You are on page 1of 3

for Lk (with the connectives exactly as in the standard case) dual for M2.

Though very close to M 2 , dM2 Q.iffers fi:om the former in two respects: the
content, E(dM2 ) = {-,o; : o: E TAUT} =/= E(M2 ), and the consequence,
FM2 =I= FdM,
Since FM = FN implies E("M) = E(N), every matrix satisfying (Kl)
also meets the requirements of (K2). At the same time, however, there
exist matrices meeting (K2) but not (Kl). Examples of matrices having
an empty contentare also known, which determine interesting consequence
relations (see Chapter 7). This obviously justifies taking into account the
relations of consequence as well as the two criteria of many-valuedness.
The examples given below illustrate the matter somewhat better.

4. Many-valuedness

The application of specialized tools of the methodology of propositional


calculi (see Wjcicki (1988)) affords possibilities for a profound investigation of many-valued constructions. The present chapter aims to present
sorne important methods of that methodology, especially those revealing
the complexity of the problem connected with a general characterization
of logical many-valuedness.

4.1

4.1.1.E. Consider the matrix

M 3 = ({0,t,1}, ..,, ...... , V,/\,::::, {t,1})


for Lk with the operations defined by the following tables:

Two cri teria

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

o
t
1

= ({O, 1},

= FN (or, equivalently, CnM = CnN)

o
o

t
t
t
t

o
o o

1
1
t
1

o
t
1

t
t
t
1

t
1

o
1

o
o

t
1
t

1
1
1
1

o o
t
t

t
1

Af3=({0,t,l}, ..,, ...... , V,/\,::::, {t,1})


for

ck

with the operations defined by the following tables:

ft
o

..,, ...... , V, /\, ::::, {O})


il

30

t
1

4.1.2.E. Consider the matrix

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


determined by two-element matrices being directly related with the basic
matrix for the classical logic, i.e. having {O, 1} as the set of elements and,
at the same time, 1 as the designated value. To get the idea let us look at
the two-element matrix
dM2

o
o

One may easily verify that this three-element matrix does not determine
many-valued (i.e. three-valued) logic even according to the condition (K2).
Due to the choice of the set D, with each h E Hom(Ck, M 3 ) the valuation
h* E Hom(Ck, Mz) corresponds in a one-to-one way such that ho: E {t, 1}
iff h*o: =l. Therefore FMs = FM2

many-valued logic if and only if for no two-element


E(M) = E(N).
many-valued logic if and only if for no two-element
FM

/\

M =(A, D)

(Kl) M determines a
matrix N for C,
(K2) M determines a
matrix N for e,

......

o t 1
o o o o
t o t t
1 o t 1

"

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


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

..,X

l
1

t
1

1
1

......

o
t
1

o
1
1

t
1
1

o o

1
1
1
1

o t
o o o
t o o

1 .1

1
1
1
1

4. Many-valuedness

32

o t 1
o o o o
t o o o
1 o o 1

/\

o
t
1

1
1

1
1

o
o'

o o

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


matrices the substitutions (i.e. endomorphisms) of the language L take the
role of valuations. Taking this in to account it is easy to verify:

1
4.2.2. (Wjcicki (1970)). Each structural consequence operation C is
uniquely determined by its Lindenbaum bundle: e = CnLc.

One may show that E(M3 ) = E(M2 ) and, accordingly, that M 2 is the
only two-element matrix which might determine FMs. Simultaneously,
FMs f:. FM2 , since, for example,
{p-> q,p}

FM

while not

{p-> q,p}

FMs

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


matrices such that e = CnK.
When a class K of matrices for a given language L. determines the
structural consequence e, i.e.if CnK =e, we shall say that K is adequate
for C. Such C is referred to as strongly finite provided that there is a finite
class of finite matrices adequate for it. Sorne criteria of strong finiteness
using congruences of the language compatible with
(a ,..., f3 implies
C(o:) = C(f3)) can be found in Wjcicki (1977).
A sufficient and necessary condition for a given structural consequence
to be determined by a single matrix formulated in terms of absolute uniformity is due to Wjcicki (1970). Such a C is uniform if for any X, Y, a
such that lr(X,a) n lr(Y) = 0 and C(Y) f:. For, a E C(X U Y) implies a E C(X). In turn, C is absolutely separable if for any family X
of sets of formulae such that for arbitrary X, Y E X, X f:. Y implies
lr(X) n lr(Y) = 0, and for any propositional variable p E Var(LJ X)

q.

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


and hq = O. Summing up, M 3 satisfies the second condition of manyvaluedness but fails to satisfy the first. From the point of view of (K2) it
determines a many-valued logic and in view of (Kl) a two-valued logic.

4.2

Structurality and many-valuedness

CnM considered in 3.3 are special examples of the operations considered


in the general theory of deductive systems originated by Tarski (1936). A
mapping C : 2For -> 2For will be referred to as a consequence opemtion of
the language ! if and only if for any X, Y s;; For

(TO)
(Tl)
(T2)

X s;; C(X)
C(C(X)) = C(X)
C(X) s;; C(Y) whenever

(as)
X

eC(X)

= (!,

implies

pEC(X),

forsome

XEX.

uniform.

4.2.3. (Wjcicki (1970)). For a structural consequence operation Can


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

s;; C(eX),

4.2.4.E. Consider the (structural) consequence C0 determined by the


twp-element class of matrices K = {M2,dM2} on Lk, i.e. Co = CnK.
Notice that C 0 (0) = 0. Subsequently, take any two formulae having no
variables in common, e.g. a = p V -,p and f3 = q V -,q. Then we have
a E Co(f3)
For and a f. Co(0)
0. This proves that C 0 is not uniform
and hence absolutely uniform and according to 4.2.3 there is no single
matrix adequate for it.

we shall say that e is structural.


It is easy to prove that each matrix consequence operation CnM is
structural. Conversely, each structural consequence C of ! and any set of
formulae X together determine a matrix
Lx

pEC(LJX)

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

Y.

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


(S)

33

Finiteness and deduction

C(X))

called a Lindenbaum matrx.


4.2.1. (Lindenbaum). The content of (!, C(0)) equals C(0).

4.3

Syntactical characterization of consequence operations refers to inference


(associated with each C is a relation l=c such that: X l=c a iff a E
C(X)). Any pair (X, a) will be referred to as a sequent and a set of
sequents as a rule (of nference). Sequents of the form (0, a) will be called

The class of all Lindenbaum matrices of a given consequence C


of !,
LC = {(!, C(X)): X

s;;

Finiteness and deduction

For},

il

4.3.3. If --+ is an implication with respect to a finite and structural


consequence e, then e= Cn(C(0),MP)

axiomatic and a set consisting of only such sequents an axiomatic rule.


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

Proof. The inclusion Cn(C( 0),MPJ(X) ~ C(X) is obvious. Assume then


that a E C(X). From the assumption it follows that there is a finite set of
formulae X1 ~X such that a E C(X1 ). Suppose that X1 = {/3, ... ,/3k}
Since a E C(/31, ... , f3k), applying the deduction theorem k-times (Ded),
we get /31 --+ (/32 --+ ... --+ (f3k-l --+ (f3k --+ a))) E C(0). Therefore,
a E Cn(C(0), MP)(x,) ~ Cn(c(0), MP(X).

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


determined by the sequent {p --+ q, p} / q and represented through the following schema: P--+ w, P/w.

4.3.4.E. Every consequence having a finite base consisting of finite rules


only (i.e. rules with finite sets of premises) is finite. An important example
of such an operation is the inferential consequence of the classical logic:
C2 = Cn({Al-Al5 },MP) where Al-A15 are schemes of the axioms in 1.4.
Since C2 = CnM2 , C2 is strongly finite as well. That 4.3.3 implies C 2 =
Cn(TAUT, MP) may also be of sorne interest (compare 3.3.1.E).

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


a set of rules of inference). We shall say that X is R-closed iff for every
a E For, Y~ X, (Y,a) E R implies a E X. The operation CnR defined
for every X ~ For by

CnR(X) = l{Y ~ For : X ~ Y, Y being R-closed}


proves to be a consequence (structural if R is structural). Every R such
that C = CnR is referred to as a base of C. In the case when R contains
axiomatic sequents, i.e. when R = RA U R 1 , we may, by putting A = {a :
(0,a) ERA}, represent CnR as Cn(A,Rl) defined for any X~ Foras
follows:

Cn(A,Rt)(X) = l{Y: AUX ~Y and Y is R 1 -closed}.


Then, the elements of A will be called axioms.
A binary connective --+ of the language ! of C, for which the following
conditiou (later referred to as the deduction theorem)
(Ded)

f3 E C(X, a) if and only if a--+ /3 E C(X)

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


way of formalizing propositional logics (!, C) with implication consists in
definiug as Cn(A, MP) or Cn(A, {SUB, MP} )1 where SUB is the substitution
rule, SUB= {P/eP: e E End(!)}. From the perspective of the present
chapter, such a formalization is very handy. Mostly, when the consequence
operation of a logic is finite (see below) the question of many-valuedness
reduces to the satisfiability of the condition (Kl ).
We say that a consequence C is finite if for every X ~ For, a E For:

(Fin)

a E C(X) if and only if a E C(X1 ) for sorne finite X 1 ~X.

4.3.2. (Wjcicki (1977)). Every strongly finite consequence operation


is finite.

",1

!I11

L._

You might also like