Universal Logic

Jean-Yves Beziau'

"Un mathemaiiclen, un mathematicien modeme en particulier, se trouve,

dirait-on, a un degte superieur de factivite consciente: il ne s'Interesse pas
seulcmcnt a la question de quot, mais aussi Ii celle du comment. It ne se
borne presque jamais a une solution-tout court-d'un probleme, il veut
avoir toujours les solutions les plus ...1es plus quoi?-les plus faeiles, les
plus courtes, les plus generales, etc."

A.Lindenbaum

Abstraet
!,\­

;~
'¥,! Universal Logic is a general siudy of logic in Ibe same way as Universal
Algebra is a general study of algebra. It is based on the fact that there is no One
'1
w Logic or Absolute Laws of Logic, but rather a type of logical structures who are
fundamental mother structures in the sense of Bourbaki. Logic is then an
autonomous field of mathematics, with its own intuitions and concepts and
which can survive and be developed without importing specific notions from
other fields of mathematics.

After criticizing the Polish approach, we give a definition of logical

structures and we show what kind of work can be done within this framework
studying the abstract form of the completeness theorem. A central notion is
emerging, the concept of excessive theory, which is the start of a new
"boumologle".

[Work supported by a grant of (he Swiss National Research Fund. The Author would like 10
thank N.C.A. da Costa and J.zygmuol for stimulating conversations and R.Sylvan for correcting
various errors in the ten and making some inleresling comments.
 i~,
':,~

Universal Logic J. Y. Beziau 75

74
We can jump from absoluteness to relativity, saying that they are no absolute
Contents laws but relative laws and various logics, but if we stay at this level we are still
I. What is Universal Logic? at the "religious stage": we have just swapped monotheism for polytheism (and
1.1. The Religious Stage this is not necessarily progress).
1.2. The example of Universal Algebra
2. The Architecture of Mathematics 1.2. The example of Universal Algebra
2.1. Logic and Mathematics
We can give a more precise account to the rejection of the law-approach,
2.2. Structures: Species and types
2.3. Ontology and Philosophy of Modem Mathematics based on the development of Universal Algebra.
3. Logical Structures In Algebra we can distinguish three levels relatively to the degree of
3.1. Critical Analysis of the Polish Approach abstraction (see Figure I). In the more concrete case an algebraic structure is
3.2 Definition of Logical Structures a structure with specific elements, according to the nature of these objects,
4 Universal Logic, an example: Abstract Completeness certain laws are valid. A first step in abstraction makes us forget the nature of
4.1 Valuations as Homomorphisms/Valuations as Maximal Theories the objects and we keep only the laws they obey, then we are into Abstract
4.2 The most Abstract Form of the Completeness Theorem Algebra.
4.3 The Emergence of the Concept of "Excessivity" The second step in abstraction, leading to General Abstract Algebra or
Universal Algebra, was taken by G. Birkhoff, when he decided to consider as
1. What is Universal Logie? an algebra any set with operations; as he explained (see e.g. his paper of 1946),
this was the only way of reaching unity among all algebras. This means that,
this time, abstraction of the laws has been realized and we are left with any
1.1. The religious stage
kind of functions. It may seem much too general. But Birkhoff succeeded in
During this century we have been witnessing the birth of a non denumerable
proving a lot of interesting results (in his famous papers of 1933 and 1935); that
infinity of new systems of logic. Some people think that most of these systems 't'
was not the case of A.N. Whitehead who wrote the Iirst monograph on
are meaningless formal mind games, which have importance only for those who Universal Algebra forty years earlier (1898) but who, according Gratzer (see his
are toying with them. Some people say that each system has its own value and
monograph, p.vi), "had no results", though he "recognized the need for universal
importance, that "classical logic is wrong", that if we want to buy cigarettes, or
algebra" .
to play with very small objects, e.g, des petits grains de poussieres, we cannot
use it. In fact what Whitehead called Universal Algebra will better be ealled today
Are there many logies or only one? and what is Logic? only Abstract Algebra. He was still at the second level considering as absolutely
-\:, valid an algebraic law such as the law of associativity. But his book is very
Traditionaly, logic was conceived as the science of the laws of reasoning: in iii
} interesting from the point of view of the philosophy of mathematics because he
the same way as the bird in the sky is flying according to certain rules, our
anticipated the idea of universal algebra by his search for generality, and
reason is moving aecording to a determined pattern (see e.g, the Logik of E.
therefore has strongly conlributed to its emergence.
Kant, Introduetion, § I The concept of logic).
Following this scheme we can look for the absolute laws of thought in the
same way as we are looking for the general equation of the universe which will
explain everything, give the key to the mystery of the world.
~s attitude seems rather nai've and primitive, and according to the
~ennmology of Auguste Cornie belongs to the "religious phase" of mankind, that
IS to say, the first and less developed stage of human being.
;:( ,
I

76 Universal Logic J. Y. Beziau 77

The Way to Universal Algebra contradiction, excluded middle, identity, or whatever categorical law from the
sky.
The three levels
Abstraction To explain exactly what Universal Logic is we must therefore explain what
is a logical structure, this will be the task of the next section.
General abstract algebra
A Groupoid Second level of 2. The Architecture of Mathematics
JI ~ (A;r) abstraction:
r is a binary function on A abstraction of the 2.1. Logic and Mathematics
laws
The relations between logic and mathematics are multiple and ambiguous.
Abstract algebra For example, is mathematical logic the logic of mathematics or the
A Group mathematic(s) of logics?
(II~(G;*) This dual aspect is well represented in the history of modem logic, with on
*is an associative operation one hand the Fregean approach and on the other hand the Boolean approach.
with neutral object and symmetric
The Fregean project was not only to study the logic of mathematics but to
First level of reduce mathematics to logic, and for him logic was not exactly mathematics but
Speci lic algebra
abstraction: rather a symbolic representation, a language more precise and more perfect than
The Group of Integers
abstraction of the the usual one.
J'~(Z;+)
elements At the opposite Boole tried to give a mathematical account to the laws of
:i; thought in general (not only of the mathematical thinking), expressing them in
Figure I a way similar to the laws of algebra.
One can think that the Fregean approach is not sufficiently mathematical
because his II language' is not sufficiently mathematical, and one can think that
Remark: if we go further and drop functionality we are at the level of the ;1':·K' the Boolean approach is not sufficiently mathematical because it cannot be used
general theory of structures. --,
'f;' for the description or the foundations of mathematics, as Louis Couturat wrote
It appears thatthc lesson of Universal Algebra is that: "
:~ at the end of L'Algebre de la Logique (1905): "One can say that the Algebra of
_ there is no One Algebra who will be the Queen of Algebra, .~ Logic is a mathematical Logic, by its form "'If' method; but one must not
'-,'
_ there is no Absolute Laws of Algebra which will rule the whole Kingdom consider it as the Logic of Mathematics:"
of Algebra for eternity, and even for the present time. If one is interested in the logic of mathematics rather than in the mathematic
of logies, the Fregean approach sounds better, that's why Frege rather than
Now let us point out the main features of the Universal Approach which appear
Boole is considered as the Father of modem logic, because he is taken as the
in Universal Algebra and which will be our guides for Universal Logic:
founder of lirst-order logic,' and, on the contrary, Boolean logic is assimilated
- unity to propositional calculus (J.Van Heijenoort, especially, has contributed to the
- generality deification of Frege and to the rejection of Boole, see e.g, his posthumous paper
- abstraction of 1992).
~ undetennination But in fact it is misleading to reduce Boole to propositional calculus. It
We say that Universal Logic is a general study of logical structures in the same appears that the Boolean project was very important and not only for logic, but
way as Univeral Algebra is a general study of algebraic structures. In particular also for mathematics in general. Boole has strongly contributed to the
there will be no Paramount Logic, such as Classical Logic, Intuitionistic Logic, development of Universal Algebra and to the modem conception of
or whatever Polar Logic, I and no Absolute Law of Logic, such as the law of mathematics. As Whitehead says in his book on Universal Algebra, Boole was

~
I

I'I
 ",:., .
I ,

78 Universal Logic J. Y. Beziau 79

the first to consider that the laws of algebra are not only concerned with criticized by Jevons (Jevons was right and today the addition in a Boolean ring
numbers or quantities, and this led to the notion of abstract algebra and is interpreted as the symmetrical difference and the disjunction is only an
furthermore to the notion of abstract structure where the nature of mathematical indirect operation defined from the other operations). This illustrates an
objects are undetermined. Thus Boole was one of the precursor of the notion of important point: if we are not keeping contact with the basic intuitions of logic
mathematical structure, concepts without which flrst-order logic is merely a and we are applying any mathematical tools already well-working in other fields
game for blind people. of mathematics we will get something which is not wbat we sougbt (this recalls
And if one is interested in the mathematic(s) of logics the Boolean project the famous story of the man looking for his lost key not where he lost it but
seems mucb more appealing that the Fregean project, even if one is interested where there is light). In a preceding paper (1994b) we bave pointed to one
only in the logic of mathematics, because the logic of mathematics is worth example of this abuse: it seems nice, from an algebraic point of view, that the
studying mathematically, that is to say in the spirit of mathematics whicb is replacement theorem holds in a logic, thus many people are studying intensional
probably not only a language, but rather the study of abstract objects given and logics in which the replacement theorem holds, without paying attention that this
sbaped by structures. mathematical property is antinomic with the coneept of intensionality.
This mathematical approach to logic has been developed especially in Of course it is always illwninating to do some importing-exporting between
Poland. Lindenbaum considered that the Fregean-style language for the different mathematical fields, but if this is to be valuable, the different fields
propositional calculus was in fact a mathematical structure, an absolute free must be really different and must bave their own concepts and intuitions.
algebra, and then Tarski reduced the Fregean-style propositional calculus to a • I
Universal Logic as the mathematic of logics may appear disgusting for
Boolean algebra (a construction known as a Lindenbaum algebra or Tarski­
those preoccupied with the problem of the foundation of mathematics, which is
Lindenbaum algebra).' From this point of view it seems that wbat was a
"~ not playing a "fundamental" role in Universal Logic, if it is playing a role at all.
drawback was not the Boolean approacb (as J. van Heijenoort claimed in the
For example, J. Porte in his book Recherches sur la theorie generaie des
paper already referred to), but the Fregean one.
systemes formels (1965) which is written in the spirit of Universal Logic, writes
A lot of tbings have then been worked out whicb bave been collected in the
(p.2): "A lot of people consider mathematical logic as the study of the
book of H. Rasiowa and R. Sikorski with a suggestive name: The mathematics
philosophical problem of 'the foundation of mathematics'. This problem will not
ofmetamathematics (1963). In fact this mathematical approacb to classical logic
be studied, [ will even not discuss the question of whether it has a scientific
was naturally extended to other logics, mainly because the mathematical
meaning or not,"
concepts used for the study of classic logic were easily adaptable to study other
However we must emphasize that Universal Logic has an interest for the
logics, if not a suggestive everlasting source of birth for new logical systems.
philosophy of mathematics and the foundation of mathematics but in a sense
And following the book of 1963 there was An algebraic approach 10 non­
which has nothing to do with the logicist, formalist or intuitionistic approaches,
classical/ogics by H. Rasiowa in 1974. and the logically oriented philosophy of mathematics.
Some people may criticize this orientation of mathematical logic saying that
Foundation of mathematics does not necessary means consistency and the
it is rather algebra than logic, and they are not necessary wrong. The problem
/,~~
reduction to a minimal formal linguistic system. It can also be taken as the
is to know whether or not logic can stand by itself as an independent part of
understanding of what is the nature of mathematics. Universal Logic will show
mathematics. us, from within (like Universal Algebra also does), as a part of mathematics,
We will argue that logic must be an independent mathematical field, that it
how such kinds of process so "fundamental" in mathematics, as those of
needs to have its own mathematical concepts which are not necessarily the same
abstraction and generalization, work.
as those of other mathematical brancbes. For example there is no reason that the :~~
notion of ultrafilter should be a key concept of logic. In §4 we will sbow that J! 2.2. Structures: Species and types
in fact it cannot be a key concept of logic and that a different key concept is
emerging in logic: the concept of excessive theory. ~ Now let us see how Logic can lake its place within mathematics.
. In fact OtC defect of an algebraisation of logic already appeared in the work According to Bourbaki, Mathematic is the study of mathematical structures,
of Boole who proceeded 10 identify "or" with the addition, for this he was r)' and there is no "s" at the end of the word because he tbinks that Mathematic is
I

80 Universal Logic J. Y. Beziau 81

not a random collection of various different things but that it has an than as a fundamental axiomatic basis from which everything on earth can be
"Architecture" (see his famous paper of 1948). derived.
There are three basic mother-structures: algebraic structures, topological Architecture of Mathematics
structures, structures of order. All the other mathematical structures can be
General Theory of Structures
constructed as "cross-structures" from the fundamental structures.
The basic structures are in fact very general, they just represent a "way of
thinking", there is an algebraic way of thinking, a topological way of thinking, Indetermination
an order-like way of thinking. A way of thinking is connected to various Abstraction
Structures Algebraic Topological Simplicity
intuitions and representations which are shaped into concepts. At this level those
of Structures Structures
are rather undetermined concepts. Then, when we go down for complexity t we
Order
reach specific well-determined structures.
~
A specific class of structures, like for example Boolean structures, can be
~
reached in various ways. Each way is a different way of looking at the same I;' a
class of structures. The class itself, independently of the way of looking .1 u,
~. -<i
will be called a species of structures. By opposition the notion of type is
connected with the way of looking. ~
]
..II J 1
Let's take another example, the Axiom of Choice can be expressed in many
(example)
ways, each formulation has a certain type, and the thing they all refer to is the
species. The Axiom of Choice in its Kuratowski-Zorn's formulation has an Species Determination
order-like type, in its Zermelo's formulation it has a function-like type. Each Decidability
formulation is based on different intuitions and concepts. And to prove the Extensional Completeness
equivalence of two different formulations is not necessarily trivial. Categoricity

To consider species independently of types is "typical" of the extensionalist

.i,}:~
approach. But at the "fundamental" level, the extensionalist approach is Figure 2
meaningless, because we are left with types. The different ways of looking do
not reveal the same things, because there is no categoricity or completeness. ,7
Thus they tum to be more important than the things they are revealing. They are ~::;:

like telescopes or microscopes which give partial information and that we can
-,;." Within this description of the Architecture of Mathematics we will consider
Logic as a class of fundamental structures.
use to reach detenninated partial informations about all kind of things. This is
This means that Logic corresponds to a way of thinking which is different
like seeing everything in black or white or in red and blue.
from the other three fundamental ones, that there are typical mathematical
Therefore we can see that they are really two different approaches in the concepts for logic which are related to typical ideas and intuitions.
"foundations of mathematics" which have totally different meanings: the
extensionalist one, set theory, and the intensionalist one, category theory (which
is not at all concerned with "catagoricity"). Of course it is possible from 2.3. The Ontology and Philosophy of Modern Mathematics
category theory to go down to determination, but this not very interesting. On According to the ontology of modem mathematics, mathematical objects are
the other hand it seems that from set theory we cannot go up to undetermina­ objects of structures. What they are depends on the structure of the structure,
tion, although the notion of set we are using in everyday mathematics is rather that is to say on the relations of the structure they are merging in. To know
undetermined, in the sense that most mathematicians don't need to know a lot what a given object is, is to know how it is related to the other objects of the
of set theory because they are using it rather as a general conceptual framework structure.
82 Universal Logic J. Y. Beziau 83

This approach is very illuminating, especially for those who are engulfed in On the other band it is easy to find intuitive reasons to reject each of these
existential torments searching for answers to questions such as: "what is a Jaws, based on concrete examples. And even without any good counter­
number?", "what is as set?". examples we can reject them in view of generality: logic has no better reason
Now this approach will also be radical as regards those never-ending to be monotonic than algebra has to be commutative.
scholastical discussions about propositions, sentences, formulas. statement, etc. This does not mean that we musl rejeet them absolutely, but that basic
To answer the question "What are logical objects?" is 10 provide a logical concepts must not depend on them. And furthermore we must not take them as
structure. a whole, but we must analyse Ibem separately.
All this means also that the philosophy of logic and the philosopby of 3) Finally we can criticize the underlying structure A (and the axiom (4) which
mathematics are not necessary connected with the so-called philosophy of is connected with it). In fact this additional aspect was not here at the beginrting
language or the so-called analytic philosophy, but seem rather closer to the most of the creation of Tarski and was put in only after the war by Los and Suszko
abstract metaphysics of Plato or Descartes as Lautman as strongly emphasized (but this addition was very natural and quite inevitable given that the concept
(see his doctoral dissertation of 1938). of matrix was already vivid during the 30s in Poland). This aspect is wbat is
called "structurality", Given the underlying structure $AI, it seems very natural
3. Logical structures to put the axiom (4). In fact structurality is a perfect expression of the idea that
logical truth doesn't depend on the content but is only a question of fonn. Los
3.1. The Polish approach, and Suszko (see their famous paper of 1958) realized that substitutions were
exactly endomorphisms, and give then a elegant and powerful treatment of what
The first tentative of development of a theory of logical structures was ',:; is generally treated confusedly as a "language", with sucb notion as concatena­
irtitiated by Tarski and was pursued mainly in Poland; it is known as the theory
tion. They have cleaned logic of linguistic accretions and that was badly needed.
of the consequence operator. -,,-
For example, in this light, it appears that what is ealled "Polish writing" is not
A structure of consequence is a structure '6~0i;Cn) where: a "writing" but the simplest (mathematical) representation of an absolute free
- A is an absolute free algebra (A;f",J algebra.
_ Cn is a function from ~A) to ~A) obeying the following four axioms: However nice the concept of strucrurality is, it seems better to reject it at
the most general level, because there are some logics which are not structural
(I) A ~ CnA,
-':r'1 (cf the quasi-formal logics of J.-L. Destouches and P. Pevrier) and because this
(2) A ~ B - Cn A ~ Cn B,
concept can be attacked for philosophical reasons by someone who is not happy
(3) CnCn A c Cn A. ';J
with this song-and-dance form/content approach. But the very reason is that the
(4) (J Cn A ~ Cn (J A, for every endomorphism (J of A.
difference of generality between the study of logics with specified underlying
':;" structures and without is enormous,
Critics l1;: I We call Abstract Logic the approach of logic which doesn't take in account

:..'I~,.'"
1) We can criticize the type of the structure, which looks like a topological type. the underlying structure (thus we don't use this word in the same sense as
This can be an advantage if wc want to apply topology to logic but it is a defect Suszko), Some concept are abstract. some are not. The idea is to work at the
if we want to develop logic by its own. ::'! abstract level, then we will bave some results which will be valid for all
'

2) We can criticize the laws for the consequence operator. Tarski chose the laws ;~~, "languages", infutitary or not, structural or not, propositional or not.
(I), (2), (3) because be had in mind the deductive relation given by what we .,~(~~ We think that we can deal only with structures, without struaurality.
called today Hilbertian systems of deduction. It is well known that a relation of
deductibility induced hy such systems obeys the laws (I), (2), (3). But if we
:~~ In fact the structural approach is closely connected with matrix theory and
thus with algebra. From this point of view it is also a reductionlst approach to
take a more general notion ofsystems of deduction, like Gentzen's systems, they >~~ logic. Then it is better to first develop independent and autonomous concepts
don't necessary hold in general. at the abstract level and to see, after. how we ean put an underlying algebraic­

l
 ,:?i;'""
I

84 Universal Logic J. Y. Beziau 85

like underlying structure. Then it will be interesting to see how our typicallogie We can say thai T« a means thai there is a way leading fmm T 10 a. In
concepts can absorb a part of another field of mathematics. But they will not general, at the abstract level, we don't know which way it is and we don't need
have necessarily to do so, because there are some logics to which it is 10 know. But we have no reason 10 snppose thai this way can be traced by a
meaningless to apply matrix theory. Then the algebraic treatment is just computer, or by any kind of mechanical process.
concerned with strocturality. This is the case of various systems of N.C.A. da Finally it will be useful to give a specific name 10 subsets of L, so we will
Costa. What is being used in this case is the Theory of Valuation of da Costa call them theories.
which is a more general approaeh, directly connected to abstract concepts (see
da Costa/Beziau, 1993, 1994). Criticisms of the proposal
Numerous criticisms can be fonuulated, we shall just select a few and give
3.2. A definition of logical structures
them an answer.
Now we give the following definition of a Logic (or Logical Structure, or
I) This defmition is too much general. This is the most general criticism thai
Afurracr Logic), could be made, and we could simply answer thai it bears the defect it is
A Logic!l is pair <L;r} where: pointing out But lei us be fair. Of course our definition ineludes a 101 of
- L is a set, parasites which ate not and will never be logics in any reaonable sense of the
_ r is a relation over ~)xll. (ie, the cartesian product of the power set of , word. BUI we can go on, without being preoccnpied with that. The example of
ll. and L). :%: Category Theory just shows thai it is not important. How many categories are
The quality and quantity of the elements of L ate left unspecified. not categories of structures? We can say that from the instant thai the basic idea
has been caught, Undelenuination is better than Overdetenuination.
Wbat about given them a name?
2) This defmition is not enough general. For example {a,a} r b identified with
Well, that's not easy. Of course we will not call them "formulas", because
(a I r b. Thus non-contraction logics are not included.
this is a very insidious name. It is connected with the distinction of formjeontent
and thus, with strocturality. It has also linguistie connotations, a fonuula refers '1l There are various ways to rectify this defect One is 10 develop our theory
.'l':';
to a string of signs, generally unintelligible. of abstraet logic within the framework of non-extensional set theory (in the
sense thai the axiom of extensionality is not valid).
In fact it is not necessary to give them a name at the abstract level. the
element of a topological space are called "points", but there is no name for Another way will be 10 consider sequences of elements of L rather than
elements of an abstract algebra, because what is primordial is not the nature of subsets of L, or other structured subsets of L. This solution, which seems the
the elements but the structure there are merged in, thus in the case of algebra, simplest, is however not more general, and we can consider this kind of
what is important is the name "function". or "operation". structures as a particular case of abstract logics.

In logic what will be fundamental is the relation r, the intuitive idea we gel 3) This notion is equivalent to the notion of strocture of consequence.
of it, and how we will call it. We will not call it a "consequence relation" Yes and No. Yes, because for every structure of consequence there is one and
because the word "consequence" is already used in the Polish approach. Some only one logic and vice-versa. No, because they are nOI of the same type. This
use such word as "entailment", or "derivability", "inference" which are rather equivalence will turn to be an advantage because then the notion of structure of
cumbersome and ugly. We will use the name "deducibility", and the expression consequence will be a bridge between logic and topology.
"relation of deducibility", because it is not erroneous to consider logic as the
4) Why not taking another stroeture? A neighbourhood strocture, like a multi­
"theory of deduction". conclusion relation, i.e, a relation on ~)x.9\L)?
There are, at least, two criticisms of this name. First some people will say
that logic is not only deduction but also induction. Seeond some people will say This question can be answered on philosophical basis, saying thai what it is
that deduction is discrete and finite, identifying it with the notion of proof. important is to know which proposition can be deduced from which set of
• ..~~---

86 Universal Logic J. Y. Beziau 87

hypotheses. We don't need ID know all of what can be deduced. The generalities Second, a bivaluation is the characteristic/unction of a maximal theory, and
we need are caught by properties of theories: completeness, consistency, the set of bivaluations is the set of characteristic functions of maximal theories.
c1osureness, etc. That why the multi-conclusion approach seems superfluous.' This second way of looking at a bivaluation is more general and more
J. Porte in his book of 1965 has studied various kind of structures, and his abstract. More general because not every logic has a "homomorphic" semantic,
work is very instructive. Of course, at the beginning, there are necessary some more abstract, because the notion of maximality does not depend on the
tdtonnemeras, the experience and the time will show what is the best solution. subjacent structure, the so-called "language".
Thus the best way to stop critics is to show how our definition works and What is the definition of a "maximal" theory? There is a definition which is
that it works well, this will be our preoccupation in the next section. not an abstract one, depending on the notion of negation, but this definition can
easily be put in an abstract shape. Let us say that a theory is limited if and only
if it is not possible ID deduce everything from it; we call LIM the class of
4. Univeral Logic, an example: Abstract Completeness limited theories of a given logic, then a maximal theory is a maximal object
with regard to the structure of partial order (LIM; c),
Abstraction and generalization can be considered by the mathematician as
the supreme virtue or as the worse defect. Abstraction can lead to enligbtment Imagine that we can prove that for every logic, the set of (characteristic
and best understanding it can also lead to General Abstract Nonsense, triviality functions) of maximal theories is a sound and complete semantic, then we
and madness. would be able ID apply this general theorem in each particular case: checking
if a set of bivaluations is the set of (charateristic functions) of maximal theories.
Our task in this section will be to show that our abstractions are good
abstractions. However as we will see it turns ID be that the concept of "maximalility" is
not the good one, and that we need another concept which is purely logical, by
But first we must say that all this stuff of abstract logic is not a made-in­
opposition to the concept of amaximal set which is quite the same as the
vaiJEAoxoxxuyla product, it has emerged progressively for a wide range of concept of ultrafilter.
logical systems which have been studied and also from various attempt ID
systematize them.
4.2. The most abstract fonn of the completeness theorem
In this section we will show how Universal Logic can be used with regard
to completeness theorem, which is without doubt a central theorem for all A theory T is said a-excessive if and only if it is a-limited (i.e. Tf a) and for
logics. As this paper is rather expository we shall not enter into very technical every b not in T, T, U (bj f- a.
details, which anyway have been worked out in some other papers refered. Given a deducibility relation f- with the two following properties:
[Monotonicity] If Ts- a and T ~ T then T's- a.
4.4.1. Valuation as homomorphimsfValuations as maximal theories
[Compactnessl lf T f- a there exists a finite subtheory To of Tsuch that Tof-a
There are two ways to look at a bivaluation of the classical propositional
we have the following theorem:
logic. First, a bivaluation is an homomorphism from the set of propositions,
taken as an absolute free algebra fi' ~ (P; V, /I, -s-, , ) , ID an algebra !!if ~ «(O,Oj, LINDENBAUM-ASSER lHEOREM Every a-limited theory can be extended
fv, f/l, f-, h), which is equivalent ID the Boolean algebra on 10,I}. The set in an a-excessive theory?
of bivaluations is the set of homomorphisms from fi' ID !!if. ft is possible to From this theorem, we have a completeness theorem for the set of
consider equivalently the set of functions from the set of generators ATOM of bivaJuations of characteristic deductive functions (the characteristic deductive
fi'to {O, I} because it is a property of an aboslute free algebra that each function function of T, 0T is defined as follows: oT(a)~1 iff Ts- a) of excessive theories
from the set of its generators to the domain of an algebra of the same type as (a theory is excessive if it is a-excessive for an object a).
a unique extension which is an homomorphism. COROLLARY: COMPLE1ENESS OF EXCESSIVE SEMANTICS
II ,';' :;,1 i i

88 Universal Logic J. Y. Beziau 89

.J
Proof. If T f a then there exists an a-excessive extension E of T. Let ~T be the theory can be extended in a maximal theory from which a is not deducible,'
cbaracteristic deductive function of E. ~T(1)~ I because E is an extension of T, works only when all excessive theories are maximal, and then it is identical to
and b,{a):O because E f a. the Lindenbaum-Asser theorem.
Remark. Some people will wonder what kind of completeness we are talking
about, especially becausewe are not presenting here any "system of deduction"
(such kind of thing with rules and proofs). But we must recall that systems of
J The concept of excessivity has plenty of good qualities. A powerful
application is its application to sequent calculus.
We have shown (in 'Excessive Theories') that excessive theories preserve the

r
deduction are special cases of abstract logics. rules of structurally standard systems of sequents.
From this and the theorem of Lindenbaum-Asser it results that we can prove
4.3. The emergence of the concept of II excessivity" immediately the completeness of a wide range of logics not necessary truth­
Now we will explain why the concept of rnaximality is not the right concept functional and not necessary Fregean (i.e. where the theorem of replacement
and why the concept of exccssivity is the good one. does not necessary holds).
We have proved (see our paper 'Excessive Theories') that the semantic of For example, let's take a logic with a connective *, called supernatural
excessive theories is a minimal complete semantic, that is to say thata smaller implication, defined by the following rules of sequent calculus (we use - in the
semantic (a class of theories strictly included in the class of excessive theories) same way as Gentzen):
cannot be complete. As it is easy to see that maximal theories are excessive
theories, this means that in the case where they are excessive theories which are
not maximal then the semantic of maximal theories is not complete. r .. G, .6. f',b - 11 I' - a,11 [' • b,.6.
_ _ _ _ _ _ _ _ _*1 _ _ _ _ _ _ _ _ _ _*r,
I', I", (a*b) - 11,11' r, f'- (a*b) /1,11'
Of course it is possible to think that in all "good" logics all excessive
theories must be maximal theories, and that was what most of the people were With our result about excessive theories we can immediately see that a set of
thinking. That's why this notion of excessivity, introduced by G. Asser a long bivaluations IB defined by the following conditions:
time ago, has never been taken seriously, until some people in Brazil were
trying to use the theory of valuation to give a semantic to logics in which there p E IB if and only if:
are excessive non-maximal theories." In fact, it is the case of intuitionistic logic. - If p(a)~ I and P(b)~O then p(a*b)~O,

We even have proved (in 'Excessive Theories') that if the law of Curry (T,-.a - If p(a)~1 and P(b)~1 then p(a*b)~I,
I- a ==::::;. T I- a) was not valid in a monotonic compact logie with a minimal
is a complete semantic tor «.
negation, then there were some excessive non-maximal theories.'
That means we can automatically translate sequent rules into conditions for
Suszko speaking about sets of bivaluations for a given inference relation is a set of bivaluations complete for the logic induced by this set of rules (and the
writing: "The adequate sets V form an interval (V, ~ V c V,) between the converse also holds: we can axiomatize automatically with sequent rules
smallest adequate set V, and the largest one V,. Some are better, some are eonditions for a set of bivaluations, given in a specific normal form).
worse." (in his paper of 1977, p.378).
Intuitively our result shows that it is suflicient to put p(a)~O if a is on the
OUf "minimal result" shows that: left hand of a sequent, and to put p(a)~ I if a is on the right. Furthermore, if a
I) the adequate sets are generally not a linearly ordered, and b occur in two different premises of a rules, like in the example, we put
2) there is no smallest adequate set in general, "and", but if it is on the same sequent we put "or".
3) on certain general conditions, there is a minimal adequate set, the set of We can conclude saying that the concept of excessivity appears as a key
excessive theories, which is, thus, not the worse. concept in Universal Logic, and using the Bourbakian dialect that it is an
It also shows that the version of Lindenbaum's theorem saying every a-limited essential part of a new "boumologie", a logical "boumologie",
 ;';5.' •
~
90 Universal Logic J. Y. Beziau 91

Notes 'This is the version required for completeness of Ihe semantic of maximal
theories, the version saying, every limited theory can be extended in a maximal
lit would be mistaken to lbink that Universal Logic is a Paramount Logic.
theory, is in general not equivalent to it, and in Ibis case doesn't guarantee
completeness (see our paper 'Excessive Theories').
'We must recall that Frege himself did not create what we call nowadays
first-order logic. Frege is taken as the originator of first-order logic because he
Bibliography
has developed the theory of quantification.
Asser, G., Einjiihrung in die mathematische Logik, B.G.Teubner,
aAccording to R.Suszko (see his Introduction to the vol.I02 of Disseruuiones Leipzig, 1959.
Mathematicae), Lindenbaum was the first to consider the language of logic as Beaulieu, L., Bourbaki, Une histoire du groupe de mathematiciens
an algebra: as an absolute free algebra, and according to Tarski (see his book francais et de ses travaca (1934-1944), These, Universite
Cylindric Algebra, p.85, nA), Tarski is the inventor of the so-called de Montreal, 1989.
~denbaum-Tarski algebras. There are two levels in the "realgebrisation" of Beziau, J.-Y., 'Logiques construites suivantles methodes de da
logic by the Polish logicians, which must not be identified, one due to Costa', Logique et Analyse, no. 131-132 (1990), pp,259­
Lindenbaum, the other one due to Tarski. The terminology Lindenbawn-Tarski 2723,
algebra is however correct if we lhink that the idea of Lindenbawn is Ihe first Beziau, J,- Y., 'Recherches sur la logique abstraite: les logiques
step leading to the Tarskian algebraisation. And Ibis must not reduce the role normales', Acta Universitatis Wratislaviensis, Serie Logika,
of Lindenbaum in the development of logic which is not very well-known and no. 16 (1994)
which is probably much more important that we can imagine, bolh on Beziau, J.- Y., 'Du Pont's paradox and intensional logic' in
conceptual and technical sides (see the papers of SJ.Surma). Proceedings of Logica'93, P. Kohii and Y, Svodoba (eds),
Czech Academy of Science, Prague, 1994, pp.62-65.
'Some people will say that Gentzen's sequent c1aculus is a counter-example. Beziau, J.-Y., 'Excessive Theories', to appear.
But in fact Gentzen's systems are systems of deduction which induce logics, and Beziau, J,- Y., 'Bourbaki etla logique', to appear.
generally they are not taken as generating multi-conclusion logics. And, contrary Beziau, J.-Y., 'au the distinctions and confusions between rules and
to some interpretations, Genzten's roles arcnot conditions on a multi-conclusion laws, proof and consequence, Gentzen and Tarski', to
relation of deducibility (see our paper "On the distinctions and confusions appear.
between, rules and laws, proof and consequence, Gentzen and Tarski"), Birkhoff, G" 'Universal algebra' in Comptes Rendus du Premier
Congres Canadien de mathemasiques. Montreal-1945,
JAB it is known this theorem is equivalent to the axiom of choice, this University of Toronto Press, Toronto, 1946, pp.3I0-326.
illustrates our remark in §22: we have here a version of the axiom of choice in Bloom, S.L. and R. Suszko, 'Classical abstract logic', Dissertationes

(
a "typical" logical framework. Mathemaiicae, voLl02 (1973), ppA3-52.
Bourbaki, N., 'L'architeeture des rnathematiques (La mathematique
'For a more detail account, see [da Costa/Beziau 1993]. Asser used the ou les mathematiques")' in Les grands courants de la
terminology "vollstandig in bezug auf' which has been translated by "relatively pensee mathematique, F. Le Lionnais (ed), Paris, Cahiers du
maximal". We don't use it because it is cumbersome. The Brazilians used the Sud, 1948, pp.35-47. [English translation: American
word "saturated", but this word is already used in model theory so we thought ", Mathematical Monthly, no. 57 (1950), pp.221-232]
it was necessary to introduce a new word. Brown, DJ. and R.Suszko, 'Abstract logic'. Dissertationes
Mathematicae, voLl02 (1973), pp.9-41.
'R. Sylvan told the author that in relevant logic the distinction is also
fundamental.
• "~ 2 ,

92 Universal Logic J. Y. Beziau 93

cia Costa, N.C.A. and J.-Y.Beziau, 'La theorie de la valuation en Surma, S.J., 'On the origin and subsequent applications of the
question' in Proceedings of the Ninth Latin-American concept of Lindenbaum algebra' in Logic, methodology and
Symposium on Mathematical Logic, M. Abad (ed), Bahia­ philosophy of science VI, LJ. Cohen, J. Los, H. Pfeiffer
Blanca, 1993. and K.-P.Podewski (eds), PWN, Warsawf North-Holland,
da Costa, N.C.A. and J.-Y.Beziau, "Iheorie de la valuation', 1994, to Amsterdam, 1982, pp.719-734.
~\
appear. Suszko, R., Introduction to Dissertationes Mamematicae, vol.l 02
Couturat, L., L'algebre de la logique, F.Alcan, Paris, 1905.
Destoucbes, J.-L., Cours de logique et philosophie generale; Centre
II (1973).
Suszko, R., 'The Fregean axiom and Polish mathematical logic in
de Documentation Universitaire, Paris, 1948. the 1920.', Studio Logica, vo1.36, pp.373-380.
Fevrier, P., 'Les relations d'incertitudes d'Heisenberg et la logique' in Whitehead, A.N., A treatise on universal algebra, CUP, Cambridge,
Travaux du IXe Congres International de Philosophie, 1898.
vol.Vl, Hermann, Paris, 1937, p.88.
Gratzer, G., Universal Algebra, Springer, New-York, 1968, 1979.
van Heinjenoort, J., 'Historical development of modern logic',
Modern Logic, vo1.2, n 3 (1992), pp.242-255. USP-Universidade de Sao Paulo
Henkin, L, J.D.Monk and A.Tarski, Cylindric algebras, North­ FFLCH-Departamento de Filosofia
Holland, Amsterdam, 1971. Caixa Postal 8105
Kant, E., Logik, 1800. 05508-900 Sao Paulo, SP
Lautman, A., Essai sur les notions de structure er d'existence en Brasil
mathematiques, Hermann, Paris, 1938. Email: beziau@cat.cce.usp.br
Lindenbaum, A., 'Sur la simplicite formelle des notions' in Actes du !
congres international de philosophie scientifique, vol. VII,
I
Logique, Hermann, Paris, 1936, pp.28-38.
Loparic, A., 'Une etude semantique de quelques ealculs \
propositionnels', Comptes Rendus de l'Academie des I
Sciences de Parisi, tom.284A (1977), pp.835-838. I
Los, J., and R.Suszko, 'Remarks on sentential logics', Indigationes
MathemaIicae, voUO (1958), pp.I77-183.
Porte, J., Recherches sur la theorie generale des systemes formels,
Gauthier-Villards, Paris f E.Nauwelaerts, Louvain, 1965.
Rasiowa, H., An algebraic approach to non-classical logics, PWN,
Warsawf North-Holland, Amsterdam, 1974.
Rasiowa, H., and R.Sikorski, 11.. Mathematics of Metamathemaiics, i." I
PWN, Warsaw, 1963. ~
:~;.I
Surma, S.J., 'The concept of the Lindenbaum algebra: its genesis' in
Studies in the history of mathematical logic, SJ.Sunna (ed),
Ossolineum, Wroclaw, 1974, pp.239-254.