5) ( )P = + + , , Ak + +
def
;
   
P def P P P P P P P P
(T6) ( ) = + + , , Ak + + ;
  
P P P P P
(T7) ( )P = + , + b , S
def

  
P P P P P
where S = Ak + + b ;
     
P def P P
(T8) (x)x = [x/y]+ , [x/y] ,S
x;y x;y
     
P P
where S = Ak [x/y]+ [x/y] and y is the first vari-
x;y x;y

 in x nor in ;  
able not occurring neither
 
P def P P
(T9) (x)x = [x/y]+ , [x/y] ,S
x;y x;y
     
P P
where S = Ak [x/y]+ [x/y] and y is the first vari-
x;y x;y
able not occurring neither in x nor in .
P
A partial structure P equipped with an interpretation mapping ()x as above, for
every context x, is called a Partial structure for LFI1 . 

In items (T8) and (T9), the following operators of quantification over relations where
used: if R Ak+1 then (R) Ak and (R) Ak are defined as
 
(R) = a Ak : (
a , b) R for every b A
 
(A) = a Ak : (
a , b) A for some b A .

When k = 0, that is, when is a sentence, three truth-values are obtained:

1 def
1 = {}, ,  ; 0 = , {},  ; and = , , {} .
def def

It is easy to see that, if x = x1 , . . . , xk is a context for the formula and y =

y1 , . . . , yr then
7.9 First-Order LFI1 and the Logic of Quasi-truth 327
   
P P P P
x;y = + Ar , Ar , b Ar .
x x x

P   P
In particular, ()x = , Ak , for every context x = x1 , . . . , xk , and so () = 0
for every P, where = ( ) for a fixed sentence . Thus, if = (
def def

) then  
P P P
(T10) ()P = , + b ,
When k = 0, quantification is given by

1 if R = A;
R, S, T  = 0 if R = A, S = ;
1
2
if R = A, S = .

1 if S = A, R = ;
R, S, T  = 0 if S = A;
1
2
if S = A, R = .

If is a sentence over a signature  and P is a partial structure over , we

P
will write P instead of , for denoting the LFI1 -interpretation of in P and
the empty context . Now, if the abbreviations 1, 0 and 21 introduced above for
the truth-values assigned to sentences in a partial structure P are identified with
the truth-values of the logical matrix MLFI1 for LFI1 (recall Definitions 4.4.38
and 4.4.39), the propositional operations over sentences (that is, for k = 0) introduced
in Definition 7.9.2 could be seen as defined over the domain of that matrix. The
following result shows that the operations defined over the propositional connectives
of the first-order language coincide with the operations of the matrix of LFI1 . As
usual, the symbol for the connectives and the symbol for the respective operators
interpreting them will be identified.

Proposition 7.9.3 Let P be a partial structure over a first-order signature , and

consider the matrix MLFI1 for LFI1 . Then, for every sentences and over ,
the following holds:
(1) (#)P = #( P ) for # {, }, where the operators are as in MLFI1 ;
(2) (#)P = ( P )#( P ) for # {, }, where the operators are as in MLFI1 .

Proof It follows as an easy consequence of the Definitions 4.4.38 and 7.9.2 (when
k = 0). 

The following definition is a variation of the notion of pragmatical satisfaction

introduced in [10] for the logic LPT1, and will constitute, as we shall see, another
perspective for the interpretation mapping provided in Definition 7.9.2.

Definition 7.9.4 Let be a formula over a first-order signature , x = x1 , . . . , xk

a context for it, P = A, ()P  a partial structure over  and a a sequence in A

interpreting x. We say that P satisfies in x and a w.r.t. LFI1 , denoted as P LFI1
x
[a ], in the following situations:
328 7 First-Order LFIs

 P P P P
(1) P LFI1
x R(t1 , . . . , tm )[ a ] iff (t1 [ a ], . . . , tm [ a ]) R+ Rb ,
whenever R(t1 , . . . , tm ) is atomic
 P P P P
(2) P LFI1
x R(t1 , . . . , tm )[ a ] iff (t1 [ a ], . . . , tm [ a ]) R Rb ,
whenever R(t1 , . . . , tm ) is atomic
 
(3) P LFI1
x [ a ] iff P LFI1 x [ a]
  
(4) P x LFI1
[ a ] iff P x LFI1
[ a ] and P LFI1 x [ a]
  
(5) P LFI1
x ( )[ a ] iff either P  LFI1
x [ a ] or P LFI1 x [ a]

LFI1 LFI1
(6) P LFI1
x ( )[ a ] iff P  x [ a ] and P  x [ a ]
  
(7) P LFI1
x ( )[ a ] iff P LFI1 x [ a ] and P LFI1 x [ a]
 
(8) P LFI1
x x[ a ] iff P  LFI1
x;y [x/y][ a ; b], for some b A, where y is
the first variable not ocurring neither in x nor in
 
(9) P LFI1
x x[ a ] iff P LFI1 x;y [x/y][ a ; b], for every b A, where y is
the first variable not ocurring neither in x nor in
  
(10) P LFI1
x [a ] iff either P LFI1 x [ a ] or P LFI1 x [ a]

LFI1 LFI1
(11) P LFI1
x ( )[ a ] iff P  x [ a ] and P  x [ a ]
  
(12) P LFI1
x ( )[ a ] iff either P LFI1 x [ a ] or P LFI1 x [ a]
 
LFI1
(13) P LFI1
x ( )[ a ] iff either P  LFI1
x [ a ] or P  x [ a]

LFI1
(14) P LFI1
x x[ a ] iff P  x;y [x/y][ a ; b], for every b A, where y is the
first variable not ocurring neither in x nor in
 
(15) P LFI1
x x[a ] iff P LFI1 x;y [x/y][ a ; b], for some b A, where y is the
first variable not ocurring neither in x nor in . 

As above, the context x will be omitted when convenient. Moreover, when x = ,


we will write P LFI1 instead of P LFI1
[].

Proposition 7.9.5 Let P = A, ()P  be a partial structure over  and let be a
formula in L  with a context x = x1 , . . . , xk , where k 1. Then:
P 
(i) + = a : P LFI1x [ a] ;
P  LFI1

(ii) = a : P x [ a] ;
P    
(iii) b = a : P LFI1 x [ a ] and P LFI1 x [
a] ;
 
(iv) if P LFI1
x [a ] then P LFI1 x  [ a ];
P P  
(v) + b = a : P LFI1 x [
a ] ;
P P   
(vi) b = a : P LFI1 x [ a] .

Proof Items (i), (ii) and (iii) will be proved simultaneously, by induction on the
P
complexity of . Moreover, just items (i) and (ii) are to be proved since b is, by
P P
definition, the complement of + relative to Ak .
If = R(t1 , . . . , tm ) is atomic, then a +P iff (t1P [ a ], . . . , tmP [
a ]) R+ P
(by Def-
P P P P
inition 7.9.2) iff (t1 [ a ], . . . , tm [
a ])
/ R Rb iff, by Definition 7.9.4, Item (2),

P LFI1
x R(t 1 , . . . , t m )[
a ].
P P P P
On the other hand, a iff, by Definition 7.9.2, (t1 [ a ], . . . , tm [
a ]) R iff
P P P P 
(t1 [
a ], . . . , tm [
a ]) / R+ Rb iff P LFI1 x R(t1 , . . . , tm )[
a ], by Definition 7.9.4,
Item (1).
7.9 First-Order LFI1 and the Logic of Quasi-truth 329

Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH).
P P 
Case 1: = . Then a + iff a , by Definition 7.9.2, iff P LFI1 x [ a ],
 
by (IH), iff P x LFI1
[ a ], by Definition 7.9.4, Item (3), iff P x LFI1
[ a ].
P P 
On the other hand, a iff a + , by Definition 7.9.2, iff P LFI1 x [ a ],
LFI1
by (IH), iff P x [ a ].
P P P 
Case 2: = . Then a + iff a + iff either P LFI1 x [ a ] or
 
P x LFI1
[a ], by (IH), iff P x LFI1
[a ], by Definition 7.9.4, Item (4).
P P  
On the other hand, a iff a b iff P LFI1 x [ a ] and P LFI1 x [ a ],

by (IH), iff P LFI1 x [ a ], by Definition 7.9.4, Item (10).
Case 3: = . Then a +P iff a +P +P iff P LFI1
 
x [ a ] and P LFI1 x

[ a ], by (IH), iff P LFI1 x [ a ], by Definition 7.9.4, Item (5).
P P P  
On the other hand, a iff a iff either P LFI1 x [ a ] or P LFI1 x

[
a ], by (IH), iff P LFI1 x [ a ], by Definition 7.9.4, Item (11).
Case 4: = . It is proved analogously to the Case 3.
P P P 
Case 5: = . Then a iff a + iff either P LFI1 x [ a ] or
 
P x LFI1
[a ], by (IH), iff P x LFI1
[a ], by Definition 7.9.4, Item (7).
P P P P P P
On the other hand, a iff a (+ b ) iff a / and a iff
  
P LFI1
x [
a ] and P LFI1 x [
a ], by (IH), iff P LFI1 x [a ], by Definition 7.9.4,
Item (13).  

P P
Case 6: = x. Then a + iff a [x/y]+ iff
x;y

 

a (
a ; b) : P LFI1
x;y [x/y][
a ; b] ,

 
by (IH), iff P LFI1
x;y [x/y][a ; b] for every b, iff P LFI1
x x[
a ], by Defini-
tion 7.9.4, Item (8). 
P P
On the other hand, a iff a ([x/y] )x;y iff
 

a (
a ; b) : P LFI1
x;y [x/y][
a ; b] ,

 
by (IH), iff P LFI1
x;y [x/y][ a ; b] for some b, iff P LFI1
x x[ a ], by Defini-
tion 7.9.4, Item (14).
Case 7: = x. The proof is analogous to the case 6.
(iv) It is proved by induction on the complexity of , by using Definition 7.9.4. The
details of the proof are left as an exercise to the reader.
(v) It is immediate from items (i), (iii) and (iv).
(vi) It is immediate from items (ii), (iii) and (iv) (by taking the contrapositive). 

Concerning sentences, the following holds as a corollary:

330 7 First-Order LFIs

Corollary 7.9.6 Let P = A, ()P  be a partial structure over  and let be a
sentence of L  . Then:
(i) P {1, 21 } iff P LFI1 ;
(ii) P {0, 21 } iff P LFI1 ;
(iii) P = 1 iff P LFI1 ;
(iv) P = 0 iff P LFI1 ;
(v) P = 21 iff P LFI1 and P LFI1 .

The notion of semantic entailment for LFI1 w.r.t. partial structures is defined as
expected.

Definition 7.9.7 (The logic LFI1 P of partial structures for LFI1 ) Let  be a first-
order signature. The logic LFI1 P generated by the semantic entailment for LFI1
w.r.t. partial structures over  is defined as follows: let  {} a set of formulas
in L  and let x = x1 , . . . , xk a context for it. We say that is a consequence of

 in LFI1 P (in context x), denoted by  LFI1 x if, for every partial structure P

over  and every sequence a in A interpreting x: P LFI1 x [a ] for every 

implies that P x LFI1
[
a ]. In particular, if  {} is a set of sentences of L  then
 LFI1 if P LFI1 for every partial structure P for LFI1 over  such that
P LFI1 for every . 

Remark 7.9.8 The situation described in Remark 7.3.13 still holds for the logic
LFI1 P . Thus, if  {} a set of formulas in L  and x = x1 , . . . , xk is a context for
it, then

 LFI1
x () LFI1 ()

where () = x1 xk and () = {() : }.

It will be convenient to consider extended partial structures defined over diagram

languages, as it was done with Tarskian structures in Definition 7.3.2.

Definition 7.9.9 (Extended partial structures) Given a partial structure P = A,

()P  over a signature , the diagram language of P, denoted by L  (P), or simply
L(P), is defined over the signature  A , which is obtained from  by adding a new
individual constant a for each element a of A. The set of terms of the diagram
language of P will be denoted by T (P), or simply by T (P).

 = A, ()P
The partial structure P is extended to a partial structure P  over  A

P
by defining (a) = a for every a A. 

As in Definition 7.3.3, given a partial structure P over , a formula in L  ,

a context x = x1 , . . . , xk for and a sequence a in A interpreting x, [ x /
a ] will
denote the sentence of L  (P) obtained from by substituting simultaneously every
free occurrence of variable xi by the constant a i , for 1 i k. Analogously, given a
term t depending on the variables of x, then t[x / a ] is the closed term obtained from
t by substituting simultaneously every occurrence of variable xi by the constant a i ,
for 1 i k.
7.9 First-Order LFI1 and the Logic of Quasi-truth 331

Finally, an important technical result, namely Proposition 7.9.13, will be proved,

as in the case of classical first-order logic, by induction on the complexity of a
formula. It is convenient to state before some lemmas concerning the extended partial

structure P.

Lemma 7.9.10 Let P be a partial structure over , t a term in T (P), x; y =

x1 , . . . , xk , y a context for t, and a ; b a sequence in A interpreting x; y. Then:
 
t P [
a ; b] = t[ a ]P [b].
x /

Proof Straightforward, by induction on the complexity of the term t. 

Lemma 7.9.11 Let P be a partial structure over , t a term in T (P), x =


x1 , . . . , xk a context for t, and a a sequence in A interpreting x. Then: t P [
a] =
P
x /
t[ a] .

Proof Straightforward, by induction on the complexity of the term t. 

Lemma 7.9.12 Let P be a partial structure over , a formula in L  (P), x; y =

x1 , . . . , xk , y a context for , and a ; b a sequence in A interpreting x; y. Then:

 LFI1 [
P a ; b]  LFI1
P

[
x /
a ][b].
x;y y

Proof By induction on the complexity of the formula .

Thus, suppose that = P(t1 , . . . , tk ) is an atomic formula. The result follows by
Definition 7.9.4, Item (1), and Lemma 7.9.10.
Suppose that the result holds for every formula with complexity less or equal than
n, for n 1 (induction hypothesis, IH).
Case 1: = #, for # {, } or = #, for # {, , }. It follows by using
(IH) and by Definition 7.9.4, observing that the definition of satisfaction is recursive
w.r.t. the degree of complexity introduced in Definition 7.1.3, and by the fact that
substitutions distribute over the propositional operators.
Case 2: = x. In this case P  LFI1 [ a ; b] iff P LFI1 [x/z][ a ; b; c] for
x;y x;y;z
every c A, by Definition 7.9.4, Item (14), iff P z  LFI1
[x/z][x /
a ][y/b][c] for
every c A, by (IH) and properties of substitution, iff P  LFI1 [x/z][ x /a ][b; c]
y;z

for every c A, by (IH), iff P  y LFI1
[x /
a ][b], by Definition 7.9.4, Item (14).
Case 3: = x. It is proved as Case 2. 

Proposition 7.9.13 Let P be a partial structure over , a formula in L  (P),

x = x1 , . . . , xk a context for , and a a sequence in A interpreting x. Then:

 LFI1 [
P a]  LFI1 [
P x /
a ].
x

Proof By induction on the complexity of the formula .

Thus, suppose that = P(t1 , . . . , tk ) is an atomic formula. The result follows by
Definition 7.9.4, Item (1), and Lemma 7.9.11.
332 7 First-Order LFIs

Suppose that the result holds for every formula with complexity less or equal than
n, for n 1 (induction hypothesis, IH).
Case 1: = #, for # {, } or = #, for # {, , }. The argument is
the same used in the proof of Lemma 7.9.12, Case 1.
Case 2: = x. In this case P  LFI1 [ a ] iff P  LFI1 [x/y][ a ; b] for every
x x;y

b A, by Definition 7.9.4, Item (14), iff P  y LFI1 
[x/y][x /
a ][b] for every b A,
by Lemma 7.9.12, iff P LFI1 [ x /a ], by Definition 7.9.4, Item (14) and definition
of substitution.
Case 3: = x. It is proved as Case 2. 

7.9.2 The Logic QLFI1

This section is dedicated to define the first-order version of LFI1 , conceived accord-
ing to the criteria introduced in Sect. 7.8. The idea is that the resulting Hilbert cal-
culus, which will be called QLFI1 , will be sound and complete with respect to the
semantics for LFI1 of partial structures. That is, QLFI1 will coincide with LFI1 P .
Because of the particular features of the semantics of partial structures, additional
axioms concerning the interdefinability of quantifiers will be added to LFI1 , in
order to obtain completeness. Notice that all the quantified LFIs presented in the
previous sections were semantically characterized by paraconsistent Tarskian struc-
tures extending the ones for QmbC (see Definition 7.3.6). That being the case, and
in order to obtain the completeness theorem, it will be proved that the semantics for
QLFI1 in terms of paraconsistent Tarskian structures coincides with the notion of
semantic entailment for LFI1 w.r.t. partial structures.
Definition 7.9.14 Let  be a first-order signature. The logic QLFI1 (over ) is
defined by the Hilbert calculus obtained by extending QmbC (recall Definition 7.1.5)
expressed in the language L  by adding the axioms (ci), (ce), (cf) and all the new
axioms of LFI1 (recall Definition 4.4.41) expressed in the language L  , plus the
following:

x x (AxEqUn)
x x (AxEqEx)


The main objective now is to prove soundness and completeness of QLFI1 w.r.t.
the semantics for LFI1 of partial structures. Because of Remark 7.9.8, and the fact
that the property therein observed also holds in QLFI1 , it will be enough to prove
soundness and completeness for derivations involving sentences.
The first result is easily obtained:
Theorem 7.9.15 (Soundness of QLFI1 with respect to partial structures for LFI1)
For every set of sentences  {} over : if  QLFI1 then  LFI1 .
7.9 First-Order LFI1 and the Logic of Quasi-truth 333

As it was discussed in Sect. 7.8, any axiomatic extension of QmbC can be seman-
tically characterized by interpretations of the  A, v, such that A is a classical
 form
Tarskian structure over  and v : SL(A) 0, 1 is a suitable valuation. In the case
of QLFI1 , the valuations are defined as follows:

Definition 7.9.16 (QLFI1 -valuations) Let A be a structure over   with

 domain A.
A QLFI1 -valuation over A is a QmbC-valuation v : SL(A) 0, 1 over A (recall
Definition 7.3.5) which also satisfies the following clauses (see Definition 4.4.42):
(vCon) v() = 1 = v() = 1 and v() = 1.
(vNeg) v() = v().
(vDM ) v(( )) = 1 v() = 1 or v() = 1.
(vDM ) v(( )) = 1 v() = v() = 1.
(vCIp ) v(( )) = 1 v() = v() = 1.
(vExUni1) v(x) = v(x).
(vExUni2) v(x) = v(x). 

Definition 7.9.17 (Interpretations for QLFI1 ) An interpretation for QLFI1 (over

the signature
 ) is a pair A, v, such that A is a structure over  and v : SL(A)
0, 1 is a QLFI1 -valuation. 

It is straightforward to extend to QLFI1 the proofs of soundness and complete-

ness of QmbC w.r.t. interpretations, obtaning so the following:

Theorem 7.9.18 (Soundness and Completeness of QLFI1 with respect to inter-

pretations) For every set of sentences  {} over a signature :

 QLFI1  QLFI1 .

Corollary 7.9.19 The logic QLFI1 is a (strong) LFI.

From now on, and when there is no risk of confusion, the expression R( a ) will
denote the sentence R( x )[
x /
a ], that is, R(a 1 , . . . , a k ), when R is a k-ary predicate
symbol, x = x1 , . . . , xk is a context and a = a1 , . . . , ak is a sequence of elements in
a set A interpreting that context.

Definition 7.9.20 (Partial structure associated to an interpretation for QLFI1 )

Let I = A, v be an interpretation for QLFI1 over the signature , where A =
A, IA . The partial structure over  associated to I is P(I) = A, ()P(I)  such
that f P(I) = IA ( f ) and cP(I) = IA (c), for every symbol for function f and every
constant c, respectively. If R is a symbol for a k-ary relation then
P(I)  
R+ = a Ak : v(R( a ) = 1, v(R( a )) = 0
P(I)  
R = a Ak : v(R(
a ) = 0, v(R(
a )) = 1
P(I)  
Rb = a Ak : v(R(
a ) = 1, v(R(
a )) = 1 . 
334 7 First-Order LFIs

In the proof of the following result, the notion of complexity of formulas given in
Definition 7.1.3 will be useful, since it states that the complexity of a formula Qx
is strictly greater than the complexity of [x/t] for every term t, where Q {, }.

Proposition 7.9.21 Let I and P = P(I) as in Definition 7.9.20. Then for every
def

sentence in L  (A) = L  (P):

1 iff v() = 1 and v() = 0;

P
= 0 if v() = 0 and v() = 1;
1
2
if v() = 1 and v() = 1.

Proof The proof will be done by induction on the complexity of the sentence . In
order to simplify the proof, it will be useful to observe that, because of clause (vNeg),
it suffices to prove the following:

1 iff v() = 0 (a);

P = 0 if v() = 0 (b);
1
2
if v() = 1 and v() = 1 (c).

Thus, if is atomic, the result holds by the very definition of P and Definition 7.9.4,
items 1 and 2.
Suppose that the result holds for every formula with complexity less or equal than
n, for n 1 (induction hypothesis, IH).
Case 1: = #, for # {, } or = #, for # {, , }. Because of Propo-
sition 7.9.3, the proof is analogous to that of Lemma 4.4.44, and so it will be omitted.
Case 2: = x. Then:
  LFI1 x, by Corollary 7.9.6, Item (iii), iff P  LFI1 
(a) P = 1 iff P y
[x/y][b] for every b A, by Definition 7.9.4, Item (8), iff P  LFI1 [x/y]
P
[y/b] for every b A, by Proposition 7.9.13, iff ([x/b]) = 1 for every b A,
by Corollary 7.9.6, Item (iii), iff v([x/b]) = 0 for every b A, by (IH), iff
v(x) = 0, by clause (vEx), iff v(x) = 0, by clause (vExUni1).
  LFI1 x, by Corollary 7.9.6, Item (iv), iff P
 LFI1 
(b) P = 0 iff P y [x/y][b]
for some b A, by Definition 7.9.4, Item (14), iff P  LFI1 [x/y][y/b] for some
P 
b A, by Proposition 7.9.13, iff ([x/b]) = 0 for some b A, by Corollary 7.9.6,
Item (iv), iff v([x/b]) = 0 for some b A, by (IH), iff v(x) = 0, by clause
(vUni).

(c) P = 21 . It is proved analogously.
Case 3: = x. It is proved as Case 2. 

The completeness of QLFI1 with respect to partial structures can now be stated:

Theorem 7.9.22 (Completeness of QLFI1 with respect to partial structures for

LFI1) For every set of sentences  {} over :

 LFI1 =  QLFI1 .
7.9 First-Order LFI1 and the Logic of Quasi-truth 335

Proof Let  {} be a set of sentences, and assume that  QLFI1 . By complete-

ness of QLFI1 with respect to interpretations (Theorem 7.9.18), there exists an inter-
pretation I = A, v for QLFI1 such that v( ) = 1, for every , but v() = 0.
Let P(I) be the partial structure over  associated to I (see Definition 7.9.20). By
Proposition 7.9.21, P(I) LFI1 , for every , but P(I) LFI1 . From this,
 LFI1 . 

7.10 First-Order P1 and Partial Structures

In this section an extension to first-order languages of Settes 3-valued logic P1 (recall

Sect. 4.4.4) will be proposed. An interesting feature of this approach is that, as in the
case of LFI1 , a semantics of partial structures will be given, besides the valuation
semantics naturally obtained. In this way, P1 could be useful for an alternative treat-
ment of the theory of quasi-truth, on the one hand, as well as a logic framework for
dealing with information systems (such as logic programming systems or databases)
because of its particularities. Indeed, the basic feature of P1, which will be preserved
in its proposed first-order extension QP1, is that paraconsistency is only allowed at
the atomic level: any contradiction involving a complex formula is explosive. This
feature has interesting consequences, both from a conceptual perspective and from
the point of view of its potential application to information systems.

Definition 7.10.1 Let  be a first-order signature. The logic QP1 (over ) is

defined by the Hilbert calculus obtained by extending QCi (see Sect. 7.8) expressed in
the language L  by adding the axioms (cw), (cv1 )(cv3 ) (recall Definition 4.4.19),
(AxEqUn) and (AxEqEx) (recall Definition 7.9.14), all of them expressed in the
language L  , plus the following:

x (cv4 )
x (cv5 )

The logic QP1, being an axiomatic extension of QmbC, can be easily semantically
characterized by interpretations ofthe form A, v, where A is a classical Tarskian
structure over  and v : SL(A) 0, 1 is a suitable valuation.
Recall that, given a structure A for a signature  and a context x = x1 , . . . , xn ,
the set of formulas of L(A) whose free variables occur in the sequence x is denoted
by L(A)x . Analogously, T (A)x is the set of terms of L(A) whose variables occur in
the sequence x.

Definition 7.10.2 (QP1-valuations) Let A be a structure

  over  with domain A. A
QP1-valuation over A is a mapping v : SL(A) 0, 1 satisfying clauses (vPred),
(vOr), (vAnd), (vImp), (vVar), (vEx) and (vUni), for QmbC-valuations over A
336 7 First-Order LFIs

(recall Definition 7.3.5), clauses (vExUni1) and (vExUni2) for QLFI1 -valuations
over A (recall Definition 7.9.16), clauses (vNeg)a , (vNeg)c , (vCon)a and (vCon)c of
Definition 4.4.20, all of them expressed over the set SL(A) , plus the following clause:
(sNega ) For every contexts ( x ; z) and (x ; y), for every sequence (  in A inter-
a ; b)
x ; y), for every atomic formula L(A)x;z and every t T (A)x;y , if
preting (
 A 
[z/t] L(A)x;y and b = (t[x ; y/
a ; b]) then

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b]). 

It should be noted that, in clauses (vNeg)a and (vCon)a of the last definition, the
formula p is now of the form P(t1 , . . . , tn ), where P is a predicate symbol of arity n
and each ti is a closed term of the diagram language of A. On the other hand, clauses
(vNeg)c and (vCon)c include now quantified sentences. In particular, it is required
that v(Qx) = 1 for every sentence in SL(A) of the form Qx, for Q = , .

Definition 7.10.3 (Interpretations for QP1) An interpretation for QP1 (over  the
signature ) is a pair A, v, such that A is a structure over  and v : SL(A) 0, 1
is a QP1-valuation. 

Lemma 7.10.4 Let A be a structure over , and let v be a QP1-valuation over A.

Then v satisfies clauses (sNeg) and (sCon) of Definition 7.3.5.

Proof Let ( x ; z) and (

x ; y) be two contexts, (  a sequence in A interpreting
a ; b)
(
x ; y), L(A)x;z and t T (A)x;y such that t is free for z in . Suppose also that
 A 
[z/t] L(A)x;y and b = (t[ x ; y/
a ; b]) .
Clause (sNeg): If is atomic then (sNeg) holds in this case, because of clause
(sNega ). Otherwise, if is complex, suppose that

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b]).

Using clause (vNeg)c , it follows that

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b]).

Clause (sCon): If is atomic, then

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b])

because of clause (sNega ). On the other hand,

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b])
7.10 First-Order P1 and Partial Structures 337

by the proof of the atomic case of the Substitution Lemma for QmbC (Theo-
rem 7.3.14). From this, it follows that

x ; y/
v(([z/t])[  = v([
a ; b]) x ; z/
a ; b])

by clause (vCon)a .
Finally, if is complex, the proof follows from clause (vCon)c . 

Corollary 7.10.5 Every QP1-valuation is, in particular, a QmbC-valuation.

Corollary 7.10.6 (Substitution Lemma for QP1) Let t be a term free for the variable
z in the formula . Suppose that ( x ; z) and (
x ; y) are contexts for and [z/t],
 A 
respectively. Let A, v be an interpretation for QP1. If b = (t[ x ; y/
a ; b]) then:


vax;
;b a ;b
y ([z/t]) = vx;z ().

Proof Identical to that for QmbC taking into account the previous corollary. 

From this, it is now easy to prove the following:

Theorem 7.10.7 (Soundness and Completeness of QP1 with respect to interpreta-

tions) For every set of sentences  {} over a signature :

 QP1  QP1 .

Corollary 7.10.8 The logic QP1 is a (strong) LFI.

A partial structures semantics for QP1 can now be defined by adapting the one
for QLFI1 given in Sect. 7.9.1. Recall from Definition 7.9.1 the notion of partial
structure over a signature .

Definition 7.10.9 Let  be a first-order signature, a formula of L  , x = x1 , . . . , xk

a context for , and P a partial structure over . The QP1-interpretation
 of in P

P P P P
and x, denoted by QP1x , is the k-ary partial relation QP1x = + , ,
  QP1x QP1 x
P
b over A defined recursively as follows (for the sake of simplicity, the sub-
QP1
x
script QP1 and the context x will be omitted, when there is no risk of confusion):
(1) is an atomic formula R(t1 , . . . , tm ), or is : then, P is constructed as in
Definition 7.9.2 items (T1) and (T3), respectively;
P
(2) is : then,
 is given by 
P def P P P
() = b , + , ;
P
(3) is : then,
 is given by 
P P P P P P
( )P = (+ b ) (+ b ), , ;
def
338 7 First-Order LFIs

P
(4) is : then,
 is given by 
P P P P P P
( )P = (+ b ) (+ b ), , ;
def

P
(5) is : then,
 is given by 
P def P P P P P P
( ) = (+ b ), (+ b ) , ;
P
(6) is x; then,
  is given by    

P P P P
(x)x = [x/y]+ [x/y]b , [x/y] ,
def

x;y x;y x;y

where y is the first variable not ocurring neither in x nor in ;
P
(7) is x; then,
  is given by     
P def P P P
(x)x = [x/y]+ [x/y]b , [x/y] ,
x;y x;y x;y
where y is the first variable not ocurring neither in x nor in .
P
A partial structure P equipped with an interpretation mapping ()x as above, for
every context x, is called a Partial structure for QP1. 

It is worth noting that the definition of the interpretation of the quantifiers in QP1
coincides with the one given by Silvestrini in [17], in the context of the logic LPT1.
Recall from Definition 4.4.17 that P1 is defined by means of the matrix M P over a
signature just contaning implication and negation. As it was done in Lemma 4.4.22,
P1 can be regarded as defined by a 3-valued matrix M P1 over the signature ,
where is defined as above, and where conjunction and disjunction are interpreted
as the operators and ! defined in Sect. 4.4.2. From this perspective, analogous of
Proposition 7.9.3 can be obtained:

Proposition 7.10.10 Let P be a partial structure over a first-order signature , and

consider the matrix M P1 for P1 over the propositional signature . Then, for every
sentences and over , the following holds:
(1) (#)P = #( P ) for # {, }, where the operators are as in M P1 ;
(2) (#)P = ( P )#( P ) for # {, }, where the operators are as in M P1 .

As it was done with LFI1 , it is possible to re-introduce the interpretation mapping

provided in Definition 7.10.9 in a Tarskian-style, that is, by means of a list of clauses
recursively defined in a way that resembles the usual definition of satisfaction in a
structure in classical first-order logic.

Definition 7.10.11 Let be a formula over a first-order signature , x = x1 , . . . , xk

a context for it, P = A, ()P  a partial structure over  and a a sequence in A
interpreting x. We say that P satisfies in x and a w.r.t. QP1, denoted as P QP1 x
[a ], in the following situations:
P P P P
(1) P QP1
x R(t1 , . . . , tm )[
a ] iff (t1 [a ], . . . , tm [a ]) R+ Rb ,
whenever R(t1 , . . . , tm ) is atomic
P P P P
(2) P QP1
x R(t1 , . . . , tm )[ a ] iff (t1 [a ], . . . , tm [a ]) R Rb ,
whenever R(t1 , . . . , tm ) is atomic
7.10 First-Order P1 and Partial Structures 339

P P P P
(3) P QP1x R(t1 , . . . , tm )[
a ] iff (t1 [ a ], . . . , tm [
a ]) R+ R ,
whenever R(t1 , . . . , tm ) is atomic
(4) P QP1x [ a ] iff P QP1 x [ a ], whenever is complex
QP1
(5) P x [ a ], whenever is complex
(6) P QP1x ( )[ a ] iff P QP1 x a ] and P QP1
[ x [ a]
QP1 QP1 QP1
(7) P x ( )[ a ] iff either P x [ a ] or P x [ a]
(8) P QP1x ( )[ a ] iff either P  QP1
x [a ] or P QP1
x [
a]
QP1 QP1
(9) P x x[ a ] iff P x;y [x/y][ a ; b], for every b A, where y is the
first variable not ocurring neither in x nor in
(10) P QP1x x[ a ] iff P QP1 x;y [x/y][ a ; b], for some b A, where y is the
first variable not ocurring neither in x nor in . 

Once again, the context x will be omitted when convenient. Additionally when
x = , we will write P QP1 instead of P QP1 [].
As in the case of LFI1 , the following useful result can be stated for the semantics
of partial structures for QP1:

Proposition 7.10.12 Let P = A, ()P  be a partial structure over  and let be
a formula in L  with a context x = x1 , . . . , xk , where k 1. Then:
 
P
(i) + = a : P QP1x [ a ] ;
 
P QP1
(ii) = a : P x [ a] ;
 
P QP1
(iii) b = a : P x [ a ] and P QP1 x [ a ] (= , if is complex);
 
P P
(iv) + b = a : P QP1 x [
a ] ;
 
P P
(v) b = a : P QP1 x [a] .

Proof As in the proof of Proposition 7.9.5, items (i) and (ii) will be proved simultane-
P
ously, by induction on the complexity of . Since b is, by definition, the complement
P P
of + relative to Ak , item (iii) follows from (i) and (ii), in case is atomic; if
is complex, it follows by Definition 7.10.9.
If = R(t1 , . . . , tm ) is atomic, the proof is identical to that given for this subcase
of Proposition 7.9.5.
Assume that the result holds for every formula with complexity k < n (induction
hypotesis, IH).
P P P
Case 1: = . If = R(t1 , . . . , tm ) is atomic, a + iff a b , by
P P P P QP1
Definition 7.10.9, iff (t1 [ a ], . . . , tm [
a ]) R Rb , iff P x [ a ], by Def-
QP1
inition 7.10.11, Item (2), iff P x [ a ], by Item (4) of the same definition.
P P P P
Otherwise, if is complex, a + iff a b = , by Definition 7.10.9,
QP1 QP1
iff P x [ a ], by (IH), iff P x [ a ], by Definition 7.10.11, Item (4), iff
P QP1x [a ], by the same item.
340 7 First-Order LFIs

P P
On the other hand (being atomic or not), a iff a + , by Defini-
tion 7.10.9, iff P QP1 x [ a ], by (IH), iff P QP1
x [a ].
P P P
Case 2: = . If = R(t1 , . . . , tm ) is atomic, a + iff (t1 [ a ], . . . , tm [
a ])
P P QP1 QP1
R+ R , iff P x [ a ], by Definition 7.10.11, Item (3), iff P x [ a ], by
Definition 7.10.11, Item (4).
P P P P
On the other hand, a iff a b , iff (t1 [ a ], . . . , tm [
a ])
P P P P P QP1
Rb , iff (t1 [ a ], . . . , tm [ a ]) / R+ R , iff P x [ a ], by Definition 7.10.11,
Item (3), iff P QP1 x [ a ].
Now, if is complex, the result is obvious from the definitions.
P P P P P
Case 3: = . Then a + iff a (+ b ) (+ b ) iff P QP1 x [ a]
and P QP1 x [
a ], by (IH), iff P QP1
x ( )[a ], by Definition 7.10.11, Item (6),
iff P QP1
x [ a ], by Definition 7.9.4, Item (4).
P P P
On the other hand, a iff a iff either P QP1 x [a ] or P QP1 x
QP1
[
a ], by (IH), iff P x [ a ], by Definition 7.10.11, Item (6).
Case 4: = . It is proved analogously to the Case 3.
P P P P
Case 5: = . Then a iff a (+ b ) iff either P QP1 x [ a]
QP1 QP1
or P x [ a ], by (IH), iff P x ( )[ a ], by Definition 7.10.11, Item (8),
iff P QP1
x [ a ], by Definition 7.10.11, Item (4).
P P P P
On the other hand, a iff a (+ b ) , iff P QP1 x [a ] and P QP1 x
a ], by (IH), iff P QP1
[ [a ], by Definition
 7.9.4, Item (8).
x
 
P P P
Case 6: = x. Then a + iff a [x/y]+ [x/y]b , iff
x;y x;y

 
a a ; b) : P QP1
( x;y [x/y][
a ; b] ,

by (IH), iff P QP1 a ; b] for every b, iff P QP1

x;y [x/y][ x x[
a ], by Defini-
tion 7.9.4, Item (9), iff P QP1
x [
a ], by
 Definition 7.10.11,
Item (4).
P P
On the other hand, a iff a ([x/y] )x;y iff
 
a a ; b) : P QP1
( x;y [x/y][
a ; b] ,

by (IH), iff P QP1 a ; b] for some b, iff P QP1

x;y [x/y][ x [
a ], by Definition 7.9.4,
Item (9).
Case 7: = x. The proof is analogous to the case 6.
(iv) and (v): It is immediate from the previous items. 
Concerning sentences, the following holds as a corollary:
Corollary 7.10.13 Let P = A, ()P  be a partial structure over  and let be a
sentence of L  . Then:
7.10 First-Order P1 and Partial Structures 341

(i) P {1, 21 } iff P QP1 ;

(ii) P {0, 21 } iff P QP1 ;
(iii) P = 1 iff P QP1 ;
(iv) P = 0 iff P QP1 ;
(v) P = 21 iff P QP1 and P QP1 .

The notion of semantic entailment for QP1 w.r.t. partial structures is defined as the
corresponding one for LFI1 .

Definition 7.10.14 (The logic QP1 P of partial structures for QP1) Let  be a
first-order signature. The logic QP1 P generated by the semantic entailment for QP1
w.r.t. partial structures over  is defined as follows: let  {} a set of formulas
in L  and let x = x1 , . . . , xk a context for it. We say that is a consequence of 
in QP1 P (in context x), denoted by  QP1 x if, for every partial structure P over
 and every sequence a in A interpreting x: P QP1 x [a ] for every  implies
that P QP1
x [a ]. In particular, if  {} is a set of sentences of L  then  QP1
if P QP1 for every partial structure P for QP1 over  such that P QP1 for
every . 

Remark 7.10.15 As in the case of the previous systems, the semantical consequence
relation can be defined in terms of sentences. That is, if  {} a set of formulas in
L  and x = x1 , . . . , xk is a context for it, then

 QP1
x () QP1 ()

where () = x1 xk and () = {() : }.

The soundness of QP1 w.r.t. partial structures can be easily proved:

Theorem 7.10.16 (Soundness of QP1 with respect to partial structures) For every
set of sentences  {} over : if  QP1 then  QP1 .

Recall from Definition 7.9.9 the notion of diagram language L(P) of a partial
 defined on it. The following
structure P, as well as the extended partial structure P

results can be proved as in the case of LFI1 .

Lemma 7.10.17 Let P be a partial structure over , a formula in L  (P), x; y =

x1 , . . . , xk , y a context for , and a ; b a sequence in A interpreting x; y. Then:

 QP1 [
P a ; b]  QP1
P [
x /
a ][b].
x;y y

Proposition 7.10.18 Let P be a partial structure over , a formula in L  (P),

x = x1 , . . . , xk a context for , and a a sequence in A interpreting x. Then:

 QP1 [
P a]  QP1 [
P x /
a ].
x
342 7 First-Order LFIs

The notion of partial structure associated to an interpretation for QLFI1 (Defi-

nition 7.9.20) can be easily adapted to QP1:

Definition 7.10.19 (Partial structure associated to an interpretation for QP1) Let

I = A, v be an interpretation for QP1 over the signature , where A = A, IA .
The partial structure over  associated to I is P(I) = A, ()P(I)  such that f P(I) =
IA ( f ) and cP(I) = IA (c), for every symbol for function f and every constant c,
respectively. If R is a symbol for a k-ary relation then
P(I)  
R+ = a Ak : v(R( a )) = 1, v(R( a )) = 0
P(I)  
R = a Ak : v(R(
a )) = 0, v(R(
a )) = 1
P(I)  
Rb = a Ak : v(R(
a )) = 1, v(R(
a )) = 1 . 

The proof of the following result is analogous to that of Proposition 7.9.21, but
now using Proposition 7.10.10 (and so Lemma 4.4.22), as well as Corollary 7.10.13.

Proposition 7.10.20 Let I and P = P(I) as in Definition 7.10.19. Then, for every
def

sentence in L  (A) = L  (P):

1 iff v() = 1 and v() = 0;

P
= 0 if v() = 0 and v() = 1;
1
2
if v() = 1 and v() = 1.

From this, the completeness of QP1 with respect to partial structures follows
straightforwardly:

Theorem 7.10.21 (Completeness of QP1 with respect to partial structures) For

every set of sentences  {} over :

 QP1 =  QP1 .

This chapter has made clear that basically the same results of classical model
theory hold for QmbC, and for first-order LFIs in general. Important results as
the Completeness, the Compactness and the Lowenhim-Skolem Theorems can be
proved for first-order LFIs as much as they can be proved in the classical case, under
certain provisos due to the inherent characteristics of the logics involved. What all this
means is not only that paraconsistent logics, and specially their first-order versions,
cannot be overshadowed by classical logic, but also that first-order LFIs are ripe
enough for serious applications as in paraconsistent set theory (treated in Chap. 8)
or in any other area where its classical correspondent falls short of its expectations.
References 343

References

1. da Costa, Newton C.A. 1993. Sistemas formais inconsistentes (Inconsistent formal systems,
in Portuguese). Habilitation thesis, Universidade Federal do Paran, Curitiba, Brazil, 1963.
Republished by Editora UFPR, Curitiba, Brazil.
2. da Costa, Newton C.A. 1974. On the theory of inconsistent formal systems (Lecture delivered
at the First Latin-American Colloquium on Mathematical Logic, held at Santiago, Chile, July
1970). Notre Dame Journal of Formal Logic 15(4): 497510.
3. Alves, Elias H. 1984. Paraconsistent logic and model theory. Studia Logica 43(1/2): 1732.
4. DOttaviano, Itala M.L. 1982. Sobre uma Teoria de Modelos Trivalente (On a three-valued
model theory, in Portuguese). Ph.D. thesis, IMECC, State University of Campinas, Brazil.
5. DOttaviano, Itala M.L. 1985. The completeness and compactness of a three-valued first-order
logic. Revista Colombiana de Matemticas, XIX(12): 7794.
6. DOttaviano, Itala M. L. 1985. The model extension theorems for J3-theories. In Methods
in Mathematical Logic. Proceedings of the 6th Latin American Symposium on Mathematical
Logic held in Caracas, Venezuela, August 16, 1983. Lecture Notes in Mathematics, vol. 1130,
ed. by Carlos A. Di Prisco, 157173. Berlin: Springer.
7. DOttaviano, Itala M.L. 1987. Definability and quantifier elimination for J3-theories. Studia
Logica 46(1): 3754.
8. Avron, Arnon, and Anna Zamansky. 2007. Many-valued non-deterministic semantics for first-
order Logics of Formal (In)Consistency. In Algebraic and Proof-Theoretic Aspects of Non-
classical Logics. Papers in Honor of Daniele Mundici on the occasion of his 60th birth-
day. Lecture Notes in Computer Science, vol. 4460, ed. S. Aguzzoli, A. Ciabattoni, B. Gerla,
C. Manara, and V. Marra, 124. Springer.
9. Carnielli, Walter A., Marcelo E. Coniglio, Rodrigo Podiacki, and Tarcsio Rodrigues. 2014. On
the way to a wider model theory: Completeness theorems for first-order logics of formal incon-
sistency. The Review of Symbolic Logic 7(3): 548578. doi:10.1017/S1755020314000148.
10. Coniglio, Marcelo E., and Luiz H. Silvestrini. 2014. An alternative approach for quasi-truth.
Logic Journal of the IGPL 22(2): 387410. doi:10.1093/ljigpal/jzt026.
11. Shoenfield, Joseph R. 1967. Mathematical Logic. Reading, MA: Addison-Wesley.
12. Mendelson, Elliot. 1987. Introduction to Mathematical Logic, 3rd ed. New York: Chapman &
Hall.
13. Chang, Chen C., and H. Jerome Keisler. 1990. Model theory. 3rd edn. Elsevier.
14. Carnielli, Walter A., Marcelo E. Coniglio, and Joo Marcos. 2007. Logics of Formal Inconsis-
tency. In Handbook of Philosophical Logic ed. by Dov M. Gabbay and Franz Guenthner, 2nd
edn, vol. 193. Springer. doi:10.1007/978-1-4020-6324-4_1.
15. Mikenberg, Irene, Newton C. A. da Costa, and Rolando Chuaqui. 1986. Pragmatic truth and
approximation to truth. The Journal of Symbolic Logic 51(1): 201221.
16. Bueno, Otvio, and Newton C. A. da Costa. 2007. Quasi-truth, paraconsistency, and the foun-
dations of science. Synthese 154(3): 383399.
17. Silvestrini, Luiz H. 2011. Uma Nova Abordagem Para A Noo De Quase-Verdade (A New
Approach to the Notion of Quasi-Truth, in Portuguese). Ph.D. thesis, IFCH, State University
of Campinas, Brazil. http://www.bibliotecadigital.unicamp.br/document/?code=000788964&
opt=4&lg=en_US.
Chapter 8
Paraconsistent Set Theory

Intuitively, a set is any collection of elements that satisfy a given property. By allowing
arbitrary properties one finds a very flexible way of specifying setsthe method of
abstraction. In this method a set is specified by giving the condition that an object
must satisfy in order to belong to the set. In this way, the basic axiom known as the
Schema of Comprehension or Principle of Comprehension proposed by Frege
in 1893 in the course of his investigations (see [1]) was the following:
If is a property (denoted by a first-order formula (x)) then there exists a set y formed by
all elements satisfying property , that is, y = {x : (x)}.

However, this form of conceiving sets, known as naive set theory, a legacy of the
early developments by Dedekind, Cantor and Frege (the last two not without rivalry),
leads to a contradiction, as it is well known from Russells paradox: just take (x)
as x / x, and the consequence there exists a set y which contains exactly those
elements x which satisfy the condition x / x is contradictory: the just obtained set
y will be such that y y y / y.
A beautiful, intuitive and useful idea lost forever? Not necessarily. A paraconsis-
tent approach to set theory makes it possible to have theories of truth and sethood in
which the mathematically fundamental intuitions about these notions are respected:
contradictions may be allowed to arise in a paraconsistent set theory, but these need
not infect the whole theory.
This chapter is dedicated to showing how core mathematics can be supported by
paraconsistent set theory, and how some of George Cantors seminal intuitions on
sets can be revisited in the light of paraconsistency, specially by means of employing
the logics of formal inconsistency. The main idea is to assume that not only sentences
can be taken to be consistent or inconsistent, but also that sets themselves can be
thought to be consistent or inconsistent. We establish the basis for new paraconsistent
set-theories (such as ZFmbC and ZFCil) under this perspective and establish their
non-triviality, provided that ZF is consistent. By recalling how G. Cantor himself, in
his efforts towards founding set theory more than a century ago, not only used a form

Springer International Publishing Switzerland 2016 345

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_8
346 8 Paraconsistent Set Theory

of inconsistent sets in his mathematical reasoning, but regarded contradictions as

beneficial, we argue that Cantors handling of inconsistent collections can be related
to ours. It is to be noted that Cantor had already proved by 1891 his now famous
Cantors Theorem, which shows that the cardinal number of the set P(S) of all
subsets of a given set S is different to the cardinal number of S. Cantors argument
shows directly that, if the universe is a set, Russells paradox obtains (see, for instance,
[2] for a short proof). This chapter is based on the ideas of the paper [3] and in the
references therein.

8.1 Antinomic Sets and Paraconsistency

Ever since the discovery of the paradoxes, the history of contemporary set theory
has centered around attempts to rescue Cantor-Freges naive theory from triviality,
traditionally by placing the blame on Principle of Comprehension (also referred
to as the Principle of (unrestricted) Abstraction), as mentioned above. Unrestricted
abstraction (which allows sets to be defined by arbitrary conditions) plus the axiom
of extensionality, and plus the laws of the underlying logic where the theory of sets
is expressed, leads to a contradiction when a weird collection such as the Russell
set or similar constructions are defined. The problem is not the weird collections by
themselvesset-theorists are used to strange objects like large cardinals, measur-
able cardinals and the like, and, in fact, hypothesizing on large cardinals enables us
to investigate the capabilities of possible extensions of ZFC. The problem is that
some weird sets, such as Russells, entail a contradiction, and in classical logic a
contradiction entails everything.
One way of escaping this mathematical catastrophe is to consider weaker forms
of separation, by patching in the Principle of Comprehension, but this seems to rem-
edy the problem just temporarily, since (as it is a well-known consequence of the
celebrated Gdels Incompleteness Theorems) there is no possibility of proving the
absolute consistency of set theory. A radical departure from this position, the paracon-
sistent one, advocates that the underlying paradoxes are not necessarily to be solved,
but that they can be made part of the rational arsenal of the working mathematician,
as well as of the philosopher interested in foundations of mathematics. This kind
of rational accommodation can be achieved if the underlying consequence relation
is paraconsistent, instead of classicalin such a case, contradictions remain, so to
speak, quarantined, and the triviality disease does not spread all over the universe.
Much has been said and written on paraconsistent set theory, chiefly because of
Russells antinomy and its alleged dangerous consequences. In the usual classical
set theory, classical negation plus an unrestricted application of the comprehension
(or separation) scheme may cause deductive trivialization of all mathematics, if one
takes seriously the dogma that mathematics must be founded on sethood.
Paraconsistent set theory, in a nutshell, is the theoretical move of maintaining
weird sets as much as we canwe call them inconsistent sets for reasons which
will be made clear in the followingand weakening the underlying logic governing
8.1 Antinomic Sets and Paraconsistency 347

sets so as to avoid the disastrous consequences of inconsistent sets. This move is in

frank opposition with traditional strategies, which maintain the underlying logic and
weaken the Principle of Abstraction. E. Zermelos proposal in 1908 of an axiomati-
zation of set theory by replacing the Principle of Abstraction with a weaker existence
axiom, the axiom of separation (Aussonderung), gave birth to ZF set theory and type
theory, but represented a radical departure from Cantors intuitions. Indeed, Cantor
regarded the antinomies as positive results which fully complemented the advance
of his investigation. Not only Cantor, but his contemporaries J. Hadamard and P.
Jourdain also held similar ideas. Somehow, Cantor had already conceived a form of
inconsistent sets (see [4]):
A collection [Vielheit] can be so constituted that the assumption of a unification of all its
elements into a whole leads to a contradiction, so that it is impossible to conceive of the
collection as a unity, as a completed object. Such collections I call absolute infinite or
inconsistent collections.
(Cantor, letter to Dedekind, 1899)

What may be surprising in many ways is that Cantor indeed reasoned with such
inconsistent sets, and regarded contradictions as beneficial, incorporating them into
his philosophy of the infinite. By taking such inconsistent collections into account,
Cantor concluded that every consistent set had to be power-equivalent to a definite
Aleph. As Dauben, in the most comprehensive biography of Cantor to date ([4],
p. 244), puts it:
In one theorem, based on the inconsistency of the system of all transfinite numbers, Cantor
had succeeded in resolving several perplexing and long-standing problems of set theory.

Could Cantors approach be related to paraconsistent set theory? And if so, to

which one? Paraconsistent set theory has been around since at least fifty years pro-
posed by da Costa, who inaugurated the topic in his [5]. In the 1980s, A. Arruda and
N.C.A. da Costa studied the systems NFn , a kind of weakened version of Quines
NF; several results had been published by them between 1964 and 1975. The first
proponents of the paraconsistent approach to set theory have found some puzzling
problems, however. Arruda showed in 1980 (see [6, 7]) that, without a restriction to
the comprehension axiom, all the set-theoretical systems in the hierarchy NFn are
trivial. She also proposed alternative systems which apparently fix the problem, con-
jecturing their equiconsistency with NF; additional results were obtained by Arruda
and Batens in 1982 (see [8]). In addition, in 1986 da Costa himself (in [7]) proposed
an improved version of his paraconsistent set theory by introducing a hierarchy called
CHUn which is obtained from Churchs set theory CHU (see [9]) by changing the
underlying first-order logic by Cn= (da Costas first-order logic Cn plus equality), and
by adding two versions of the axioms involving negation, one using strong (classi-
cal) negation and another using weak (paraconsistent) negation (see [10]). He proves
that CHU is consistent if and only if each CHUn is non-trivial. An interesting fea-
ture of CHUn is that both the universal set and the Russell set are allowed. More
details of this proposal and other features and applications of the diverse da Costas
paraconsistent set theories can be found in [11].
348 8 Paraconsistent Set Theory

Another approach to paraconsistent set theory was proposed in [12] by R. Routley

(who changed in 1983 his name to R. Sylvan), by introducing a system for set theory
based on relevant logic in which the full comprehension principle holds. In 1989,
Brady (see [13]) proved that there is a non-trivial model for Routleys paraconsistent
set theory (including some contradictions such as Russells), and that therefore is
definitely non-trivial. Bradys results represent an important step in the development
of Routleys set theory, not only by producing a model of dialetheic set theory which
shows its non-triviality, but because Routleys set theory includes the metaphysical
perspective that the universe is comprised of, aside from real objects, nonexistent,
contradictory or even absurd objects that one can imagine or think about. Seen from
this perspective, paraconsistent set theory is instrumental for the whole of rationality,
far beyond the foundations of mathematics.
In a more recent investigation (see [14]), the question of the existence of natural
models for a paraconsistent version of naive set theory is reconsidered. There it is
proved that allowing the equality relation in formulas defining sets (within an exten-
sional universe) compels the use of non-monotonic operators, and that fixed points
can be used to obtain certain kind of models. Already in 1994, a naive set theory with
paraconsistent basis was investigated in [15]; in 2010, Weber (see [16]), by revisiting
Sylvans ideas, introduced an axiomatic system for naive set theory (i.e., with a full
comprehension principle) in a paraconsistent setting based on relevant logic. In [17]
a non-trivial (but inconsistent) set theory based on unrestricted comprehension was
also proposed, formalized by means of an adaptive logic.
The main difficulty with all such views on set theory from a paraconsistent view-
point is that they seem to be all quite unconnected and apparently ad hoc. However,
it can be argued that in essence they are really similar. In this chapter we take up the
discussion, showing that a wide approach to antinomic set theory can be given by
means of the paraconsistent LFIs.
This chapter expounds a new axiomatic paraconsistent set theory based on some of
the first-order LFIs examined in Chap. 7, by admitting that sets, as well as sentences,
can be either consistent or inconsistent. A salient feature of such a paraconsistent
set theory, inherited from LFIs, is that only consistent and contradictory objects will
explode into triviality (provided that ZF is non-trivial). Moreover, if we declare that
all sets and sentences are consistent, we immediately obtain traditional ZF set theory,
and nothing new.
This permits one to recover in principle, within our paraconsistent set theory, all
the concepts and definitions concerned with the predicativist program for the foun-
dations of mathematics. One could also think, although this direction has not been
yet explored, about defining some new constructs by referring to constructs which
were introduced by previous definitions by allowing some degrees of circularity
(impredicativity). This approach would introduce a form of controlled predicativism,
that might be interesting by itself. The inclusion of a consistency predicate for sets
constitutes one of the main differences with previous approaches to paraconsistent
set theory such as da Costas proposals.
8.1 Antinomic Sets and Paraconsistency 349

Inconsistent objects such as the ones Cantor was fond of, on the other hand,
can be used to entail beneficial consequences, as Cantor himself pointed out. Our
inconsistent sets are, in their way, akin to the inconsistent sets conceived by Cantor,
in the sense that the essentially unfinished sets devised by him can be seen as a
particular kind of the inconsistent sets our theory supports.

8.2 LFIs Predicating on Consistency

The basic system of paraconsistent set theory proposed here is called ZFmbC, and
consists of the first-order (plus equality) version QmbC of mbC, as introduced
in Chap. 7, over a first-order signature ZF which contains, besides the equality
predicate , a binary predicate (for membership), and an unary predicate C (for
consistency of sets).

Definition 8.2.1 (ZFmbC) The system ZFmbC is the first-order theory with equal-
ity obtained from the logic QmbC over ZF (see Definition 7.7.1) by adding the
following set-theoretic axiom schemas:
Axioms from ZF:1

z(z x z y) (x y) (Ext)
yx(x y z(z x z a)) (PwSet)
yx(x y z((z a) (x z)) (Union)
w(( w) (x)(x w x {x} w)) (Infty)
FUN by(y b x(x a (x, y))) (Repl)

Axioms (Ext), (PwSet), (Union), (Infty) and (Repl) are called, respectively, exten-
sionality axiom, powerset axiom, union axiom, axiom of infinity, and replacement
axiom.

The weak Regularity Axiom:

C(x) (y(y x) y(y x z(z x z y))) (Reg)

The Unextensionality Axiom:2

(x y) z((z x) (z
/ y)) z((z y) (z
/ x)) (UnExt)

1 The
notation for the axioms of this group will be explained below.
2 Fromnow on x y abbreviates the formula (x y) and x / y abbreviates the formula
(x y).
350 8 Paraconsistent Set Theory

Axioms for the consistency predicate:

x(x y C(x)) C(y) (Cons0)

x(C(x) (x x)) (Cons1)
x((x x) C(x)) (Cons2)

Notation used in the axioms: In the axiom of infinity (Infty), stands for the strong
empty set = {x : (x x)}; {a} = {x : (x a)}; and a b = {x : (x
def def def

a)(x b)} (all these terms are legitimated, that is, they can be constructed from the
given axioms, see Remark 8.2.2). Finally, for the replacement schema axiom (Repl),
let (x, y) be a formula where variables x and y occur free, and where variable b
does not occur free (other variables can occur free in ). Let z be a fresh variable.
Then FU N denotes the following formula:

FUN = xyz((x, y) (x, z) (y z)).

def

Remark 8.2.2 Given that ZFmbC is an axiomatic extension of QmbC, it enjoys the
deduction meta-theorem, DMT, as stated in Theorem 7.2.5 and in Corollary 7.2.6.
It is worth noting that the first five axioms of ZFmbC, when added to classical
first-order with equality, together with the usual Regularity Axiom, constitute the
well-known set theory ZF.
From (Repl), the Separation Axiom can be derived in ZFmbC:

bx((x b) ((x a) (x))) (Sep)

where (x) is a formula with variable x occurring free, and where variable b does
not occur free (other variables can occur free in ). Indeed, it is enough to consider
(x, y) = (x) (x y) in (Repl), where y does not occur free in (x) and
def

y = b. The set b created from (Sep) is denoted by {x : (x a) (x)}. The strong

empty set = {x : (x x)} is defined in ZFmbC by the Separation axiom
and by extensionality, given that  ((x a) (x x)) (x x). From ,
(PwSet) and (Repl) it can be obtained {a, b} = {x : (x a) (x b)} for every

In particular {a} is legitimated, by taking a = b. Finally a b = {a, b}
def
a and b.
where x is a notation for the set y created from axiom (Union). Observe that all
the constructions above work in ZFmbC exactly as in ZF, given that the connectives
and (the only ones that do not preserve logical equivalences) are not used in the
above derivations.

Some immediate properties of equality and membership, most of them as a conse-

quence of Regularity, are the following:

Proposition 8.2.3 Let n 2. In ZFmbC it holds that:

(i)  x((x x) (x
/ x) (x x));
8.2 LFIs Predicating on Consistency 351

(ii)  x(C(x) (x x));

(iii)  x(C(x) (x / x));
n
(iv)  x1 . . . xn ( i=1 C(xi ) ((x1 x2 ) (x2 x3 ) . . . (xn x1 )));
n
(v)  x1 . . . xn ( i=1 C(xi ) ((x1 x2 ) (x2 x3 ) . . . (xn x1 )));
(vi)  C( );
(vii)  ( ).

Proof (i) Assume that (x x)(x / x). But then it follows that z((z x)(z / x))
and so (x x), by (UnExt).
(ii) Assume that C(x). It follows, then, that C({x}), by (Cons0) and extensionality.
By regularity (replacing x by z such that z {x}) and by classical logic it follows
that (x x). The result follows by the Deduction meta-theorem DMT.
(iii) It follows from (ii) and the fact that  , by Proposition 2.4.8(iii).
(iv) Assume that C(x1 ) . . . C(xn ). Then C({x1 , . . . , xn }), by (Cons0) and exten-
sionality. From this and by regularity (replacing x by z such that z {x1 , . . . , xn }) it
follows, by reasoning in classical logic, that ((x1 x2 )(x2 x3 ). . .(xn x1 )).
The result follows by DMT.
(v) By (iv) and by  (see Proposition 2.4.8(iii)).
(vi) Since x (x ) then x((x ) C(x)). The result follows by
axiom (Cons0) and by MP.
(vii) This follows from (vi) and (Cons1). 

Remark 8.2.4 It is worth noting that  x(x x), by axiom (AxEq1) (recall
Definition 7.7.1). However, it is also a direct consequence of the extensionality axiom:
from the theorem (z x) (z x) it follows that z((z x) (z x)) and
so (x x), by the extensionality axiom and by MP. Thus, axiom (AxEq1) could
be eliminated from the theory ZFmbC (see Definition 8.2.1), since it can be derived
from the others.
Items (ii) and (iv) in Proposition 8.2.3 suggest that ZF is not contained in ZFmbC,
despite the latter being able to define the classical negation . This is a slight dis-
tinction from other paraconsistent set theories, such as da Costas CHUn systems.
Indeed, ZF can be regarded as a subsystem of CHUn , since their underlying para-
consistent logic is an extension of the classical logic, and the axioms of ZF are
derivable in such systems. On the other hand, in ZFmbC (and in its extensions) the
Regularity Axiom (Reg) is weaker than the original one from ZF: it only applies to
consistent sets. However, as we shall see in the next proposition, ZF can be seen as a
subsystem of ZFmbC up to the assumption that any set is consistent. That is, a kind
of Derivability Adjustment Theorem (or DAT) between ZF and ZFmbC, as obtained
between certain LFIs and CPL in the previous chapters, can be stated (see Theorem
8.2.5 below).

For every formula in the language of ZF let be the corresponding formula

in the language of ZFmbC obtained from by replacing every occurrence of by
. If  is a set of formulas in the language of ZF then  denotes the set of all the
formulas of the form , for a formula in . Then:
352 8 Paraconsistent Set Theory

Proposition 8.2.5 (DAT between ZF and ZFmbC) With the notation as above, let
 be a set of closed formulas of ZF and let be a closed formula of ZF. If  
holds in ZF then xC(x),   holds in ZFmbC.

Proof Since ZF is compact, has a conjunction and satisfies the Deduction meta-
theorem for closed formulas, and since preserves implication and conjunction
, it is sufficient to prove the result for theorems of ZF. Observe that, if is an
instance of an axiom of ZF other than the Regularity Axiom, then is derivable
in ZFmbC. On the other hand, if is an instance of the Regularity Axiom of ZF,
then xC(x)  holds in ZFmbC. Since ZF and ZFmbC have the same infer-
ence rules, the result follows easily by induction on the length of a derivation in
in ZF. 

At this point we can already refer to sets x which are not consistent, that is,
satisfying C(x). This does not mean that these are inconsistent sets: indeed, at this
stage, within the realm of ZFmbC, non-consistency is not the same as inconsistency,
both for formulas and for sets. The next proposition shows how non-consistent sets
could appear, namely through violation of basic properties of the identity predicate
or of the membership predicate, as expressed in Proposition 8.2.3.

Proposition 8.2.6 In ZFmbC it holds that:

(i)  (x x) C(x);
(ii)  (x x) C(x);
(iii)  C(x) ((x x) (x x));
(iv)  C(x) ((x x) (x / x)).

Proof (i) Assume that (x x). Note that (x x), by axiom (AxEq1), and so
(x x)(x x). From this it follows that (x x), by Proposition 2.3.3(1). The
result follows by (Cons2) and by the Deduction meta-theorem DMT, by observing
that the inference rules for quantifiers were not used in the derivation.
(ii) Assume that (x x). Suppose firstly that C(x). From Proposition 8.2.3(iii) it
follows that (x / x), and so (x x) (x / x). But then it follows that (x x), by
Proposition 8.2.3(i). Then C(x), by (i). By using DMT it follows that (x x) 
C(x) C(x). It is clear that (x x)  C(x) C(x). By adapting the proof
of item (3) of Theorem 7.2.1 to ZFmbC it follows that (x x)  C(x).
(iii) From C(x) and (x x) (x x) it follows that (x x) (x x) (x x),
by (Cons1), and then it follows , by (Ax10). But then C(x)  ((x x) (x
x)), by Proposition 2.4.8(ii). The result follows once again by DMT.
(iv) From C(x) and (x x) (x / x) it follows that C(x) and (x x), by
Proposition 8.2.3(i). From this it follows that C(x) and (x x) (x x), which
leads to , by the proof of item (iii). Then C(x)  ((x x) (x / x)), by
Proposition 2.4.8(ii). The result holds by the Deduction meta-theorem. 

Proposition 8.2.7 In ZFmbC it holds that:

(i)  (( ) ( ));
(ii)  (( ) ( )).
8.2 LFIs Predicating on Consistency 353

Proof (i) From ( ) ( ) it follows that ( ) and so x((x

) (x / )), by (UnExt). From this, x((x ). But  x(x ) and so
 x(x ) (by Proposition 7.2.2(x)), which leads to the bottom. Thus it follows
that  (( ) ( )).
(ii) This follows from the fact that  , by Proposition 2.4.8(iii). 

A main concern of the study presented in this chapter is to determine whether or

not the set theory ZFmbC could support contradictions without trivialization, that
is, if it is indeed paraconsistent. Specifically, the following facts would be expected
in ZFmbC:
(x x), (x x)  for some formula , and

(x x), (x
/ x)  for some formula .

On the other hand, we have the following:

Corollary 8.2.8 In ZFmbC it holds:

(i) C(x), (x x), (x x)  for every formula ;
(ii) C(x), (x x), (x
/ x)  for every formula .

Proof (i) It is a consequence of (Cons1) and Ax10).

(ii) It is a consequence of Proposition 8.2.3(ii). 

In order to guarantee that C(x) is a necessary condition in the above deriva-

tions (which, in turn, guarantee that ZFmbC is in fact paraconsistent) the following
question should be answered: is ZFmbC non-trivial? As we shall prove in Corol-
lary 8.3.13, this system, as well as its extensions introduced herein, are indeed non-
trivial, provided that ZF is consistent.
Different from other approaches to paraconsistent set theory (see, for instance,
[8, 16]) the universal set V = {x : x x} cannot be defined in ZFmbC, in order
to avoid the triviality of such a set theory. Indeed, consider the following argument,
adapted from the proof of Theorem 2.1 in [6], pp. 45: if V were definable in ZFmbC,
then x(x V ) would be a theorem and so it would be possible to define the set

A = {x : (x V ) (x)}

for every formula in which A does not occur free, by the Separation (or Compre-
hension) axiom of ZF. In particular, the strong Russell set

R = {x : (x V ) (x x)}

would be definable in ZFmbC, where is the strong (classical) negation definable

in mbC as usual. But then R = {x : (x x)}, by extensionality and the theorem
x(x V ). From this, (R R ) (R R ) would be demonstrable and so
(R R ) (R R ) would be also demonstrable, trivializing ZFmbC.
354 8 Paraconsistent Set Theory

On this basis, one may ask if it is possible to start from a weaker paraconsistent
system in which the strong negation is not definable. As Arruda has pointed out in [6],
p. 8, the Curry paradox follows in a paraconsistent set theory without strong negation
if one allows the unrestricted version of the Comprehension axiom and considers
the set
B = {x : (x x) }

such that is any formula in which B does not occur free. The only requirements
for the underlying  implication must satisfy Modus Ponens (MP)
 logic is that the
and the schema ( ) ( ) must be derivable. Indeed, let be
the formula B B . By the very definition of the set B , the following holds:

 ( ) (1)

 ( ) . (2)
 
Since  ( ) ( ) (by hypothesis about the logic) then, by
(MP) with (1), it follows that

 ( ). (3)

Using (MP) with (2) it follows that  and so, by (MP) with (3) it follows that  .
In particular, any closed sentence would be a theorem of the given logic, and so the
logic would be trivial (assuming that the schema x is derivable).
Adapting Arrudas argument once again, we can show the following: even by
adopting the Separation Axiom of ZF instead of the unrestricted version of the
Comprehension Axiom, if the universal set V were definable in a paraconsistent set
system without strong negation but with the properties of the system above, then
the set
C = {x : (x V ) ((x x) )}

would be definable in that system, by the Separation Axiom of ZF. But the latter set
is equal to the set B , by extensionality and by the fact that  (x V ) for every x.
This means that the set B would be definable, and so the system would be trivial:
the proof is analogous to the one given above.

8.3 Some Extensions of ZFmbC

The next, natural step, is to consider extensions of ZFmbC by taking stronger LFIs
and appropriate axioms for the consistency operator C for sets. Being axiomatic
extensions of QmbC, all of the systems considered here will satisfy the deduction
meta-theorem, DMT, as stated in Theorem 7.2.5 and in Corollary 7.2.6.
8.3 Some Extensions of ZFmbC 355

The first system to be defined is based on mCi (recall Definition 3.5.12), by adding
also some axioms for C which are coherent with the rationale behind mCi.

Definition 8.3.1 (ZFmCi) The system ZFmCi is obtained from ZFmbC by adding
the following axioms, for n 0:

( ) (ci)

n+2 n
(n )
x(C(x) (x x)) (Cons3)
x(C(x) (x x)) (Cons4)


The first two axioms transform the underlying paraconsistent logic mbC into the
stronger logic mCi, in which the inconsistency operator can be defined by =
def

and so is equivalent to (recall Chap. 3, Sect. 3.5). Axioms (Cons3) and

(Cons4) intend to strengthen C, the consistency operator for sets, in order to obtain
an inconsistency operator for sets I dual to C, defined by I (x) = C(x).
def

Lemma 8.3.2 In mCi it holds:  ( ) .

Proof It is a consequence of Propositions 3.2.3 and 3.1.10(1), and the fact that mCi
extends mbCci. 

Proposition 8.3.3 In ZFmCi it holds that:

(i)  C(x) (x x);
(ii)  (x x) C(x);
(iii)  C(x) (x x);
(iv)  C(x) (x x);
(v)  ( ).

Proof (i) Assume C(x), (x x) and (x / x). By Proposition 8.2.3(i) it follows

that C(x) and (x x) and so C(x), (x x) and (x x), by axiom (AxEq1). By
(Cons1) and MP we get (x x), (x x) and (x x) and then it follows , by
(Ax10). By DMT and definition of it follows that C(x)  ((x x) (x / x))
and then C(x)  (x x), by Lemma 8.3.2. The result follows by DMT.
(ii) Assume (x x). By (ci) and MP it follows that (x x) (x / x). Then
(x x), by Proposition 8.2.3(i). Therefore (x x) (x x), by axiom (AxEq1)
and so (x x), by Proposition 2.3.3(1). Then by (Cons2) and MP it follows that
C(x). The result follows by DMT.
(iii) Assume C(x). Then (x x), by (Cons4), and so (x x) (x / x), by
(ci). By Proposition 8.2.3(i) it follows that (x x), as desired.
(iv) From C(x) it follows that (x x) (x / x), by the proof of item (iii). From
this we get (x x), by the properties of conjunction.
(v) It follows from item (i) and from Proposition 8.2.3(vi). 
356 8 Paraconsistent Set Theory

Proposition 8.3.4 In ZFmCi it holds that:

(i)  (C(x) C(x));
(ii)  (C(x) C(x));
(iii)  C(x).

Proof (i) Assume C(x). Then (x x), by (Cons1). On the other hand C(x) 
(x x), by (Cons3). Thus C(x) C(x)  (x x) (x x). But
(x x) (x x) implies , since  holds in mCi for every , by
Theorem 3.5.15(i). The result holds by the deduction meta-theorem DMT and the
definition of .
(ii) If follows from (i) and Proposition 2.4.8(iii).
(iii) If follows from (i) and Lemma 8.3.2. 

By recalling that I (x) = C(x), it is easy to prove the following:

def

Proposition 8.3.5 In ZFmCi it holds that :

(i)  x(I (x) (x x));
(ii)  x(I (x) (x x));
(iii)  x(I (x) (x x));
(iv)  x(I (x) (x x));
(v)  x(I (x) ((x x) (x x)));
(vi)  x(I (x) ((x x) (x
/ x)));
(vii)  x(I (x) (x x));
(viii)  x(I (x) (x x)).

Recall that the logic Ci is obtained from mCi by adding the axiom

(cf)

or, equivalently, by adding to mbC the axioms (ci) and (cf) (see Definition 3.5.17).
This suggests the definition of the following system:

Definition 8.3.6 (ZFCi) The system ZFCi is obtained from ZFmCi by adding
axiom (cf) or, equivalently, the system obtained from ZFmbC by adding axioms
(ci), (cf), (Cons3) and (Cons4). 

On the other hand, recall that the system Cil is obtained from Ci by adding the
axiom

( ) (cl)

(see Definition 3.5.17). Thus, consider the following extension of ZFCi, coherent
with Cil:
Definition 8.3.7 (ZFCil) Let ZFCil be the system obtained from ZFCi by adding
axiom (cl) plus
8.3 Some Extensions of ZFmbC 357

x(((x x) (x x)) C(x)) (Cons5)

x(((x x) (x
/ x)) C(x)) (Cons6)

Proposition 8.3.8 In ZFCil it holds that:
(i)  x(C(x) ((x x) (x x)));
(ii)  x(C(x) ((x x) (x / x)));
(iii)  x(C(x) (x x));
(iv)  x(C(x) (x x)).
Proof (i) This is immediate from Proposition 8.2.6(iii) and axiom (Cons5).
(ii) It follows from Proposition 8.2.6(iv) and axiom (Cons6).
(iii) From (x x) it follows that ((x x) (x x)), by Proposition 2.3.3(2),
and so we get C(x), by (Cons5). The converse is simply (Cons1 ).
(iv) From (x x) it follows that ((x x) (x / x)), by Proposition 2.3.3(2).
Then it follows C(x), by (Cons6). The converse is a consequence of Proposition
8.3.3(i). 
Remark 8.3.9 (1) At this point, it can be interesting to summarize the list of equiv-
alences relating the consistency predicate C, the identity predicate and the mem-
bership relation proved up to now in ZFCil:

C(x) (x x) C(x) ((x x) (x x))

C(x) (x x) C(x) ((x x) (x
/ x))
C(x) (x x) C(x) (x x)
C(x) (x x) C(x) (x x)

Additionally,  C(x) (x / x) but the converse should not be valid, assuming the
non-triviality of ZFCil. Some of the equivalences above should not be valid in the
weaker systems ZFmbC and ZFmCi, assuming that they are non-trivial.
(2) The fact that (x x), the consistence of (x x), is equivalent to x be a consistent
set, expressed by C(x) (and, in particular, the fact that ( ) is a consistent
sentence) deserves some explanation. As discussed in the previous chapters, in the
realm of LFIs, the consistency of a sentence means that it is reliable, or it has
a classical or expected behavior (with respect to the principle of explosion). In
semantic terms, it means that is true whenever the truth-values of and are
different (it holds from mbCciw on, see Sect. 3.1). In particular, if is always true
and its negation is always false, or if is always false, then it is a consistent sentence.
This is why the assumption C(x) of consistency of a given set x forces to have (x x)
as being a consistent claim: it must be always false, under such assumption.
As we can see, the expedient of separating sets and sentences into consistent and
inconsistent, and moreover of using very cautious logics (that is, logics with weaker
negations, but yet endowed with semantics in the best sense), permits us a finer
control of the most basic reasoning underlying the foundations of (even inconsistent)
mathematics.
358 8 Paraconsistent Set Theory

A fundamental question to be answered is the following: Is ZFCil (and therefore

its subsystems) non-trivial? Of course the answer to this question is related to the
consistency of ZF, as we will show below.
Recall that a theory  in a logic L with an explosive negation is inconsistent
if   and   for some formula ; otherwise it is consistent.3 On the other
hand, a theory  in a logic L is trivial if   for every formula ; otherwise, it
is non-trivial. In case is explosive in L , then  is inconsistent if and only if it is
trivial.
As a result, the question of the consistency of ZF (seen as a theory in first-order
classical logic, where the negation is explosive) is equivalent to the statement of
its non-triviality.
On the other hand, a theory based on a paraconsistent logic can be inconsistent
but non-trivial. As such, the relevant question about ZFCil is to determine its non-
triviality (seen as a theory of first-order Cil) instead of its inconsistency. As we shall
see in Corollary 8.3.13, the non-triviality of ZFCil depends on the consistency of
ZF, and so paraconsistent set theory based on LFIs is no more unsafe than the set
theory ZF. This will be established by means of the following technical result, in the
same vein of the famous Gdel embedding of classical logic into intuitionistic logic:
Proposition 8.3.10 Let t be a mapping from the language of ZFCil to the language
of ZF recursively defined as follows:
t (x y) = (x y);
t (x y) = (x y);
t (C(x)) = (x x);  
t ( ) = (t () t ()) for , , ;
t () = t ();
t () = (t () t ());
t (x) = x t (); and
t (x) = x t ().
Then t is an embedding of ZFCil into ZF, that is: for every formula in the language
of ZFCil, if  in ZFCil then  t () in ZF.
Proof It is sufficient to observe that, if is an instance of an axiom of ZFCil, then
t () is derivable in ZF. On the other hand, any application of an inference rule in
ZFCil is transformed by t into an application of the same inference rule in ZF. Thus,
any derivation of obtained in ZFCil can be transformed into a derivation of t ()
in ZF. Details are left to the reader. 
Corollary 8.3.11 Let t be the mapping from the language of ZFCil to the language
of ZF defined in Proposition 8.3.10. Then t is a translation from ZFCil to ZF with
respect to closed sentences. That is: for every set of closed sentences  {} in the
language of ZFCil, if   in ZFCil then t[]  t () in ZF.

3 Weare using here the traditional terminology. If the negation is not explosive then it would be
more appropriate to speak of contradictory and non-contradictory theories, respectively.
8.3 Some Extensions of ZFmbC 359

Proof Observe that ZFCil is finitary and ZF is monotonic. By the Deduction meta-
theorem DMT for sentences, which holds both in ZF and in ZFCil (see Corol-
lary 7.2.6), the result is an easy consequence of Theorem 2.4.2. 

Theorem 8.3.12 If ZF is consistent then ZFCil cannot prove any contradiction.

That is: is not provable in ZFCil, for every .

Proof Suppose that is derivable in ZFCil for some formula . By Proposi-

tion 8.3.10, the formula t () is derivable in ZF. But this implies that t ()t ()
is derivable in ZF and so ZF is inconsistent. 

Corollary 8.3.13 If ZF is consistent then ZFCil is non-trivial.

Proof Suppose that ZF is consistent. By Theorem 8.3.12, no contradiction can be

proved in ZFCil. Then ZFCil is non-trivial, since some formula cannot be derivable
in it. 

Assuming the consistency of ZF, Theorem 8.3.12 turns evident a basic feature
of LFIs: contradictions are not proved in these systems, but they can be used as
hypothesis without trivializing.
Consider now the following axioms that could be added to ZFCil:

yx(x y (x)) (cNC1)

x(x) yx(x y (x)) (cNC2)
C(y) x(x y C(x)) (crp)

where (cNC) stands for consistent Naive Comprehension and (crp) stands for
consistency for sets retropropagation. (In (cNC1) and (cNC2) it is assumed that y
does not occur free in (x).) The basic idea is that either (cNC1) or (cNC2) could
make it possible to recover the main constructions of [16]. However, in the presence
of the embedding t from Proposition 8.3.10, it is easy to see that t maps any instance
of (cNC1) into a sentence that implies the existence of the universal set V in ZF.
Indeed, let be an instance yx(x y (x)) of (cNC1). Then t () is the
formula yx(x y (x)), where (x) is the formula (t ((x)) t ((x))).
Since  (x) in ZF then t ()  yx(x y). That is, the universal set V would be
definable in ZF from t ().
By its turn, any instance of (cNC2) is mapped by t into a sentence that implies an
instance of the unrestricted version of the Comprehension Axiom, and thus able to
obtain, in particular, the strong Russell set. Indeed, let be any instance x(x)
yx(x y (x)) of (cNC2). Then t () is the formula x (x) yx(x
y t ((x))), where (x) is defined as above. Suppose that neither nor C occur
in (x). Then, clearly t () = . Since  x (x) in ZF then t ()  yx(x y
(x)) in ZF. That is, the unrestricted version of the Comprehension Axiom would
be obtained in ZF from such formulas t (), which would produce the existence of
the strong Russell set and the universal set V .
360 8 Paraconsistent Set Theory

On the other hand, t maps any instance of (crp) into a sentence derivable in ZF.
Indeed, if is an instance C(y) x(x y C(x)) of (crp) then t () is the
formula (y y) x(x y (x x)), which is clearly provable in ZF.
Accordingly, axiom (crp) is the only acceptable one to be added to ZFCil, among
the three axioms above, as far as we maintain the mapping t as an embedding that
guarantees equi-non-triviality.
Another question concerns the existence of the universal set of all consistent sets,
in the form
cons = {x : C(x)}.

Intuitively, this would constitute a kind of universal set. Moreover, if cons is legit-
imated, then using again the embedding t the universal set V would be legitimated
in ZF. Indeed, let be the formula yx(x y C(x)), which guarantees the
existence of the set cons. Then t () is the formula yx(x y (x x)), which
establishes the existence of the universal set V in ZF.
Observe that, in the presence of (Cons0), it follows that C(cons) and therefore
cons cons, trivializing the system in the presence of the Regularity Axiom (see
Proposition 8.2.3(ii)).

8.4 Inconsistent Sets and Proper Classes

In 1899, Cantor discovered, or became aware of,4 the famous paradox about the
cardinal number of the set of all sets. In intuitive terms, it may be described as
follows: on the one hand, there must be a greatest possible cardinal, call it C. On
the other hand, if this cardinal number is a set, the cardinal number of the power set
of C is strictly larger than the cardinal number of C (this is now known as Cantors
theorem). Hence, either this cardinal number does not exist, or it is not a set. This
paradox, together with Burali-Fortis paradox of 1897 (which proves that naively
constructing the set of all ordinal numbers leads to an antinomy) led Cantor to
formulate a concept called limitation of size according to which collections such
as the one of all ordinals, or of all sets, was an inconsistent multiplicity, too large
to be a set. Such collections later became known as proper classes. In the words of
Dauben in [19]:
Anything that was too large to be comprehended as a well defined, unified, consistent set was
declared inconsistent. These were absolute collections, and lay beyond the possibility of
mathematical determination. This, in essence, is what Cantor communicated first to Hilbert
in 1897, and somewhat later to Dedekind in his letters of 1899.

According to [19], Cantor himself believed, in his early period, that the idea
of the actual infinite could not be consistently formulated and so had no place in

4 Itis worth noting that in [18] it is shown that Burali-Fortis paradox, generally regarded as the
first of the set-theoretical paradoxes, was neither created by Burali-Forti nor by Cantor. It arose
gradually and only acquired its contemporary form in the hands of Bertrand Russell in 1903.
8.4 Inconsistent Sets and Proper Classes 361

rigorous mathematics, and he took seriously Kroneckers criticisms,5 which were

echoed in his major work on set theory, [20]. This work was a truly mathematical-
philosophical investigation of the infinite, as the title indicates, and was where he
declared his famous pronouncement that the essence of mathematics is exactly its
freedom. Such a sense of freedom, together with his method of using in a positive
way what seemed to be paradoxical, as explained by Kanamori in the quotation
below from [21], is what raises doubts whether an inconsistent multiplicity would
coincide with what we today know as a proper class:
Cantor in 1899 correspondence with Dedekind considered the collection  of all ordinal
numbers as in the Burali-Forti Paradox, but he used it positively6 to give mathematical
expression to his Absolute. He defined an absolutely infinite or inconsistent multiplicity
as one into which  is injectible, and proposed that these collections be exactly the ones
that are not sets. He would thus probe the very limits of sethood using his positive concept
of power!

Maddy, in her search for a realistic theory of sets and classes (see [22]), tries to
provide a theory of both sets and classes (along the lines of J. Knig), claims for a
theory where:
(1) classes should be real, well-defined entities;
(2) classes should be significantly different from sets
Maddy is not alone in questioning whether the Fregean formulation of naive
set theory (which was later refuted by Russells paradox) would really be a faithful
interpretation of the Cantorian conception of sets. As Weir (see [23], p. 766) observes:
. . .it may well be seriously mistaken to think of Cantors Mengenlehre as naive. . .

As we know today, in ZF set theory the notion of class is metalinguistic (actu-

ally, ZF does not refer to classes at all), whereas in other set theories, such as von
Neumann-Bernays-Gdel (NBG), the notion of class is axiomatized and inter-
preted as an entity that is not a member of any other entity. Indeed, in NBG classes are
the basic objects, and a set is defined as a class that is an element of some other class.
Like NBG, Morse-Kelley (MK) set theory admits proper classes as basic objects, but
whereas NBG is a conservative extension of ZF, MK is strictly stronger than both
NBG and ZF. Other set theories, such as New Foundations (NF) and the theory of
semisets, treat the relationship between sets and classes in still different ways. The set
theoretical proposals of P. Finsler, centered around his concept of circle-free sets,
are regarded in principle as incoherent (see [24]). The idea was reworked (apparently
independently) by W. Ackermann, and makes for a quite different theory from the
standard one (it is considered to be the first genuine alternative set theory). Acker-
manns set theory, as explained in [24], is a theory of classes in which some classes
are sets, and indeed the notion of set is indefinablethere is no simple definition of

5 L.Kronecker, a prominent German mathematician who had been one of Cantors teachers, even
attacked Cantor personally, calling him a scientific charlatan a renegade and a corrupter of the
youth! .
6 Emphasis in the original.
362 8 Paraconsistent Set Theory

which classes are sets. It turns out, however, to be essentially the same theory as ZF.
Coming now to our question, whether inconsistent sets can be equated (or at least
somehow regarded as) proper classes, we see no reason why Cantors inconsistent
multiplicities could be so simply reduced to the notion of proper classes. Firstly,
as it is clear from the above overabundance of approaches to sets and classes, there
is no unique sense of what proper classes would be. One criticism might be that
there is even less agreement on what an inconsistent collection would be, but we
are not defining what an inconsistent collection israther, we are postulating their
rational possibility (which is much less than their existence), and showing how they
can be handled from a coherent logical standpoint. Our approach views inconsistent
situations as possible evidence that something may have gone wrong, as a sign that
contradictions may appear, but not as seal of condemnationonly in this way, we
believe, can one explore positively the very limits of sethood, as Cantor did.
However, even if not coincident, the notions of being consistent (as formalized
here) and being a set are related: in fact, let set (x) = y(x y) be the usual notion
def

of sethood in a theory of classes enriched with a consistency predicate C(x) as the

one proposed in this chapter. In case  set (x) C(x), then the following will be
obtained: x x  set (x)  C(x) (x x)  and so x x cannot be satisfied for
any entity x. Of course this is the case in classical set (or class) theories because of the
Regularity Axiom, but in a paraconsistent set theory (x x) and (x / x) should not
be necessarily excluded. Therefore the non-classical character of our theory would
vanish by requiring the relationship between sethood and consistency in the above
way. An interesting point is that the converse relation can be added to the theory
without any further problems:

C(x) set (x) (cs)

(consistency as sethood axiom).

On the other hand, if class(x) = y(x y) denotes the proper-classhood
def

predicate as usual, it would be possible to consider the following principle:

class(x) C(x) (pci)

(proper-classhood as inconsistency axiom)

together with (or as an alternative to) axiom (cs).
Our theoretical plan for set theory by predicating on consistency has no need
to overpopulate the usual formal ontology of sets, so we do not need any kind of
strange new objects (aside from the usual ones of ZF). If we consider what is called
dialetheia (see [25]) as a sentence which is both true and false, our theory is even
free of dialetheias since in our setting a sentence can be seen as true and its negation
as also true (at least provisionally) without either of them falling into the anathema
of falsity. But we might purposedly add, for instance, what may be called the Russell
set axiom, and keep harvesting its consequences without falling into triviality:
8.4 Inconsistent Sets and Proper Classes 363

x((x x) (x
/ x)) (Russ)

From Proposition 8.2.6 it follows that the Russell sets are inconsistent:
Corollary 8.4.1 Let R be a Russell set, that is,  (R R) (R
/ R). Then
 C(R).
Postulating the Russell set axiom (Russ) is not really necessary for the devel-
opment of a paraconsistent set theory. However, we can indulge in a dialetheistic
assumption if we want to enrich set theory towards an antinomic extension. Although
the postulation of a Russell set takes us closer to Routleys set theory, we could as well
remove this axiom from our systems while tolerating inconsistent sets as premises.
In this sense, our approach is not metaphysically committed. According to da Costa
in [7]:
The main concern to paraconsistent set theory is not to make possible the existency, and
thereby the investigation of some sets which can cause trouble in naive set theory, such as
Russells set, Russells relations and the set of all non-k-circular sets (k = 1, 2, . . .). On the
contrary, the most important characteristic of paraconsistent theories is that they allow us to
handle the extensions of inconsistent predicates which may exist in the real world or are
inherent in some universes of discourse in the fields of science and philosophy.

On the other hand, the inclusion of a Russell set in our systems would require
a new proof of non-triviality of the system relative to the consistence of ZF (see
Corollary 8.3.13). Indeed, if ZF is consistent then the mapping t defined in Propo-
sition 8.3.10 would no longer be an embedding between the system extended with
(Russ) into ZF. In fact, t maps the axiom (Russ) into a sentence not derivable in ZF
(see Theorem 8.3.12).
Related to the question of postulating the existence of some sets, observe that the
existence of the strong empty set such that  x(x ) is guaranteed by the
axioms of ZFmbC, and in fact this set plays an important role in the theory. What
about the existence of the weak empty set? It would be a set such that  x(x ),
that is, satisfying  x(x / ). Since it is not obvious that the existence of such a
set can be proved even in ZFCil, its existence should be postulated by an specific
axiom:
yx(x / y) (WES)

The convenience of postulating the existence of such an object should be ponder-

ated. It is clear that, as happens with Russell sets, its existence is not necessary, nor
directly relevant for the development of our paraconsistent set theory.

8.5 On Models

Although models of (classical) set theory can be seen as special cases of models of
a first-order language, some models can also be built using proper classes. But what
would be the models of ZFCil? In what follows, we explore some preliminary ideas
364 8 Paraconsistent Set Theory

on this question, with the understanding that there is much to be done. Along this
section, the symbol  will be used for denoting membership in the metalanguage,
while will denote the membership relation symbol in the object (first-order)
language.
Libert considers in [14] a class of structures suitable for paraconsistent set theories.
Such interpretation structures are pairs of the form M = M; []M  such that M is
a nonempty set and []M : M p (M) is a function, where p (M) = {X, Y  :
def

X, Y M and X Y = M}. The equality symbol is interpreted in the standard

way. For every b M, the pair [b]M = [b]+ +
M , [b]M  is such that [b]M = {a M :

a M b} is the positive extension of b, while [b]M = {a M : a / M b} is its
negative extension. A strong extensionality is required, namely that the function []M
be injective, and so for every a, b M it holds that: [a]+ +
M = [b]M and [a]M = [b]M

iff a = b. This leads naturally to a 3-valued paraconsistent logic. It is noteworthy to

stress the similarity between Liberts interpretation structures and the partial models
underlying the theory of quasi-truth, analyzed in Sect. 7.9 of Chap. 7.
It is also possible to think of a theory of quasi-sets (see for instance [26, 27])
as a theory of indistinguishable objects, motivated mainly by the assumption that
one cannot meaningfully apply the notion of identity to quantum particles. In [28]
other, more general, models are proposed (expanding on [15]), in such a way that
any formula defines a set (thus recovering the whole theory of Frege), but Russells
paradox is tolerated on account of an underlying paraconsistent three-valued logic
(equivalent to LFI1 and so to J3 and to LPT1).
From an algebraic perspective, a generalization of Boolean-valued models of set
theory by means of a 3-valued algebra was proposed in [29]. Such algebra is a model
of a certain paraconsistent set theory that validates the negation-free fragment of ZF.
As mentioned in Chap. 4, this 3-valued algebra originates a 3-valued paraconsistent
logic called PS3 which again coincides with J3, LFI1 and MPT, among others
(see Remark 4.4.12).
Such observations make it very natural to consider, generalizing the classical state
of affairs, special cases of models of first-order paraconsistent language as models of
ZFCil, and indeed (though we shall not go into details here) exploring such models
seem to be a promising approach to understanding the distinction between consistent
and inconsistent sentences, the distinctions between consistent and inconsistent sets,
and the relations between inconsistent sentences and inconsistent sets.
Another point which would deserve attention is the possibility of studying a variant
of our paraconsistent set theory in the direction of the non-well-founded set theories,
obtained by replacing the Regularity Axiom (also known in this context as the axiom
of foundation) by one or more axioms implying its negation. The resulting would
certainly give an interesting theory of circular sets, but nothing has been done on this
regard yet.
Part of our interest is to compare our approach with several other proposals. By
admitting that sets, as well as sentences, can be inconsistent, and that only consistent
and contradictory objects may cause triviality, our view offers support to proposals
such as the one by Byers [30], where thinking about math requires creativity that is
8.5 On Models 365

promoted, rather than neglected, using forms of thought connected to contradiction.

Aside from Russells antinomy, it is natural to extend our considerations on Berrys
and to the Burali-Fortis antinomies.
Moreover, a criticism that can be raised against category theory is that most of its
mechanisms are just inspired on ZFC set theory. There is no reason not to consider
a wider version of category theory that would include inconsistent situations, such
as (simultaneously) commuting and non commuting diagrams, and the like.
Inconsistent mathematics (and, in more general terms, even paraconsistent cate-
gory theory) may have instigating applications to quantum mathematics and to the
study of quantum measurement, in particular to dynamic systems in which there are
discontinuous jumps, as suggested by Mortensen in [31].
It is hard to conceive of mathematics today without the idea of completed infi-
nite, and if it had not been stated by a mathematician of the intellectual stature of
Kronecker, we would consider as pure insanity his claim (see e.g. [19]) that:
I dont know what predominates in Cantors theory philosophy or theology, but I am sure
that there is no mathematics there.

Nobody, however, can be sure where is there new mathematics, or what kind of
exciting mathematics will be revealed by yet unthought entities as contradictory and
inconsistent mathematical objects.
For a concrete example in the somewhat recent history of mathematics, it is
well known among mathematicians that important methods in analysis and for the
search for solutions to certain differential equations are differentiation and Fourier
transformation. Due to the fact that not all functions are differentiable or have a
Fourier transform, the theory of distribution famously proposed by Schwarz (see
[32]) and subsequentlty further developed, embeds the theory of classical functions
in a larger class of objects, the so called distributions (or generalized functions)
which give a well-defined meaning to objects such as the Dirac delta function.7
Generalized functions, however, exhibit some contradictory features when
regarded from the point of view of classical analysis. Physicists and mathemati-
cians choose to tolerate such self-contradictory objects, as they have greater utility
in certain cases. Better for the human intellect and for the development of science.

References

1. Frege, Gottlob. 1982. Grundgesetze der Arithmetik, Band I/II (in German). Verlag Hermann
Pohle, Jena, 1893. Partial translation of Band I by Montgomery Furth as The basic laws of
arithmetic. Berkeley: University of California Press.
2. Crossley, John N. 1973. A note on Cantors theorem and Russells paradox. Australian Journal
of Philosophy 51(1): 7071.
3. Carnielli, Walter A., and Marcelo E. Coniglio. 2016. Paraconsistent set theory by predicating
on consistency. Journal of Logic and Computation 26(1): 97116. doi:10.1093/logcom/ext020.

7 L. Schwarz was awarded the Fields Medal in 1950 for his work on the theory of distributions.
366 8 Paraconsistent Set Theory

4. Dauben, Joseph W. 1990. Georg Cantor: His Mathematics and Philosophy of the Infinite.
Princeton University Press.
5. da Costa, Newton C.A. 1993. Sistemas formais inconsistentes (Inconsistent formal systems,
in Portuguese). Habilitation thesis, Universidade Federal do Paran, Curitiba, Brazil, 1963.
Republished by Editora UFPR, Curitiba, Brazil.
6. Arruda, Ayda I. The Russell paradox in the systems NFn. In [Arruda, Ayda I., Newton C. A.
da Costa, and Antonio M. A. Sette, eds. 1980. Proceedings of the Third Brazilian Conference
on Mathematical Logic, Recife 1979. Sociedade Brasileira de Logica, Campinas], pp. 112.
7. da Costa, Newton C.A. 1986. On paraconsistent set theory. Logique et Analyse, 29(115): 361
371.
8. Arruda, Ayda I., and Diderik Batens. 1982. Russells set versus the universal set in paraconsis-
tent set theory. Logique et Analyse 25(98): 121133.
9. Church, Alonzo. 1974. Set theory with a universal set. In Proceedings of the Tarski Symposium,
ed. by Leon Henkin, 297308. Providence, RI, USA: American Mathematical Society.
10. da Costa, Newton C.A. 2000. Paraconsistent mathematics. In Frontiers of Paraconsistent Logic,
ed. by Diderik Batens, Chris Mortensen, Graham Priest, and Jean-Paul Van Bendegem, 165
180. Hertfordshire: Research Studies Press.
11. da Costa, Newton C.A., Jean-Yves Bziau, and Otvio Bueno. 1998. Elementos de Teoria
Paraconsistente de Conjuntos (Elements of Paraconsistent Set Theory, in Portuguese). Coleo
CLE, vol. 23. Campinas, Brazil: CLE-Unicamp.
12. Routley, Richard. 1977. Ultralogic as universal? Relevance Logic Newsletter 2:5090 and
138175, reprinted in [255].
13. Brady, Ross T. 1989. The non-triviality of dialectical set theory. In Paraconsistent Logic:
Essays on the Inconsistent, ed. by Graham Priest, Richard Routley, and Jean Norman, 437
471. Philosophia Verlag: Mnchen.
14. Libert, Thierry. 2005. Models for a paraconsistent set theory. Journal of Applied Logic 3(1):
1541.
15. Hinnion, Roland. 1994. Naive set theory with extensionality in partial logic and in paradoxical
logic. Notre Dame Journal of Formal Logic 35(1): 1540.
16. Weber, Zach. 2010. Transfinite numbers in paraconsistent set theory. Review of Symbolic Logic
3(1): 7192.
17. Verdee, Peter. 2013. Strong, universal and provably non-trivial set theory by means of adaptive
logic. Logic Journal of the IGPL 21(1): 108125.
18. Moore, Gregory H., and Alejandro Garciadiego. 1981. Burali-Fortis paradox: a reappraisal of
its origins. Historia Mathematica 8(3): 319350.
19. Dauben, Joseph W. 2005. The Battle for Cantorian Set Theory. In Mathematics and the Histo-
rians Craft: The Kenneth O. May Lectures, ed. by Michael Kinyon and Glen van Brummelen,
221241. New York: Springer, Canadian Mathematical Society Books in Mathematics.
20. Cantor, Georg. 1996. Grundlagen einer allgemeinen Mannigfaltigkeitslehre. Ein mathematisch-
philosophischer Versuch in der Lehre des Unendlichen (in German). B.O. Teubner, Leipzig,
1883. Translated to English as: Foundations of a general theory of manifolds: A mathematico-
philosophical investigation into the theory of the infinite. In From Kant to Hilbert: A Source
Book in the Foundations of Mathematics, ed. William B. Ewald, vol. 2. Oxford: Oxford Uni-
versity Press.
21. Kanamori, Akihiro. 1996. The mathematical development of set theory from Cantor to Cohen.
The Bulletin of Symbolic Logic 22(1): 171.
22. Maddy, Penelope. 1983. Proper classes. The Journal of Symbolic Logic 48(1): 113139.
23. Weir, Alan. 1998. Naive set theory is innocent!. Mind 107(428): 763798.
24. Holmes, M. Randall. 2014. Alternative axiomatic set theories. In The Stanford Encyclopedia
of Philosophy ed. by Edward N. Zalta. Spring 2014.
25. Priest, Graham, Koji Tanaka, and Zach Weber. 2013. Paraconsistent logic. In The Stanford
Encyclopedia of Philosophy, ed. by Edward N. Zalta, Fall 2013.
26. Bueno, Otvio. 2000. Quasi-truth in quasi-set theory. Synthese 125(12): 3353.
References 367

27. da Newton, C.A. 2007. Costa and Dcio Krause. Logical and philosophical remarks on quasi-set
theory. Logic Journal of the IGPL 15(56): 421431.
28. Esser, Olivier. 2003. A strong model of paraconsistent logic. Notre Dame Journal of Formal
Logic 44(3): 149156.
29. Lwe, Benedikt, and Sourav Tarafder. 2015. Generalized algebra-valued models of set theory.
The Review of Symbolic Logic 8(1): 192205. doi:10.1017/S175502031400046X.
30. Byers, William. 2007. How mathematicians think: Using Ambiguity, Contradiction, and Para-
dox to Create Mathematics. Princeton University Press.
31. Mortensen, Chris. 2013. Inconsistent mathematics. In The Stanford Encyclopedia of Philoso-
phy, ed. by Edward N. Zalta. Spring 2013.
32. Schwartz, Laurent. 19501951. Thorie des Distributions (in French). Paris: Hermann & Cie.
Chapter 9
Paraconsistency and Philosophy of Science:
Foundations and Perspectives

9.1 An Epistemological Understanding of Paraconsistency,

and Its Significance for Science

The previous chapters have been devoted to establishing several cornerstones of LFIs,
from basic propositional logics and their extensions to quantified generalizations, to
new set theories and their potentialities. The subtle relationship between the para-
consistent and the fuzzy paradigms have been investigated by means of combining
fuzzy logics with consistency operators, as has the relationship between the paracon-
sistent and the intuitionistic paradigms. Important issues concerning algebraizability,
semantics and proof theory of paraconsistent logics have also been addressed.
In all such approaches the importance of the consistency operator (or better oper-
ators, since each logic has its own concept of consistency) has been duly emphasized.
We have already presented in Chap. 1 some philosophical motivations for developing
paraconsistent logics in general, and in particular LFIs. But there are some aspects of
paraconsistency that still deserve further attention: the potential of LFIs for applica-
tions in engineering, computer science, probability theory and linguistics, to mention
only a few strategic areas.
This chapter takes up again, even if briefly, some philosophical problems related to
paraconsistency. We survey here a view of the nature of contradictions that provides
a philosophical justification for LFIs, and for paraconsistent logics in general. Such
a view, as has been said, may be held independent of ones beliefs concerning real
contradictions, and therefore is not committed to dialetheism, the thesis according
to which there are true contradictions. We argue that contradictions can be naturally
understood as an epistemic phenomenon, a position that cannot be maintained if
contradictions are characterized in ontological terms.
Concerns about contradictions lead naturally to more central questions about
the nature of logic: what are the principles of logic about? Are they about reality,
thought or language? The fundamental question is whether logic has an ontological,
an epistemological or a linguistic character.

Springer International Publishing Switzerland 2016 369

W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5_9
370 9 Paraconsistency and Philosophy of Science: Foundations

The view that logic is primarily about language is related to the idea that
metaphysical problems should be avoided. Concentrating on the study of language,
of course, is a way, above all, of avoiding ontological issues. But the linguistic con-
ception of logic has still independent interest, and indeed questions about the nature
of consistency and inconsistency require attentionin linguistics there seem to be
several types of inconsistency, as we shall briefly discuss in Sect. 9.2.7. But we do not
here discuss issues of the linguistic character of logic in any depth. The main reason
for restricting attention here to the epistemological and the ontological aspects of
logic is that the central issue, when it comes to paraconsistent logic, has not so much
to do with language but, rather, whether or not there are contradictions in reality.
In todays philosophy it is almost a truism to repeat that contradictions appear
in several situations: pieces of information contradict each other, norms, laws and
regulations conflict, databases clash and scientific theories collide.
The present chapter discusses the occurrence of contradictions in scientific the-
ories and is concerned to understand them as possessing an epistemological, rather
than an ontological, character. In Chap. 1 we present an interpretation of LFIs
(and of paraconsistent logics in general) as logics having epistemic character and, at
the same time, having a descriptive power in the sense that they permit us to describe
and represent correct reasoning in real-life situations. This summarizes the position
with respect to paraconsistency endorsed in this book.
If contradictions are epistemological (and do not belong to reality), as we have
extensively argued, the rejection of the Principle of Explosion by paraconsistentists is
analogous to the rejection of law of excluded middle by intuitionists. In both cases, the
formal system may be understood as having epistemological character that combines
a descriptive with a normative approach to logic. This epistemological approach, we
claim, is of natural and immediate interest for a correct understanding of the modus
operandi of the sciences.
This does not mean, of course, that ontologically contradictory scientific phe-
nomena or objects, say in Physics, Chemistry, or Biology could not existwhat we
mean is that our approach to paraconsistency does not need this presupposition: if
there were real contradictions in certain sciences, then those sciences would have to
develop their own specialities, say, physics of contradictory particles or genetics
of contradictory organisms. But this is not what we see around us.
What we witness in the absolute majority of cases, without reasonable doubt,
are contradictory theories, or at least contradictory beliefs, but not contradictory
phenomena or objects. In this way, a useful, applicable account of contradictions
involving entities such as propositions, beliefs, theories or hypothesis may start from
the notion of evidence, a notion weaker than truth. In this way, if one knows that a
proposition A is true, one has evidence that A is true, but not the converse. Evidence
for A should be understood as reasons for a justified belief that A is true (see [1])
in the sense that evidence that A is true means that there are reasons justifying
the belief that A is true. Some examples of real situations will help to illustrate this
approach.
9.2 Consistency and Contradiction in Scientific Theories 371

9.2 Consistency and Contradiction in Scientific Theories

It is well-recognized that scientific research deals almost routinely with

contradictionsexamples are abundant, from the disputes in the beginning of Cal-
culus involving Isaac Newton and Gottfried Wilhelm Leibniz, to Bohrs theory of the
atom and classical electrodynamics. Well-known cases are W. Paulis derivation of
a contradiction from Bohrs postulates and from P. Ehrenfests adiabatic principle,
and the several kinds of alleged contradictions in Newtonian cosmology. There are,
however, different ways in which a scientific theory might be regarded as incon-
sistent or contradictory, as analyzed in [2]: some assumptions of the theory may
conflict with other parts of the same theory, or with other well-established theories.
Second, the theory might be explicitly contradictory. Third, the theory might use
some inconsistent mathematical tools. In physics, for instance, it sometimes occurs
that two different theories, coherent in and of themselves, yield contradictions when
put together (see e.g. Chap. 5 of [3], and [4]). Hence, since contradictions imply
triviality in the framework of classical logic, it seems obvious that scientists need a
non-explosive logic in order to give a global account of them. It is worth noting in
this regard that when the physicist considers two incompatible theories, even if it is
a matter of revising one of them, (s)he is already reasoning in a paraconsistent way,
since (s)he limits the application of the principle of explosion.
Before addressing the question of the nature of contradictions in scientific theories,
a brief excursion into Kantian negative results with respect to human knowledge
will be helpful.

9.2.1 The Heritage of Kant

Kant, more than 200 years ago, presented fundamental insights about the limits of
human reason that we essentially endorse. Kants aim was to provide a foundation
for metaphysics, separating, so to speak, the good metaphysics from the bad. In brief,
he concluded that human reason has limits that should not be surpassed. When such
limits are transcended, we may get into error and contradiction, which is a sign that
something has gone wrong. There is an unbridgeable distinction between reality as
it is (things-in-themselves) and reality as it presents itself to us by means of our
experience (phenomena). Hence, and this is the point we want to emphasize, there
are aspects of reality that are inaccessible to human knowledge.
In what follows, with a frankly Kantian inspiration, and based on the account
given in [5], we sketch an interpretation of scientific systematic investigation where
the incidence of contradictions is compatible with an epistemic approach to para-
consistency.
Scientific research involves three levels: (i) Reality (with a capital R), (ii) empir-
ical phenomena and (iii) scientific theories. Reality represents those aspects of reality
that are inaccessible to human reason; this idea, of course, corresponds to the Kantian
372 9 Paraconsistency and Philosophy of Science: Foundations

notion of things-in-themselves. Empirical phenomena establish a mediated relation-

ship with Reality. It is based on this level that theories are elaborated, since it is here
that scientific experiments and data collection take placeeverything, of course,
mediated by perception, our conceptual apparatus and measuring instruments. There
is not, nor can there be, direct, non-mediated access to Reality.
In fact, it is not possible to establish conclusively that the schema above gives the
correct account of the construction of scientific theories. Nevertheless, we take it as
the best account of the practice of scientific research and the results of the empirical
sciences. We also agree that it is a somewhat skeptical view, but we think it is the
right measure of skepticism that millennia of philosophical reflection suggest as the
most plausible position.
Now it may be asked: in the schema above, where do contradictions occur? Of
course they occur in theories, since it is a fact that we deal with contradictory theories
(see Chap. 5 of [3], and also [57]). Aside from this, it is not implausible that
contradictions might also occur in empirical phenomena, concerning the way reality
presents itself to us mediated by our experience.
The central question is, however, whether there are contradictions in Reality
(with a capital R). These would be what we call real contradictions. However, in
order to answer this question once and for all we would have to have unconditioned
access to Reality, and according to the schema presented above this is not possible.
The claim, therefore, that there are contradictions in Reality cannot be anything but
a conjecture.
Surely, we cannot simply extend the features of empirical phenomena and scien-
tific theories to Reality: the occurrence of contradictions in phenomena and theories
does not imply the existence of contradictions in Reality. What we have on our hands,
indeed, is a typical philosophical problem that has no prospect of a definite solution.
Given that Reality is inaccessible, we cannot, from the fact that a given theory has
some internal contradictions, decide whether it is Reality that is contradictory or if
such a theory is going to be corrected in the course of further investigations.
It seems to us that claiming that there are real contradictions, in the sense explained
above, is a bit too metaphysical, in the bad sense of the term, the kind of speculative
philosophy Kant tried to fight against in the 18th century. We acknowledge, of course,
that we cannot conclusively affirm the opposite, that there are no real contradictions.
It would imply that we have direct access to Reality, a refutation of the schema
presented above, and we are not any kind of meta paraconsistentists so as to
endorse simultaneously a view and its refutation.
Notice, however, that in any circumstances in which we have full and unprob-
lematic access to phenomena, there is no sign of contradiction. Assuming that
nature is uniform and regular, an assumption that plays an essential role in scientific
research, it seems that there is no evidence to suppose that there are contradictions in
Reality. To insist on this point, trying to impose what cannot be rationally established
is like trying to establish the existence of time travel by Reductio ad Absurdum: if
time travel is possible, where are the tourists from the future?1 In any case, the more

1 Quote attributed to Stephen Hawking.

9.2 Consistency and Contradiction in Scientific Theories 373

important point for science is not whether the contradictions found by scientists are
real, but what can be made out of them.

9.2.2 Some Historical Examples

There are several examples where scientists hold contradictory propositions and end
up profiting from them. The controversy surrounding the movement of the luminif-
erous aether of the 19th century (see e.g. [8]) is a nice example of such a situation.
Experiments suggested that the aether could perform some kinds of motion relative
to matter, and thus the speed of light should depend on the speed of this motion.
Attempts to establish special relativity theory, led by physicists, have been con-
fronted with two seemingly contradictory experiments related to such aether wind:
an experiment of 1885, apparently confirming Fresnels stationary aether, and an
1887 experiment which seemed to confirm Stokes dragged aether.
However, instead of being rejected as non sense, such contradictory evidence
influenced Einsteins views on Electrodynamics and on the Principle of the Constancy
of Light. Einstein himself stated that such experiments, as well as Lorentzs theory of
1895, influenced his thinking. He borrowed, as he said, that principle from Lorentzs
stationary aether (which implies in particular the constancy of light in the aether
frame), but he recognized that this principle together with the principle of relativity
makes unnecessary references to aether. As A. Einstein wrote in 1907 and in later
papers, the apparent contradiction between those principles led to certain adaptations
of the notion of Lorentzs local time that revealed to be very fruitful.
The role of contradictory theories may even have a salutary effect in the expected
reconciliation between Einsteins general relativity and quantum mechanics, as put
in [9] (where contradictory theories are referred to as inconsistent theories):
The important point, then, is that while two mutually inconsistent theories cannot of course
both be true, they may both be approximately truethat is, both may emerge as (of course
different) limiting cases of some further, superior theory, just as current physicists expect
General Theory of Relativity and Quantum Mechanics both to emerge as limiting cases from
the eventual synthesis [9]

Our position is in agreement with [2], who claims that inconsistency in science is
a topic that realists of all stripes would be well advised to think more about:
My tentative conclusion is that anyone persuaded by the indispensability argument for sci-
entific and mathematical realism, should also (perhaps reluctantly) sign up for belief in
inconsistent objects. Note that this is not an argument that the world is inconsistent or that
the world contains inconsistent objects, just that there were times when we had warrant to
believe in such inconsistent objects.
374 9 Paraconsistency and Philosophy of Science: Foundations

9.2.3 The Beginning of Quantum Theory and

Paraconsistency

There are evidences, as in [10], that the beginnings of Quantum Theory could be
explained by understanding the position of N. Bohr and other physicists in terms of
a model based on a certain paraconsistent logic. But it is a different story to argue
that the logic behind science should be paraconsistent, and to recognize the role of
contradictions in science. The positions are not mutually exclusive, but we think that
while the latter is starting to be recognized, the former still has a long way to go.
In 1895 the astronomer H. von Seeliger concluded that basic assumptions of
Newtonian cosmology, specifically those concerning the infinity of the universe
(with stars uniformly distributed), lead to a contradiction under the action of New-
tonian forces, and therefore that the universe cannot be stable. An explanation of
why such a contradiction went unnoticed for two hundred years is ventured in [11].
Basically, it went unnoticed because of a confusion about non-convergent series
(it is simply not true that any infinite series converge to a real number, or sum to
infinitysome series are just indeterminate). Not every scientist, however, seems
to be totally convinced that all the alleged examples of inconsistent science in the
philosophical literature can really be reconstructed as such at all, or that the ones
that can be reconstructed are relevant for the philosophy of science.2 H. von Seeliger
(see [13], Chap. 1) proposed a small change in the inverse square law of gravita-
tional force which in some ways parallels (or anticipates) Einsteins cosmological
constant in his proposals for General Relativity and the contemporary hypothesis
of dark energy. What we see then, independently of the debate about whether the
logic behind science should be paraconsistent, is that contradictions have acted as
a significant driving force for the development of science.

9.2.4 Mercurys Orbit and a Non-existent Planet

By the end of the 19th century, a puzzling incongruence in Mercurys orbit was
observed: the data showed that the elliptical orbit shape was rotating or precess-
ing by small amounts (precession is a change in the orientation of the rotation axis
of a rotating body). A similar perturbation had been observed before: the recently
discovered planet Uranus had also deviated markedly from the orbit that New-
tonian theory prescribed for it. This contradiction between the theory and the data
(at any rate, always reducible to a contradiction between two theoretical frameworks,
since the data also compose a theory) lead the French astronomer U.J.J. Le Verrier
in 1846, recalculating orbits based on Newtons theory of gravitation, to predict

2 An acerbic criticism of those philosophers and logicians who have craved for, relied on, hinted
at or watched over inconsistent science (henceforth: iMongers), comparing them to warmongers,
as if science could be saved from contradiction by blaming the ones who point to the problem, is
advanced in [12].
9.2 Consistency and Contradiction in Scientific Theories 375

the position of a mass that would be perturbing Uranus orbit. The position was so
accurately predicted that when the astronomer J. Galle and his group, at the Berlin
Observatory, searched more carefully the sky, they indeed discovered a new planet,
the eighth planet of the Solar system, a planet that would come to be called Neptune.
The fact that such temporary contradiction was used for good profit represented
an enormous triumph not only for Le Verrier, but also for Newtons theory, and
gave scientists a confidence that all anomalies in planetary orbits could be explained
by the Newtonian theory, and that such contradictions could be used as support for
discovering new celestial objects.
The most obvious hypothesis, along with other competitors, was that, as much as
for Uranus, a yet-to-be-discovered celestial body was needed to explain the strange
misbehavior of Mercurys orbit: possibly a large number of small objects, or perhaps
even another planet, inside the orbit of Mercury.
The new, unborn member of the Solar system even had a name, Vulcan, and this
theory started to be known as the Vulcan hypothesis. After some time it became
clear that Vulcan did not exist (see, e.g., [14]), and that astronomers were now facing
a second contradiction, not about theory and data anymore, but about the theory
and an apparently coherent new hypothesis. In the case of Uranus, a contradiction
revealed that the Newtonian theory was correct, and even generated a new scientific
discovery. But in the case of Vulcan, a contradiction was pointing to a failure in
what seemed to be a brilliant, new method of discoveryor was Newtonian theory
at stake?
Given the impossibility of arriving at a satisfactory Newtonian explanation for the
anomalous precession, some researchers began to think that perhaps gravity would
not be exactly an inverse square law. Might Newtons law of gravitation require
precisification such that the power 2 be replaced by, say, 2.00000016, calculated
with more precision than was possible in Newtons time?
This idea was not, in the end, plausible, since it did not square well with other laws,
such as those concerning the conservation of mass and energy. The net result of this
other contradiction was, this time, a triumph for Einsteins General Relativity. The
explanation for this discrepancy was found by Einstein, who tried to convince physi-
cists that the extra precession arises unavoidably from the fundamental principles of
general relativity, see [15], specially Chaps. 5 and 18 (details are too technical to be
given here). Einstein is reported to have said that seeing the anomaly in Mercurys
orbit emerging from his geometrical field equations had given him palpitations of
the heart.
But that is not the end of the story. More recently, efforts have been made by sev-
eral experimentalists to explain some or all of Mercurys precession by oblateness
(or flatenning) in the shape of the sun (e.g. [16], see also [17]) in some ways contra-
dicting General Relativity. The issue gets more complicated, since Mercury is not the
only object in the solar system that exhibits anomalous precession. Contradictions
of different types, thus, have played a non-negligible role in the establishment of
celestial mechanics, and are no less influential in relativistic celestial mechanics.
376 9 Paraconsistency and Philosophy of Science: Foundations

9.2.5 Contradictions in Phlogiston, the Imponderable

The theory of phlogiston, an imponderable substance whose properties were inca-

pable of being detected by senses, was a great help in the beginnings of the chemistry
of the 18th Century, since it made it possible to given a theory explaining combustion.
Two main opponents, J. Priestley and A.-L. Lavoisier, quarreled about phlogiston
for years. Lavoisier began, in 1772, a series of experiments designed to convince
the chemistry community that the concept of phlogiston was superfluous. Priestley
conducted similar experiments, with Lavoisier and Priestley arriving at contradictory
conclusions. One of the crucial experiments was the thermal decomposition of the
red oxide of mercury (HgO) in 1774, which produced an eminently respirable air
able to support combustion and living organisms better than ordinary air. To Priest-
ley, this gas was dephlogisticated air: ordinary air deprived of phlogiston, precisely
because the current view was that fire burns out, and animals die in an enclosed space,
because the air becomes saturated with phlogiston. To Lavoisier, this gas was oxygen
(details in [18]).
A serious problem was that the phlogiston hypothesis not only led to contradictory
conclusions, but was itself contradictory. Indeed, the calx formed when certain met-
als, such as magnesium, burn weighs more than the metal from which it formed, but
the contemporary hypothesis was that burning substances allegedly lose phlogiston.
Supporters of the phlogiston theory responded with the additional hypothesis that
metallic phlogiston has negative weight, while other combustibles contain phlogiston
with positive weight (perhaps a not so absurd idea, given the alleged imponderability
of phlogiston, but suspicious enough to start to convince the chemical community
about the superiority of the oxygen theory of combustion).
The story was of course much more complicated, and theories of dephlogisticated
air still competed with the oxygen theory, but other experimental results further
contributed to the invalidation of the phlogiston hypothesis. However, the difficulty
remained (see [18]) in demonstrating the non-existence of such a hypothetical entity
as phlogiston.3 But for our purposes it is worth remarking that the attitude of those
18th century scientists was obviously of an epistemic nature regarding the many
contradictions found: they concentrated on the difficulties involving the relevant
experiments, and no one ever concluded that the fire burn and does not burn, or
anything of that sort. The fact was that the phlogiston theory, even if contradictory,
gave a great impetus to chemistry (see [18], p. 1919):
The picture that emerges from even this abbreviated history is that of two competing theories,
one of which eventually prevailed by a steady accretion of new experimental results linked
to a more coherent explanatory framework.[xv] Lavoisiers balance sheet method, which
accounted for the weights of all reactants and products and showed that nothing ponderable
was created or destroyed, was very effective in converting other chemists to his system. Yet
the rigorous application of this method required Lavoisier to adopt some premises of the
scheme he had set out to overthrow. For example, in order to account for the three states of

3 The reader should perhaps compare with an analogous difficulty in demonstrating the non-existence

of dialetheias, qua ontologically true contradictions; dialetheias may be useful as hypothetical

entities to simplify the explanation of paraconsistent logic, but they are hardly necessary.
9.2 Consistency and Contradiction in Scientific Theories 377

matter and for the heat released in oxidation reactions, Lavoisier posited that all substances
contain varying amounts of caloric, the material basis of heat. Since he obtained nearly
perfect agreement between the weights of reactants and products to the extent that some of
his opponents had understandable doubts about his claim[xvi] Lavoisier found it necessary
to assume that caloric, although real, was an imponderable fluid. Shadows of phlogiston!

The phlogiston theory was the first chemical theory (as opposed to the predomi-
nant mechanical theory) which could explain the qualitative changes characteristic
of chemical processes. Contradictory as it was, phlogiston theory was useful and
decisively contributed to the founding of modern chemistry. And its shadows still
remain.

9.2.6 The Special Theory of Relativity

A good example of a provisional contradiction in Physics is the problem faced by

Einstein just before he formulated the special theory of relativity in 1905.4 It is well
known that there was an incompatibility between classical, Newtonian mechanics
and Maxwells theory of electromagnetic field. This is a typical case of two non-
contradictory theories that, when put together, yield contradictory results.
Classical mechanics gives a description of bodies changing position in space and
time. It is intuitively understood, and works very well, with respect to slow objects
(we will see soon what would be a criterion for an object to be or not to be slow).
Accordingly to classical mechanics (CM), given two systems of coordinates K
and K  , K  in uniform linear motion (ULM) with respect to K, an object o in ULM
with respect to K will also be in ULM with respect to K  . Let v be the velocity of K 
with respect to K, and (x, y, z, t) the space-time coordinates of o in K at the time t.
The corresponding space-time coordinates (x  , y  , z  , t  ) of o in K  relatively to K
are given by the following equations, called Galilean transformations:

x  = x vt; y  = y; z  = z; t  = t.

From the equations above, the so-called theorem of the addition of velocities may
be obtained:
w  = w v,

where w is the velocity of o w.r.t. K and w is the velocity of o w.r.t. K. All of this
expresses the very familiar fact that if an object o is moving inside a train with a
velocity w  w.r.t. the train, and the train is moving with velocity v w.r.t. the rails
(both in ULM, w  and v in the same direction), the velocity w of o w.r.t. the rails is
the sum of w  and v.

4 Thediscussion in this subsection is based upon material from [19], mostly following A. Einstein
himself in [20] and Chap. 15 of [21].
378 9 Paraconsistency and Philosophy of Science: Foundations

In the second half of 19th century, the physicist J.C. Maxwell formulated the
so-called theory of electromagnetic field (EM) that gives a unified account of the
phenomena of electricity, magnetism and light. According to this theory,
(i) the velocity of light (c) is equal to 300,000 km/s, and
(ii) c is independent of the motion of its source.
Now let us modify a bit the example above. Suppose that instead of an object
moving inside the train, we are concerned with the light emitted by the headlight of
the train. According to classical mechanics, the velocity w of the light with respect
to the rails would be the sum of the velocity of the train and the velocity of light,
that is,
w = c + v.

However, according to Maxwells theory, the velocity of light does not depend on
the velocity of the train, that is,
w = c.

We have, thus, that classical mechanics and the theory of electromagnetic field prove
a pair of contradictory propositions:

C M  (w = c), while E M  w = c.

So, if we put the two theories together, we get a trivial theory, if the underlying logic
is classical.
In the situation described above, two contradictory propositions A and A hold
in the sense that both may be proved from theories that were supposed to be correct.
This fact may be represented by the attribution of the value 1 to both A and A. But
clearly, the meaning of this should not be that both are trueactually, we know this
is not the case, and nobody has ever supposed that it could be the case. The meaning
of the simultaneous attribution of the value 1, as we suggest, is that at that time there
was evidence for both in the sense, mentioned above, of some reasons for believing
that both are true, because there were evidence that the results yielded by both CM
and EM were true.
What we have just seen above is that classical mechanics was not compatible with
Maxwells equations, or more precisely, the latter were not invariant under Galilean
transformations. This situation has been solved by Einstein. By the end of 19th
century, H.A. Lorentz had already presented a group of equations, called Lorentz
transformations,

x vt t vx/c2
x =  ; y  = y; z  = z; t  =  ,
1 v 2 /c2 1 v 2 /c2

and the interesting fact is that Maxwells equations are invariant under Lorentz
transformations. Einstein then rewrote Newtons equations in such a way that the
theory so obtained, the special relativity theory, was fully compatible with the theory
9.2 Consistency and Contradiction in Scientific Theories 379

of electromagnetic field.5 From the new equations, a different theorem of addition

of velocities may be proved:
wv
w =
1 wv/c2

The contradiction has been solved in the following way: as velocity grows,
time slows down and space shortens. So, the relation between space and time
(roughly speaking, space divided by time) that gives velocity, remains the same,
because both have decreased. What the special theory of relativity shows is that clas-
sical mechanics is a special case of the former. Newtons equations work well for
slow objects, that is, objects moving in such a way that the value of wv/c2 may be
ignored.
We have just seen an example of what we call epistemic contradictions. We want
to call attention to the fact that the logical framework Einstein was working in was
not classical. He had two different theories at hand, CM and EM that, when put
together, yielded a non-explosive contradiction. Later, according to the special theory
of relativity, the contradiction disappeared. Although there was some reasons to
believe that both w = c and (w = c) were true, only one, the latter, has been
proved true. The value 1 attributed to (w = c) later became 0.

9.2.7 Mathematics, and the Meaning of Objects that Mean

Nothing

In mathematics, of course, the notions of contradiction and consistency, as well as

non-contradiction and inconsistency, play a perennial role, as mathematical objects
derive their meaning from relations they bear to other objects or theories, and not
from experiments or measurements.
Examples are abundant from geometry to number systems, set theory, and other
branches of mathematics. Among the vast literature, [22] offers a collection of papers
with several cases, from Dedekind to Gdel, relevant for the philosophy of mathe-
matics. Some of these papers focus on the role of several degrees of contradiction
involved in some mathematical concepts and their reception for the logical, philo-
sophical and mathematical community.
Under the general designation of paradoxes, [23] studies in detail several cases
of contradictions in commonly held notions in mathematics, misconceptions, true
statements that seem to be false and false statements that seem to be true. Notable
examples are Russells paradox, the Liar paradox and Currys paradox, for which a
vast literature exist and about which there is still heated debate.

5 Actually,what Einstein did was to consider that the mass of a body increases with velocity, and
of course this changed the whole thing. We are not going into more details, but a friendly and
accessible presentation may be found in Chap. 15 of [21].
380 9 Paraconsistency and Philosophy of Science: Foundations

Other examples dealing with numbers, functions, logarithms, continuity, infinite

series, sets, curves and geometrical objects are carefully discussed in [23], and among
a bunch of interesting cases, the dispute around the computation of log(1), the log-
arithm of 1, involving Johann Bernoulli, Gottfried Wilhelm Leibniz and Leonhard
Euler is especially fascinating. While Leibniz claimed it was imaginary (i.e., a com-
plex number), Bernoulli insisted it should be a real number. The solution was given
by Euler in 1749, employing the famous Euler-Cotes formula

ei = cos( ) + i sin( ) .

Now, ei(+2n) = 1 hence log(1) assumes not one, but several (in fact infinite)
values: i( + 2n ), for n = 0, 1, 2, . . . . This analysis gives rise to a new
mathematical notion: that of multifunction (also called multivalued function). In
fact, the logarithm of a complex (or imaginary) number a + bi (and, in particular,
of a negative number, when a < 0 and b = 0) produces not a single, but an infinite
number of values, namely the set
    
log a 2 + b2 + i arg(a + bi) + 2n : n is an integer .

This means that both, Bernoulli and Leibniz were wrong, but the dispute around the
contradictory nature of log(1) helped Euler to introduce a new category of mathe-
matical object indispensable in contemporary science, specifically, the multivalued
functions.6
But mathematicians, logicians and philosophers continued to debate about the
imaginary numbers, a mathematical tool used since as early as the 16th century to
solve equations that were unsolvable in any other way, and which generated perplexity
on its connections with real numbers until the early 19th century. In an address on the
topic of completeness and the imaginary, delivered to the Gttingen Mathematical
Society in 1901, invited by D. Hilbert and F. Klein, E. Husserl expounded his ideas,
in frank opposition to G. Freges views that expressions failing to denote objects
are unsuitable for use in science, even if expressed in a logically perfect way. As
[24] puts it (pp. 150151), Husserl recognized that contradictory aspects of some
mathematical entities, imaginary numbers being a case in point, may not necessarily
be negative or devoid of meaning7 :
Questions regarding imaginary numbers, he continued, had come up in mathematical contexts
in which formalization yielded constructions which arithmetically speaking were nonsense
but which, astonishingly, could nonetheless be used in calculations. It became apparent that
when formal reasoning was carried out mechanically as if these symbols had meaning, if
the ordinary rules were obeyed, the results did not contain any imaginary components, then
these symbols might be legitimately used. And this could be empirically verified (p. 432).

6 Recall that multivalued operations, defining swap structures, are essential in order to semantically

characterize some LFIs, as shown in Chap. 6.

7 In the text below, the page numbers refers to the notes for Husserls lecture published as an appendix

to the Philosophie der Arithmetik, E. Husserl, Pfeffer, Halle, 1891.

9.2 Consistency and Contradiction in Scientific Theories 381

Husserl did not believe that general logic could shed light on the mystery because of the
importance logicians accord to working with clear, precise, unambiguous concepts so that
contradictions do not sneak in unnoticed. Logicians would ban contradiction, he said. For
them contradictions only serve to show that a concept does not have an object, and contra-
dictory concepts but yield contradictory consequences to which no object will correspond.
But with the imaginary in mathematics that is plainly not the issue (p. 433).

This facet of imaginary numbers already astonished A. Whitehead, who in 1911

admitted ([25], pp. 6162):
[]in some mysterious way, symbols which mean nothing can by appropriate manipulation
yield valid proofs of propositions.
The mistake of Whitehead was to suppose that because something has contradic-
tory aspects, it means nothingindeed, the imaginary numbers as we know them
today have widened the mathematical spectrum: they are essential in control the-
ory via Laplace transformations, complex valued functions are analyzed via Fourier
analysis, certain real-valued improper integrals are computed with the help of com-
plex numbers, and quantum mechanics, and advanced cryptography, employ them
as well. It is hard to deny that complex numbers mean something.
There are even new geometric properties holding between euclidean figures
(such as triangles and ellipses) that were simply unthinkable before the advent of
imaginary numbers which can be established via imaginary numbers, as illustrated by
Mardens Theorem (originated circa 1892 in papers by M. Bcher and J. Siebeck).
This theorem (see [26] for an elementary proof) describes a surprising geometric
connection between the roots of a cubic polynomial p with complex coefficients and
the roots of the its derivative p  . To wit, if the roots of p are non-collinear points A, B,
and C in the complex (Argand) plane, the roots of p  are the foci of a unique ellipse
inscribed in the triangle ABC and tangential to the sides at their midpoints. Mar-
dens theorem establish a kind of generalization of Rolles theorem for real-valued
functions in the complex domain.
What is more surprising is that, not only do we commonly see mathematical
theories flirting with the contradictory, but some mathematicians themselves teeter
on the edge of contradiction, as J. Hintikka puts it in [27], p. 39:
More generally but perhaps somewhat less conspicuously, my analysis throws some light on
the curious history of the axiom of choice. One of the curiosities here is that a large number
of the vocal critics of the axiom of choice were in a subsequent analysis caught in flagrante
delicto as having used the principle tacitly in their own research. This is documented carefully
in Moore. [28].8 Such inconsistencies are not what we should expect of highly sophisticated
mathematicians, and they require therefore some deeper explanation.
Perhaps we should admit that traces of contradiction are inherent within human
rationality, and instead of admonishing people it would be more productive to better
understand this phenomenon.
On a lighter note, [29] recalls several cases in mathematical thinking where ambi-
guity, paradox and contradiction are always present, from the most elementary to the

8 Thenote refers to G.H. Moore: 1982, Zermelos Axiom of Choice, Springer-Verlag, Berlin-
Heidelberg-New York, pp. 6476.
382 9 Paraconsistency and Philosophy of Science: Foundations

most advanced level, and that ambiguity, contradictions, and paradoxes are inherent
in mathematical creativity. With examples from geometry, arithmetic, the idea of
zero and of infinity, the book shows that it is better to live with ambiguous and con-
tradictory mathematical ideas than to get rid of them, and that abstraction consists
essentially in the creation and utilization of ambiguity.
Paraconsistent logic, and LFIs in particular, has found a good number of appli-
cations in theoretical computer science, in areas such as artificial intelligence,
description logics, programming languages, software engineering, databases and
information systems engineering, just to name the most popular. The cases are sim-
ply too many to be treated individually, but perhaps what is more surprising is to
find applications in linguistics, as discussed by [30]: recognizing that Chomskys
theory allows for the temporary tolerance of contradiction between the hypotheses
and the data, the authors warn that in linguistics there are several types of incon-
sistency, and their paper develops what they call a novel metatheoretical framework
which provides tools for the representation and evaluation of inconsistencies in lin-
guistic theories. The proposed approach distinguishes between strong and weak
inconsistency. Strong inconsistency is destructive, in that it leads to trivialism. In
contrast, weak inconsistency may be constructive, because (they say) it is capable
of accounting for the simultaneous presence of seemingly incompatible structures.
The authors of [30] conclude too hastily, however, that paraconsistent logic cannot
grasp the dynamism of the emergence and resolution of weak inconsistencies. That
distinction is precisely the contribution of LFIs, as this book aims to show. In any
case, misunderstandings about paraconsistent logic and confusions between para-
consistency and dialetheism of this sort will hopefully diminish, but the positive side
is that paraconsistency and LFIs also have promising applications in linguistics.

9.3 Quasi-truth and the Reconciliation of Science and

Rationality

The concept of pragmatic truth, later referred to as quasi-truth, or partial truth, was
introduced in [31] as an attempt to loosen the tight constraints of the traditional,
Tarskian account of formalized truth, with the intention of capturing the meaning
of pragmatist and anti-realist (or empiricist) theories of truth in the philosophy of
science. This semantical approach was already analyzed, from a formal perspective,
in Sect. 7.9 of Chap. 7.
A more refined, and thorough account of a partial-truth approach to models and
scientific reasoning is given in [3]. The philosophical rationale behind this proposal
is to overcome the gulf between our actual knowledge-gathering activities and the
philosophical characterization of these activities (see [3], p. 3). Scientific represen-
tation is indeed imperfect, incomplete, partial and not rarely contradictory, as noted
by many (and as our examples suggest). The quasi-truth approach is realistic, and
coherent with scientific practice: the concept of quasi-truth is weaker than truth, and
theories can have, in principle, degrees of quasi-trueness, without being absolutely
9.3 Quasi-truth and the Reconciliation of Science and Rationality 383

true. If this view seems appropriate for an anti-realist view of natural sciences, it is
debatable whether it is also good for mathematics, logic, statistics, and other formal
sciences, such as theoretical computer science, information theory, game theory, the-
oretical linguistics and cognitive science. Nothing, however, in principle impeaches
a partial-truth account of such disciplines.
One of the main problems concerning scientific rationality is to try to understand
how science develops, and how theories are selected and substituted in the long term,
as put in [32]:
How to entertain with the best rational attitude the periods when contradictions and appar-
ently irreconcilable opposition between theories coexist?

This problem involves two main interconnected issues according to [32]:

(a) How should we rationally understand theory change in science? In particular, how should
we accommodate, in rational terms, the apparent lack of a cumulative development in scien-
tific development? What is at stake here are the criteria of theory selection, and the difficul-
ties of rationally accommodating scientific development given the presence of radical theory
changes in science. This immediately raises the second issue: (b) How should we make sense
of episodes that apparently challenge the rationality of scientific theorizing? For example,
how should we understand the various situations in which scientists, or even mathematicians,
entertain inconsistent theories? Are these simply cases of scientific irrationality?

Noting the blatant diversity of cases in which dramatic theoretical changes are
involved, leaving scientists and mathematicians facing the trouble of (even if tem-
porarily) entertaining contradictory theories, a model of scientific rationality is out-
lined in [32]. The intention is to accommodate these two issues, making explicit the
role of such contradictory theories and yielding an account of scientific rationality
that is able to make better sense of scientific and mathematical activity.
Now, if scientific theories can rationally be taken to be quasi-true, the underlying
logic is necessarily paraconsistent, on pain of trivialism, a disaster to be avoided at
all costs.
A paraconsistent logicin fact, a first-order three-valued LFI, called LPT1
providing support for quasi-truth, was proposed in [33], and proven to be sound
and complete with respect to a certain semantics of triples. A revised version of
this proposal was given in Sect. 7.9 of Chap. 7. The strategy in [33] avoids con-
structing total structures, with a minimal detour from the Tarskian notion of sat-
isfaction. Then it is proved that LPT1 coincides (up to language) with LFI1,
the quantified version of LFI1 proposed by Carnielli et al. in [34], and also with
the quantified version of J3 studied several decades ago by I.M.L. DOttaviano
(see [3538]). After this, LPT1 is compared with the logic LP, introduced by
G. Priest in [39] from F. Asenjos proposal in [40], as a formal framework for studying
antinomies.
The logic of paradox LP, studied in more detail in Sect. 4.4.5, is one of the main
3-valued paraconsistent logics introduced in the literature, and has a certain kinship
with ukasiewicz and Kleenes 3-valued logics. It was proved in [33] that LPQ, the
first-order version of LP, is a fragment of LPT1, and so the latter is a conservative
extension of the former. The three-valued propositional logic underlying LPT1,
384 9 Paraconsistency and Philosophy of Science: Foundations

called MPT, was described in Sect. 4.4.7 of Chap. 4, while J3 and LP were briefly
described in Sects. 4.4.3 and 4.4.5 of that chapter, respectively.
The notion of quasi-truth, or partial truth, is therefore, closely connected to the
paraconsistent LFI paradigm and constitutes a non-dogmatic overture to the dynam-
ics of theory change in science, tolerant to the flounderings of scientific practice.
In [41] and [42] this question is studied under the perspective of the AGM theory
of Belief Change based on LFIs (a good general reference for Belief Change is
S. Hanssons book [43]).
Paraconsistent Belief Revision systems apply their tools to elicit the very notion of
rationality within a paraconsistent setting. It is possible to explain, in particular, the
role the consistency operator, as introduced by LFIs, has to play within a dynamic
context. By considering the existence of contradictions as a natural consequence
of the dynamics of rational thinking, the strictures set by the Belief Revision sys-
tems operations within a paraconsistent approach are to be reinterpreted. Thus, such
approach could also be taken as a paradigm for scientific reasoning. One innovation
of [41, 42] is to understand consistency as an epistemic attitude, thus clearing the
way for further inquiries about the epistemological features of paraconsistency, as
the ones discussed in Chap. 1 and in the present chapter.

9.4 An Evidence-Based Approach to Paraconsistency

A pair of contradictory sentences may be understood as conveying conflicting infor-

mation, or conflicting scenarios, or imposing a rational deadlock, and when one
designs a paraconsistent logic aiming at giving rational tools to deal with them, the
question of acceptance of a contradictory pair of sentences without falling into trivi-
alism requires some philosophically palatable position; the notion of evidence seems
particularly promising as a vantage point from which to understand paraconsistent
logics.
Evidence may be understood, as it is usual in epistemology, as the relevant ingre-
dient for justified belief [1] taking into account that justified belief is, in traditional
terms, a necessary condition for knowledge. From this point of view, the idea of
preservation of evidence suggests itself as a topic to be further developed in LFIs,
in an anlogous way to the idea of preservation of certain constructions, as in the
well-known BHK interpretation of intuitionistic logic.
On the other hand, however, a paraconsistentist approach to evidence may also
be characterized from a probabilistic point of view. The aim of evidence theory is to
give an account of reasoning under epistemic uncertainty, that may happen due to a
lack of data and/or a defective understanding of the available data related to some
scenarios. It is distinguished from randomic uncertainty, that is due to the inherently
aleatory character of some scenarios.
One of the key points of evidence theory is that in imprecise events, uncertainty
about an event can be quantified by the maximum and minimum probabilities of that
9.4 An Evidence-Based Approach to Paraconsistency 385

event. This directly connects evidence theory to probability theory, and therefore to
Bayesian conditionalization.
In [44] the authors investigate a paraconsistent approach to theory of probability
based on the logics of formal inconsistency. The paper shows that LFIs naturally
encode an extension of the notion of probability able to express the notion of prob-
abilistic reasoning under contradictions by means of appropriate notions of condi-
tional probability and paraconsistent updating, via a version of Bayes Theorem for
conditionalization. The paper argues that the dissimilarity between the notions of
inconsistency and contradiction plays a central role in an extended notion of proba-
bility that supports contradictory reasoning.
Actually, evidence theory and Bayesian theory of subjective probability are simul-
taneously generalized by Dempster-Shafer theory, which concerns belief and plau-
sibility. Beliefs from different sources can be combined (by means of the so-called
Dempster rule of combination) with various operators used to model specific situa-
tions of belief diffusion.
An analogous treatment can be given using paraconsistent probability. Another
approach to an evidence-based interpretation to paraconsistency, related (though not
identical) to paraconsistent probability, is the game-theoretical view of paraconsis-
tency, still to be developed in full. A game-theoretical account (by means of dialogical
logic) of paraconsistency is defended in [45]. Dialogical logic makes it possible to
accommodate the occurrence of contradictions in two (or more) persons reasoning,
and contributes positively to debates concerning the ontological versus epistemo-
logical nature of contradictions. Rahman and Carniellis paper has attracted some
attention (see e.g. [46]) as an relevant first step into reformulating paraconsistent
logic in a dialogue format.
Dialogical logic is not the only approach to paraconsistency from the point of view
of game theory. For instance, in a paraconsistent game-theoretical scenario the truth of
a sentence can be defined in terms of the lack of winning strategies for the Opponent,
instead of in terms of existence of winning strategies of the Proponent; a similar
view is defended in [47], but with different assumptions (namely, the existence of
true contradictions or dialetheias). A related possibility is granted by the definition of
so-called team semantics in [48], although the idea of a society producing semantics
had been introduced (years before) in [49].

9.5 Summing Up

A logic has epistemic rather than ontological character when its subject matter refers
not only to truth, but also to some other concept strictly related to reason. This is the
case of intuitionistic logic, which is concerned with truth attained in a specific way, by
means of a constructive proof. And we claim that this is also a way of understanding
paraconsistency in general, and particularly the logics of formal inconsistency. The
latter, we may say, is concerned with truth, since classical logic can be recovered,
but it is also concerned with a notion weaker than truth, and it is precisely this
386 9 Paraconsistency and Philosophy of Science: Foundations

notion weaker than truth that allows an intuitive and plausible understanding of the
acceptance of contradictions in some contexts of reasoning.
Intuitionistic logic is a special sort of paracomplete logics, that is, logics in which
there is a model M and a sentence A such that both A and its negation do not hold
in M. Mathematicians deal with lack of information in the sense that there are many
unsolved mathematical problems. This is one of the reasons for the rejection of the
law of excluded middle by intuitionists. In empirical sciences, on the other hand,
although obviously many things are not known, the researcher sometimes deals
with conflicting information, and very often with contradictory information. Thus
(s)he sometimes may have to provisionally consider two contradictory claims, one
of which in due time will be rejected. Both intuitionistic and paraconsistent logics
may be conceived as normative theories of logical consequence with an epistemic
character. Notwithstanding, both are also descriptive: the first, according to Brouwer
and Heyting, is intended to represent how the mind works in constructing correct
mathematical proofs, while the latter, we argue, represents how we draw inferences
correctly when faced with contradictions.
In fact, it is not surprising that we find a kind of duality in the motivations for
intuitionistic and paraconsistent logics. Dual-intuitionistic logics have been shown
in [50] to be essentially paraconsistent, and similar results in the other direction
have been investigated. Nevertheless, the dual of Heytings well-known intuitionistic
logic gives rise to a new paraconsistent logic, that is, one that is not a familiar
paraconsistent logic. Another dualities between paraconsistency and intuitionism,
even from a topological perspective, were also investigated in [5153]. These results,
in any case, show an intrinsic relationship between both paradigms, but there is of
course much to be investigated in this regard.
Conceptions of consistency as coherence among a collection of statements, or
as continuity or persistence of a collection of statements in time, could also be
studied from a paraconsistent point of view. It is worth noting that the ideas presented
here indicate that there is much more to be explored in Hegel than just the (maybe
misleading) idea of true contradictions.
Although classical first-order logic is a powerful tool for modeling reasoning, the
fact that it does not handle contradictions in a sensible way, due to the insistence on
the Principle of Explosion, is an intrinsic drawback. In essence, classical logic is too
brutal to sense contradictions and to appraise their meaning. In theoretical computer
science, for example, facts and rules of knowledge bases, as well as integrity con-
straints, can produce contradictions when combined, even if they are sound when
separate, and this is also the case for description logics. For this reason, the devel-
opment of paraconsistent tools has turned out to be an important issue in working
with description logics and expandable knowledge bases, as well as with large or
combined databases (see [34, 54]).
The new approach to consistency delivered by the logics of formal inconsistency
has also generated a good deal of interest in the field of inferential probability and
confirmation theory. Because of the well-known debate in the philosophical literature
on the long-standing confusion about probability when confronted with confirma-
tion (see e.g. [55, 56]), notions such as coherence, credence, consistent or coherent
9.5 Summing Up 387

individual profiles versus group profiles, etc. (see [57]), can be fruitfully approached
from the point of view of the logics of formal inconsistency, although there is still a
long way to go. If this book helps in any way to clarify the strength of paraconsistent
logics in the endeavour of reasoning it will have been successful.

References

1. Kelly, Thomas. 2014. Evidence. In The Stanford Encyclopedia of Philosophy, ed. Edward N.
Zalta. Fall 2014 edn.
2. Colyvan, Mark. 2008. The ontological commitments of inconsistent theories. Philosophical
Studies 141(1): 115123.
3. da Costa, Newton C.A., and Steven French. 2003. Science and Partial Truth: A Unitary
Approach to Models and Scientific Reasoning. London: Oxford University Press.
4. Meheus, Joke (ed.). 2002. Inconsistency in science. Dordrecht: Kluwer Academic Publishers.
5. da Costa, Newton, C.A., and Dcio Krause. 2014. Physics, inconsistency, and quasi-truth.
Synthese 191(13): 30413055.
6. da Costa, Newton C.A., and Dcio Krause. 2004. Complementarity and paraconsistency. In
Logic, Epistemology, and the Unity of Science. Logic, Epistemology, and the Unity of Science,
vol. 1, ed. by Sahid Rahman, John Symons, Dov M. Gabbay, and Jean Paul van Bendegem,
557568. Dordrecht: Springer.
7. Vickers, Peter. 2013. Understanding Inconsistent Science. Oxford University Press.
8. Born, Max. 1956. Physics in my Generation. London: Pergamon Press.
9. Worral, John. 2011. Underdetermination, realism and empirical equivalence. Synthese 180(2):
157172.
10. Brown, Bryson. 1992. Old quantum theory: A paraconsistent approach. Proceedings of the
Biennial Meeting of the Philosophy of Science Association 2: 397441.
11. Vickers, Peter. 2009. Was Newtonian cosmology really inconsistent? Studies in History and
Philosophy of Modern Physics 40(3): 197208.
12. Muller, Fred A. 2014. The slaying of the iMongers. Studies in History and Philosophy of
Modern Physics 48: 5255.
13. Byrd, Gene, Mauri Valtonen, Arthur Chernin, and Veli Teerikorpi. 2012. Paths to Dark Energy:
Theory and Observation. Berlin: De Gruyter.
14. Baum, Richard, and William Sheehan. 2003. In Search of Planet Vulcan: The Ghost in Newtons
Clockwork Universe. Basic Books.
15. Schutz, Bernard. 2004. Gravity from the Ground Up: An Introductory Guide to Gravity and
General Relativity. Cambridge: Cambridge University Press.
16. Dicke, Robert H., and H. Mark Goldenberg. 1967. Solar oblateness and general relativity.
Physical Review Letter 18: 313316.
17. Hawking, Stephen W., and Werner Israel (eds.). 1989. Three Hundred Years of Gravitation.
Cambridge: Cambridge University Press.
18. Labinger, Jay A., and Stephen J. Weininger. 2005. Controversy in chemistry: How do you
prove a negative? The cases of phlogisto and cold fusion. Angewandte Chemie (International
Edition) 44(13): 19161922.
19. Carnielli, Walter A., and Ablio Rodrigues. 2016. Paraconsistency and duality: between onto-
logical and epistemological views. In Logic Yearbook 2015. College Publications.
20. Einstein, Albert. 1916. Relativity: The Special and General Theory, 2013 edn. Emporum Books.
21. Feynman, Richard P., Robert B. Leighton, and Matthew Sands. 2010. The Feynman Lectures
on Physics, vol. 1. Basic Books.
22. Hintikka, Jaako (ed.). 1995. From Dedekind to Gdel: Essays on the Development of the
Foundations of Mathematics. Synthese Library, vol. 251. Kluwer Academic Publishers.
388 9 Paraconsistency and Philosophy of Science: Foundations

23. Kleiner, Israel, and Nitsa Movshovitz-Hadar. 1994. The role of paradoxes in the evolution of
mathematics. The American Mathematical Monthly 101(10): 963974.
24. Hill, Claire O. Husserl and Hilbert on completeness. In [Hintikka, Jaako, ed. 1995. From
Dedekind to Gdel: Essays on the Development of the Foundations of Mathematics, Vol. 251
of Synthese Library. Kluwer Academic Publishers], pp. 143163.
25. Whitehead, Alfred. 1911. An Introduction to Mathematics. Cambridge: Cambridge University
Press. Reprinted by Oxford University Press, Oxford, 1958.
26. Kalman, Dan. 2008. An elementary proof of Mardens theorem. The American Mathematical
Monthly 115(4): 330338.
27. Hintikka, Jaako. Standard vs. nonstandard distinction: A watershed in the foundations of math-
ematics. In [Hintikka, Jaako, ed. 1995. From Dedekind to Gdel: Essays on the Development of
the Foundations of Mathematics, Vol. 251 of Synthese Library. Kluwer Academic Publishers],
pp. 2144.
28. Bochvar, Dmitri A. 1981. Ob odnom trechzna c nom iscislenii i ego primenenii k analizu
paradoksov klassiceskogo funkcionalnogo iscislenija (in Russian). Matmaticeskij Sbornik
46(2): 287308, 1938. Translated to English by M. Bergmann as On a three-valued logical
calculus and its application to the analysis of the paradoxes of the classical extended functional
calculus. History and Philosophy of Logic 2: 87112.
29. Byers, William. 2007. How Mathematicians Think: Using Ambiguity, Contradiction, and Para-
dox to Create Mathematics. Princeton University Press.
30. Kertsz, Andrs, and Csilla Rkosi. 2013. Paraconsistency and plausible argumentation in
generative grammar: A case study. Journal of Logic, Language and Information 22(2): 195
230.
31. Mikenberg, Irene, Newton C.A. da Costa, and Rolando Chuaqui. 1986. Pragmatic truth and
approximation to truth. The Journal of Symbolic Logic 51(1): 201221.
32. Bueno, Otvio, and Newton C.A. da Costa. 2007. Quasi-truth, paraconsistency, and the foun-
dations of science. Synthese 154(3): 383399.
33. Coniglio, Marcelo E., and Luiz H. Silvestrini. 2014. An alternative approach for quasi-truth.
Logic Journal of the IGPL 22(2): 387410. doi:10.1093/ljigpal/jzt026.
34. Carnielli, Walter A., Joo Marcos, and Sandra de Amo. 2000. Formal inconsistency and evo-
lutionary databases. Logic and Logical Philosophy 8: 115152.
35. DOttaviano, Itala M.L. 1982. Sobre uma Teoria de Modelos Trivalente (On a three-valued
model theory, in Portuguese). Ph.D. thesis, IMECC, State University of Campinas, Brazil.
36. DOttaviano, Itala M.L. 1985. The completeness and compactness of a three-valued first-order
logic. Revista Colombiana de Matemticas XIX(12): 7794.
37. DOttaviano, Itala M.L. 1985. The model extension theorems for J3-theories. In Methods in
Mathematical Logic. Proceedings of the 6th Latin American Symposium on Mathematical
Logic held in Caracas, Venezuela, 16 August 1983. Lecture Notes in Mathematics, vol. 1130,
ed. by Carlos A. Di Prisco, 157173. Springer.
38. DOttaviano, Itala M.L. 1987. Definability and quantifier elimination for J3-theories. Studia
Logica 46(1): 3754.
39. Priest, Graham. 1979. The logic of paradox. Journal of Philosophical Logic 8(1): 219241.
40. Asenjo, Florencio G. 1966. A calculus for antinomies. Notre Dame Journal of Formal Logic
16(1): 103105.
41. Testa, Rafael R. 2014. Reviso de Crenas Paraconsistente Baseada em um Operador For-
mal de Consistncia (Paraconsistent Belief Revision Based on a Formal Operator of Consis-
tency, in Portuguese). Ph.D. thesis, IFCH, State University of Campinas, Brazil. http://www.
bibliotecadigital.unicamp.br/document/?code=000935185&lg=en_US.
42. Testa, Rafael R., Marcelo E. Coniglio, and Mrcio M. Ribeiro. 2015. Paraconsistent belief
revision based on a formal consistency operator. CLE e-Prints, 15(8). http://www.cle.unicamp.
br/e-prints/vol_15,n_8,2015.html.
43. Hansson, Sven O. 1999. A Textbook of Belief Dynamics. Applied Logic Series, vol. 11. Dor-
drecht: Kluwer Academic Publishers.
References 389

44. Bueno-Soler, Juliana, and Walter A. Carnielli. 2015. May be and may be not: paraconsistent
probabilities from the LFI viewpoint. Submitted.
45. Carnielli, Walter A., and Sahid Rahman. 2000. The dialogical approach to paraconsistency.
Synthese 125(12): 201232.
46. Van Bendegem, Jean-Paul. 2001. Paraconsistency and dialogue logic: Critical examination and
further explorations. Synthese 127(12): 3555.
47. Galliani, Pietro. 2012. Dialetheism, game theoretic semantics, and paraconsistent team seman-
tics. ArXiv e-prints. http://arxiv.org/abs/1206.6573.
48. Vnnen, Jouko. 2007. Dependence Logic: A New Approach to Independence Friendly Logic.
Cambridge: Cambridge University Press.
49. Carnielli, Walter A., and Mamede Lima-Marques. 1999. Society semantics and multiple-valued
logics. In Advances in Contemporary Logic and Computer Science. Proceedings of the XI
Brazilian Conference on Mathematical Logic, May 1996, Salvador, Bahia, Brazil. Contem-
porary Mathematics, vol. 235, ed. by Walter A. Carnielli and Itala M.L. DOttaviano, 3352.
American Mathematical Society.
50. Brunner, Andreas, and Walter A. Carnielli. 2005. Anti-intuitionism and paraconsistency. Jour-
nal of Applied Logic 3(1): 161184.
51. Goodman, Nicolas D. 1981. The logic of contradiction. Zeitschrift fr Mathematische Logik
und Grundlagen der Mathematik 27(8/10): 119126.
52. Priest, Graham. 2009. Dualising intuitionistic negation. Principia 13(3): 165184.
53. Baskent, Can. 2013. Some topological properties of paraconsistent models. Synthese 190(18):
40234040. doi:10.1007/s11229-013-0246-8.
54. Grant, John, and V.S. Subrahmanian. 2000. Applications of paraconsistency in data and knowl-
edge bases. Synthese 125(12): 121132.
55. Fitelson, Branden. 2012. Accuracy, language dependence, and Joyces argument for probabil-
ism. Philosophy of Science 79(1): 167174.
56. Fitelson, Branden. 2006. Logical foundations of evidential support. Philosophy of Science
73(5): 500512.
57. Fitelson, Branden. 2007. Likelihoodism, Bayesiansim, and relational confimation. Synthese
156(3): 473489.
Index

A BHK interpretation, 384

Ackermann, W., 2, 361 Blok, W., xiii, xiv, 43, 119, 121, 129, 130,
Aether theory, 373 135, 136, 139, 140, 144, 145, 151,
AGM theory of Belief Change, 384 155, 171, 180, 193, 199, 207, 209,
Algebra, 121 210, 237
BL, 193 Bcher, M., 381
MTL, 193 Bochvar, D.A., 10, 138, 143
De Morgan, 213 Bohr, N., 324, 371, 374
Lindenbaum Boole, G., 2, 129
of TM4, 228 Brady, R., 348
Ockham, 279 Brouwer, L.E.J., 13, 14
t-norm Bueno, O., 324
continuous, 192 Bueno-Soler, J., 230
left-continuous, 192 Burali-Fortis paradox, 360
tetravalent modal, 213 Burali-Forti, C., 360, 361, 365
Almukdad, A., 179 Byers, W., 364
Alves, E.H., 35, 40, 293
Anellis, I., 121
Aristotle, vii, xi, 11, 12, 16, 17 C
Arruda, A., 347, 354 Cantor, G., vii, xvi, 3, 345347, 349, 360
Asenjo, F., 11, 149, 383 362
Avron, A., 39, 40, 251253, 256, 260, 272, Cantors Theorem, 346
279, 293 Cardano, G., vii
Axiom of separation (Aussonderung), 347 Cardinal of a first-order signature, 316
Cardinal of a first-order structure, 316
Carnap, R., 2
B Carnielli, W.A., 141, 158, 280, 383
Baaz, M., 112 C-system, 73
Bar-Hillel, Y., 2 Chisholm, R., 233
Bar-Hillel-Carnaps paradox, 2 Chomsky, N., 382
Batens, D., 77, 141, 232, 347 Chuaqui, R., 142, 158, 324
Beall, J.C., 24 Church, A., 52
Belnap, N., xiv, 171, 176, 211214, 216 Closed theory, 36
Bernoulli, J., 380 Cohen, K.J., 10
Berry, G.G., 365 Cohen, P., 3
Bziau, J.-Y., 115, 116 Columbus, C., 11
Springer International Publishing Switzerland 2016 391
W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5
392 Index

Commutative monoid, 192 Dialogical logic, 385

Complex number, 380 Dirac, P., 365
logarithm of a, 380 Distinguished truth-values, 122
Complexity DOttaviano, I.M.L., xiii, 141, 142, 282, 293,
of a formula, 33, 50, 7880, 294 383
of a term, 294 Dugundji, J., 63, 251
Concepts, 2, 52 Dugundji-like theorem, 126, 129
Congruence, 130 w.r.t. Nmatrices, 253, 272
compatible with a theory, 131 Dugundjis Theorem, 121, 122
logical, 131 Dummet, M., xiv, 191
trivial, 131 Dunn, J.M., xiii, 141, 176, 211214, 216
Coniglio, M.E., 142, 158, 280
Consistency
for sets, 349 E
retropropagation, 359 Ehrenfest, P., 371
operator, 32 Einstein, A., 373375, 377379
in a core fuzzy logic, 198 Empty set
propagation, 104 strong, 350
w.r.t. a subsignature, 205 weak, 363
strong, 115 Equality relation, 318
retropropagation Esteva, F., 191, 193
strong, 132 Euclid, vii
Context Euler, L., vii, 380
for a formula, 303 Evidence, 370, 384
for a term, 303 Ex Contradictione Sequitur Quodlibet
Cornwall, J., 2 (ECSQ), 2
Explosion law, 31
Currys paradox, 354, 379

F
D Fidel structure
Da Costa, N.C.A., xiii, 3, 9, 11, 25, 35, 40, for bC, 248
48, 71, 104, 111, 121, 141, 142, 144, for Ci, 248
145, 158, 205, 279, 282, 293, 324, for Cil, 249
347, 348, 363 for Cila, 250
Dauben, J.W., 347, 360 for mbCci, 244
dC-system, 73 for mbCciw, 242
de Amo, S., 141, 158, 383 for mbCcl, 245
De Clercq, K., 141 for mbC, 238
Dedekind, R., xiv, 345, 360, 361, 379 for N4, 181
Deduction Meta-Theorem (DMT), 34, 298, Lindenbaum
300 for mbCciw, 243
De Morgan, A., xiv, 129, 179 for mbCci, 245
Dempster-Shafer theory, 385 for mbCcl, 246
Derivability Adjustment Theorem (DAT), for mbC, 241
46, 190, 207, 220, 352 for N4, 183
Derivation Fidel, M., xiv, 40, 180, 181, 184, 238, 256,
in mbC, 34 279
in TM4, 227 Figallo, A., 224
Descartes, R., viii Figallo, M., 217, 218
Designated truth-values, 122, 252 Finsler, P., 361
Diagram language, 303 First-order language, 294
Dialetheia, 17, 362, 376, 385 First-order signature, 294
Dialetheism, 3, 17, 25, 151, 369, 382 First-order structure, 302
Index 393

extended, 303 for QCi, 323

Fodor, J., 192 for QmbC , 304
Font, J.M., 135, 214, 217 canonical, 313
Frege, G., viii, x, xvi, 2, 5, 12, 13, 345, 346, for QmbC , 320
364, 380 for QP1, 336
Fresnel, A.J., 373 Ivlev, Y.V., 40, 251

G J
Galle, J., 375 Jansana, R., 135
Game-theoretical view of paraconsistency, Jaskowski, S., 7, 10, 136, 140, 141, 229
385 Jevons, W.S., 129
Gauss, C.F., vii Johansson, I., xiii, 171, 175, 179, 187, 221
General relativity, 374 Jourdain, P., 347
Generalized functions, 365
Gdel, K., xiv, 3, 191, 200, 379
Gdels Incompleteness Theorems, 346 K
Godo, L., 191, 193 Kanamori, A., 361
Kant, I., 19, 25, 371, 372
Kapsner, A., 10
H Kearns, J.T., 251
Hadamard, J., 347 Kleene, S., 11, 149
Hjek, P., xiv, 191, 193 Klein, F., 380
Halldn, S., xiii, 10, 136, 138, 159 Kolmogorov, A., 175
Hansson, S.O., 384 Knig, J., 361
Harman, G., 6 Kripke model
Hawking, S., 372 for IPL+ , 172
Hegel, G.W.F., 17, 19, 25 for imbC, 188
Henkin theory, 311 for LFIs, 230
Henkin, L., 231 for Min, 176
Heyting, A., 14 Kripke, S., xiv, 173
Hilbert, D., 2, 360, 380 Kronecker, L., 361, 365
Hintikka, J., 381
Humberstone, L., 176
Husserl, E., 380 L
Hyperalgebra, 251 Landini, P., 224
Hyperoperation, 251 Lattice
Hyperstructure, 251 bounded
comutative, 193
distributive, 216, 222
I De Morgan, 213
Identity relation, 318 implicative, 181
Imaginary number, 380 residuated
logarithm of a, 380 comutative, 192
Implication divisible, 192
contrapositive, 224 prelinear, 193
deductive, 31 Lavoisier, A.-L., 376, 377
Inconsistency operator, 95 Leibniz, G., 380
for sets, 355 Leibniz, G.W., 371, 380
in a core fuzzy logic, 208 Lev, I., 39, 40, 251
Inconsistent mathematics, 365 Level valuations, 251
Inconsistent set, 347 Le Verrier, U.J.J., 374, 375
Interpretation Lewin, R.A., 145
for QLFI1 , 333 Liar paradox, 379
394 Index

Libert, T., 364 ukasiewicz, J., xiv, 10, 16, 121, 141, 142,
Lindenbaum, A., xiii, 37, 130, 131, 227 149, 159, 200, 229
Logic
adequate for an Nmatrix, 252
(Blok and Pigozzi) algebraizable, 130 M
classical propositional, 44 Maddy, P., 361
discussive, 141 Marcos, J., 141, 155, 158, 280, 282, 383
finitary, 30 Mardens Theorem, 381
fuzzy Marden, M., 381
basic BL, 193 Marty, F., 251
core, 195 Mathematical Fuzzy Logic, 191
gently explosive, 9, 32 Matrix semantics, 122
Johanssons minimal, 175 Maximal non-trivial set w.r.t. a formula, 36
maximal, 155 Maxwell, J.C., 377, 378
strong, 155 Mercury, 374
weak, 155 Mikenberg, I., 142, 145, 158, 324
modal Monteiro, A., xiv, 39, 213
anodic, 230 Monteiro, L., 213
cathodic, 230 Moore, G.H., 381
tetravalent, 214 Mortensen, C., 237, 365
Monoidal t-norm based (MTL), 193 Multialgebra, 251
monotonic, 30 Multifunction, 251, 380
Nelsons paraconsistent (N4), 180 Multivalued function, 251, 380
of deontic inconsistency (LDI), 233 Multivalued operation, 251
of formal inconsistency (LFI), 31
strong, 32
weak, 32
paracomplete, 8 N
paraconsistent, 7 Negation
boldly, 176, 220 classical, 7
paranormal, 39 paracomplete, 8
partially explosive, 176 paraconsistent, 8
Positive classical, 34 strong, 7
Positive intuitionistic, 111 strong (Nelson), 10
preserving degrees of truth, 196, 214, Nelson, D., xiii, 10, 106, 108, 160, 171, 179,
222, 223 180, 187, 212, 233, 279
reducible to a signature, 72 Newton, I., 371, 374, 375, 378, 379
standard, 30 Nmatrix, 252
Strict monoidal t-norm based (SMTL), associated to a swap structure
197 for bC, 267
structural, 30 for Ci, 271
Tarskian, 30 for mbC, 255
paraconsistent, 31 for mbCci, 264
Logical matrix, 122 for mbCciw, 261
Loparic, A., 112 Non-deterministic algebra, 251
Lorentz, H.A., 373, 378 Non-deterministic matrix, 252
s, J., 37 Norm
Loureiro, I., 213, 224 continuous, 191
Lwe, B., 142
Lowenhim, L., xv
Lowenhim-Skolem Theorem for QmbC O
downward, 318 Odintsov, S., 180, 181, 184, 238, 253
upward, 318 Ono, H., 180
Index 395

P R
Paraconsistent Belief Revision, 384 Reduct of a first-order structure, 316
Paraconsistent probability, 385 Rescher, N., 251
Paraconsistent set theory, 347 Restall, G., 24
Paraconsistent updating, 385 Rius, M., 214, 217
Paraconsistent version of Bayes Theorem, Rivieccio, U., 180
385 Robinson, A., 4
Partial relation, 325 Routley, R., 348
Partial structure, 325 Russell set, 363
associated to an interpretation for strong, 353
QLFI1 , 333 Russells paradox, 346, 379
associated to an interpretation for QP1, Russell, B., 2, 5, 24, 346, 360, 361, 363365
342
extended, 330
for LFI1, 325, 326 S
for QP1, 338 Schrder, E., 129
Partial truth, 324, 382 Schumacher, vii
Pauli, W., 371 Schtte, K., 141
Peirce, C.S., 121, 129 Schwarz, L., 365
Phlogiston, 376 Schwarze, M.G., 145
Phlogiston hypothesis, 376 Segerberg, K., 138, 141, 145
Sette, A.M.A., xiii, 144, 145, 155, 158, 335
Pigozzi, D., xiii, 43, 119, 121, 129, 130, 135,
Sextus Empiricus, x
136, 139, 140, 144, 145, 151, 155,
Siebeck, J., 381
171, 180, 193, 199, 207, 209, 210,
Signature
237
, 30
Polarities, 211
+ , 30
Popper, K., 10
0 , 30
Possible translations algebraizability, 119
1 , 30
Possible translations semantics, 280
2 , 30
for bC, 281
 , 30
for Cila, 283
 , 30
for Ci, 281
c , 30
for mbC, 281
Silvestrini, L.H., 142, 158
for mCi, 281
Skolem, T., xv
Pragmatic truth, 324, 382
Slater, H., 23
Priest, G., xiii, 2, 11, 25, 150, 151, 383 Snapshots, 255
Priestley, J., 376 Society semantics, 289
Principle of (unrestricted) Abstraction, 346 Splicing of logics, 289
Principle of Comprehension, 346 Splitting of logics, 289
Principle of Explosion (PE), 2 Stokes, G.G., 373
Proper class, 361 Subcontraries
Proposition surrogates, 211 sentences, 23
Propositional language, 30 Substitution, 30
Propositional signature, 29 Lemma
Propositional variable, 30 for QmbC , 306
for QmbC , 320
for QP1, 337
Q multiple, 303
Quantum Theory, 374 Swap structure
Quasi-matrices, 40 for bC, 267
Quasi-truth, 324, 382 for Ci, 270
Quesada, F.M., 11 for mbC, 254
Quine, W.V.O., 5, 16, 24, 347 for mbCci, 263
396 Index

for mbCciw, 260 for mbCcici , 89

Sylvan, R., 151, 348 for mbCciwci , 70
for mbCcicl, 88
for mbCciw, 64
T for mbCcl , 84
Tarafder, S., 142 for mbC ci, 95
Tarski, A., xiii, 5, 121, 130, 131, 227, 324 for mbCcici cl, 89
Theory for mbCci, 66
contradictory, 7 for mbCcl, 82
explosive, 7 for QCi, 323
trivial, 7 for QmbC , 303
Thomason, R., 180 for QmbC , 320
Translation between logics, 43 for QP1, 335
conservative, 43 for PC, , 231
Truth-preserving semantics, 194, 214 for PC , 231
T-(triangular) norm, 191 for PI, 232
left-continuous, 191 over an F-structure for bC, 248
Twist structure over an F-structure for Ci, 249
for N4, 184 over an F-structure for Cil, 249
full, 184 over an F-structure for Cila, 250
Lindenbaum over an F-structure for mbCci, 244
for N4, 185 over an F-structure for mbCciw, 242
over an F-structure for mbCcl, 246
over an F-structure for mbC, 239
U over an F-structure for N4, 182
Universal set, 353 over an MTL-algebra, 194
Uranus, 374 over an Nmatrix, 252
over a matrix logic, 122
over a swap structure for mbC, 255
V
over a twist structure for N4, 184
Vakarelov, D., 180, 184, 279
Variant formulas, 295
Valuation
Vasiliev, N.A., 10
extended, 305
Von Seeliger, H., 374
for LFI1 , 160
Vulcan, 375
for QLFI1 , 333
Vulcan hypothesis, 375
for P1 , 146
for Cn , 113
for Cio, 132
for Ciore, 133 W
for LFI2, 152 Wajsberg, M., 91, 144
for LPT0, 166 Weber, Z., 348
for mbC ciw, 94 Weir, A., 361
for mbCclND, 275 Whitehead, A., 381
for mbC , 53 Wjcicki, R., 43, 122, 196
for mCi, 102
for mbC, 35
for mbC , 97 Z
for mbC cew, 99 Zamansky, A., 293
for mbCcew, 99 Zermelo, E., 347
Index of Logic Systems

Symbols CPL , 45
C1+ , 116 CPLW , 52
C1s , 110 LFI1 P , 330
C2 , 75 LFI1 , 159
C , 112 LFI1 , 160
Cn , 113 P1 , 146
Cn , 293 QLFI1 , 332
Cn= , 293 QP1, 335
CLim , 115 QP1 P , 341
M4m , 214 QmbC , 319
PS3 , 144 mbCciw, 64
M4mc , 225 MTL, 193
P1 , 52 mbCcew, 99
F OU R, 212 CLuN, 77
W3+ , 282 CPL, 44
LBDL , 222 Ci, 103
LDMA , 223 Cil, 103
LDML , 222 Cilo, 116
LTMAc , 225 Cilore, 154
LTMA , 214 Cio, 131
Ciaeci , 123 Ciore, 133
Cilaeci , 126 H3, 136
Cl , 128 J3, 141
mbC ciw, 94 K3, 149
mbC ci, 95 LFI1, 159
mbC cew, 99 LP, 150
mbC , 96 LPQ, 151
mbC , 77 LPT0, 165
mbCcici , 89 MPT, 163
mbCciwci , 70 MPT0, 164
CPL+ , 34 N4, 180
mbC , 51 P1, 145
M B , 215 PI, 77
T ML, 217 QCi, 323
BD, 212 QmbC , 295
mbCcici cl, 89 S3, 138
Springer International Publishing Switzerland 2016 397
W. Carnielli and M.E. Coniglio, Paraconsistent Logic: Consistency, Contradiction
and Negation, Logic, Epistemology, and the Unity of Science 40,
DOI 10.1007/978-3-319-33205-5
398 Index of Logic Systems

TM4, 226 K,, , 231

3, 142 K, , 231
bC, 103 K , 231
imbC, 188 PC, 231
mCi, 102 PC, , 231
mbC, 33 PC , 231
mbCcicl, 88 PI , , 232
mbCcl, 82 BL, 191
mbCclND, 275 SMTL, 197
mbcCe, 103 Ci , , 233
LPT1, 383 LFI1, 159
ZFCil, 356 LFI2, 152
ZFCi, 356 bC , , 233
ZFmCi, 355 mbCcl , 84
ZFmbC, 349 mbC , , 233