-RXUQDORI3KLORVRSKLFDO/RJLF

'XDOHTXLYDOHQWWZRYDOXHGXQGHUGHWHUPLQHGDQGRYHUGHWHUPLQHGLQWHUSUHWDWLRQVIRU
?/XNDVLHZLF]
VYDOXHGORJLF?/
0DQXVFULSW'UDIW


0DQXVFULSW1XPEHU

/2*,5

)XOO7LWOH

'XDOHTXLYDOHQWWZRYDOXHGXQGHUGHWHUPLQHGDQGRYHUGHWHUPLQHGLQWHUSUHWDWLRQVIRU
?/XNDVLHZLF]
VYDOXHGORJLF?/

$UWLFOH7\SH

2ULJLQDO5HVHDUFK

.H\ZRUGV

0DQ\YDOXHGORJLF?/XNDVLHZLF]YDOXHGORJLFWZRYDOXHG
XQGHUGHWHUPLQHGDQGRYHUGHWHUPLQHGLQWHUSUHWDWLRQVSDUDFRQVLVWHQWORJLFV

&RUUHVSRQGLQJ$XWKRU

*HPPD5REOHV3K'
8QLYHUVLGDGGH/HQ
/HQ&DVWLOOD\/HQ63$,1

&RUUHVSRQGLQJ$XWKRU6HFRQGDU\
,QIRUPDWLRQ
&RUUHVSRQGLQJ$XWKRU
V,QVWLWXWLRQ

8QLYHUVLGDGGH/HQ

&RUUHVSRQGLQJ$XWKRU
V6HFRQGDU\
,QVWLWXWLRQ
)LUVW$XWKRU

*HPPD5REOHV3K'

)LUVW$XWKRU6HFRQGDU\,QIRUPDWLRQ
2UGHURI$XWKRUV

*HPPD5REOHV3K'
)UDQFLVFR6DOWR3K'
-RV00QGH]3K'

2UGHURI$XWKRUV6HFRQGDU\,QIRUPDWLRQ
$EVWUDFW

?/XNDVLHZLF]WKUHHYDOXHGORJLF?/LVRIWHQXQGHUVWRRGDVWKHVHWRIYDOXHGYDOLG
IRUPXODVDFFRUGLQJWR?/XNDVLHZLF]
VYDOXHGPDWULFHV)ROORZLQJ:RMFLFNLLQ
DGGLWLRQZHVKDOOFRQVLGHUWZRDOWHUQDWLYHLQWHUSUHWDWLRQVRI?/?WH[WTXRWHGEOOHIWZHOO
GHWHUPLQHG?WH[WTXRWHGEOULJKW??/DDQG?WH[WTXRWHGEOOHIWWUXWK
SUHVHUYLQJ?WH[WTXRWHGEOULJKW??/EGHILQHGE\WZRGLIIHUHQWFRQVHTXHQFHUHODWLRQVRQ
WKHYDOXHGPDWULFHV7KHDLPRIWKLVSDSHULVWRSURYLGHE\XVLQJ'XQQVHPDQWLFV
GXDOHTXLYDOHQWWZRYDOXHGXQGHUGHWHUPLQHGDQGRYHUGHWHUPLQHGLQWHUSUHWDWLRQVIRU
?/?/DDQG?/E7KHORJLF?/LVD[LRPDWL]HGDVDQH[WHQVLRQRI5RXWOH\DQG
0H\HU
VEDVLFSRVLWLYHORJLFIROORZLQJ%UDG\
VVWUDWHJ\IRUD[LRPDWL]LQJPDQ\YDOXHG
ORJLFVE\HPSOR\LQJWZRYDOXHGXQGHUGHWHUPLQHGRURYHUGHWHUPLQHGLQWHUSUHWDWLRQV
)LQDOO\LWLVSURYHGWKDW?WH[WTXRWHGEOOHIWZHOOGHWHUPLQHG?WH[WTXRWHGEOULJKW??/
XNDVLHZLF]ORJLFVDUHSDUDFRQVLVWHQW

5HVSRQVHWR5HYLHZHUV

:HVLQFHUHO\WKDQNWKHWZRUHIHUHHVRIWKH-3/IRUWKHLUFRPPHQWVDQGVXJJHVWLRQVRQ
RXUSDSHU
,QWKLVILQDOYHUVLRQDOOVXJJHVWLRQVPDGHE\5HIHUHHRQWKHVHFRQGYHUVLRQRIRXU
SDSHUKDYHEHHQDFFHSWHGDQGLQFRURSRUDWHGLQWKHWH[W

Powered by Editorial Manager and Preprint Manager from Aries Systems Corporation

 0DQXVFULSWH[FOXGLQJDXWKRUQDPHVDQGDIILOLDWLRQ

Journal of Philosophical Logic manuscript No.

(will be inserted by the editor)

Dual equivalent two-valued under-determined and

over-determined interpretations for ukasiewiczs 3-valued
logic 3
Gemma Robles Francisco Salto Jos M.
Mndez

Received: date / Accepted: date

Abstract ukasiewicz three-valued logic 3 is often understood as the set of 3-valued

valid formulas according to ukasiewiczs 3-valued matrices. Following Wojcicki, in
addition, we shall consider two alternative interpretations of 3: well-determined
3a and truth-preserving 3b dened by two dierent consequence relations on the
3-valued matrices. The aim of this paper is to provide (by using Dunn semantics) dual
equivalent two-valued under-determined and over-determined interpretations for 3,
3a and 3b. The logic 3 is axiomatized as an extension of Routley and Meyers
basic positive logic following Bradys strategy for axiomatizing many-valued logics by
employing two-valued under-determined or over-determined interpretations. Finally, it
is proved that well determined ukasiewicz logics are paraconsistent.

Keywords Many-valued logic; ukasiewicz 3-valued logic; two-valued underdetermined and over-determined interpretations; paraconsistent logics.

G. Robles
Dpto. de Psicologa, Sociologa y Filosofa
Universidad de Len
Campus de Vegazana s/n
24071, Len, Spain E-mail: gemmarobles@gmail.com
F. Salto
Dpto. de Psicologa, Sociolog y Filosofa
Universidad de Len
Campus de Vegazana s/n
24071, Len, Spain E-mail: francisco.salto@unileon.es
J. M. Mndez
Universidad de Salamanca
Campus Unamuno, Edicio FES
37007, Salamanca, Spain E-mail: sefus@usal.es

1 Introduction

ukasiewicz three-valued logic 3 was dened in a two-page paper in 1920 (cf. [15]).
The philosophical motivation of the proposed logic and the interpretation of the third
truth value are clearly stated: The indeterministic philosophy [...] is the metaphysical
substratum of the new logic ([15], p. 88). The third logical value may be interpreted
as possibility ([15], p.87). 3, understood as the set of all three-valued formulas, was
rst axiomatized by M. Wajsberg in 1931 (see [35]). On the other hand, ukasiewicz
generalized this 3-valued logic to q-valued logics (with q nite) and also to an innitevalued logic in 1922 (see [17] and [33]). An axiomatization of the innite-valued logic
$ was provided by Wajsberg in 1935 but his proof was never published (see [30], [32]).
The rst published proof was that of Rose and Rosser (cf. [24], [33]). Regarding the
q-logics, it seems that they were rstly shown nitely axiomatizable in some cases
by Lindenbaum, and, in general, by Wajsberg in 1935 (see [32], p. 333). Since then,
dierent axiomatizations of the nite-valued q-logics have been proposed. Among
those in a Hilbert-style form, the ones by Tokarz and Tuziak are to be remarked (see
[30], [32]). Tokarz axiomatizes q by adding one-variable axioms to $ as axiomatized
by Wajsberg. Tuziak, on the other hand, presents particular axiomatic systems for
each q by generalizing a completeness theorem by Pogorzelski and Wojtilak (see [32]).
Concerning the axiomatization of 3, Avron provides a simple one which is, in addition,
a well-axiomatization (cf. [1]). We, on our part, will dene a simple axiomatization
of 3 by extending with independent axioms Routley and Meyers basic positive logic
B+ (cf. [20], [25] and [4]).
Now, let us precisely state what ukasiewicz three-valued logic 3 refers to in
this paper.
As it is well known, a logic is, according to the Polish logical tradition, equivalent
to a consequence relation of some kind, provided it fullls Tarskis standard conditions
(cf., e.g., [37], Chapter 1, 2, where Tarskis concept of logical consequence is clearly
explained). As Wojcicki puts it: a logic is dened by its derivability relation rather
than by its sets of theorems ([36], p. 202). In this sense, there are essentially (but not

exclusively), two ways of dening a (semantical) consequence relation given a set of

truth-values V with an ordering $ dened in it, a subset G a V of designated values
and a set of functions F from the set of ws to V:
1. For any set of ws [ and any w D, [ D i for all i M F, if i ([) M G, then
i (D) M G.
2. For any set of ws [ and any w D, [ D i for all i M F, i ([) $ i (D) [i ([) =
inf {i (E) : E M [}].
It is commonly understood that ukasiewicz logics are those determined by the
relation dened in (1). But, as Wojcicki remarks, ([37], 13), the ukasiewicz logics
determined by the relation dened in (2) also deserve to be called ukasiewicz logics
([37], p.42). Actually, the referred author concludes that there are at least two kinds
of ukasiewicz logics: truth-preserving ukasiewicz logics (determined by the relation
dened in (1)) and well-determined ukasiewicz logics (determined by the relation
dened in (2)). Consequently, ukasiewicz three-valued logic 3 is here understood in
three dierent senses:
i. As the set of three-valued valid formulas according to the matrices M3 dened by
ukasiewicz (cf. Denition 3 below).
ii. As truth-preserving three-valued 3 (determined by the relation (1) in M3).
iii. As well-determined three-valued 3 (determined by the relation (2) in M3).
The aim of this paper is to provide a two-valued semantics for the three versions of
3 recorded above with (equivalently) under-determined or over-determined interpretations. An under-determined interpretation is a function from sentences to the proper
subsets of the set {W> I }; and an over-determined interpretation is a function from
sentences to the non-empty subsets of {W> I } ({W } and {I } represent truth and falsity in the classical sense). Thus, under-determined interpretations assign {W }, {I } or
empty set to sentences; and over-determined interpretations assign {W }, {I } or {W> I }.
It will be shown that 3 in the sense (i) can be dually interpreted either by underdetermined or over-determined interpretations. Then, consequence relations equivalent

to those dened in (ii) and in (iii) (referred, of course, to the matrices M3) will be dened by under(over)-determined interpretations. A consequence of these results is that
ukasiewiczs third-value can legitimately be thought of as equivalently representing
either indeniteness or contradictoriness.
The two-valued semantics with either gaps or gluts here dened is based on
Dunns semantics for rst degree entailments (see [ 9], [10]), that goes back to Dunns
doctoral dissertation (see [8]). As noted by Dunn himself ([9], p.150) essentially equivalent semantics are dened in [26] and [34].
At this point, it is interesting to note the following:

Remark 1 (First degree entailment fragment of 3) In [10], Dunn shows that the rst
degree entailment of 3 can equivalently be characterized by either under-determined
or over-determined interpretations.

This is not the place to discuss the dierent interpretations of the q (or innite)
(truth) values of many-valued logic in general and ukasiewicz logics in particular (we
refer the reader to, e.g., [14], Chapter III 5; [18], esp., Chapters 2, 4, 5 and 10; [13], 2.5
and [33] 1.3, 5). Let us however make a couple of remarks on the famous Suszkos
thesis and the so-called Bivalent semantics.
As it is known, Suszkos thesis can be formulated as follows: any logic with a
structural consequence relation operator conforming to Tarskis standard conditions [...]
is logically two-valued ([31], p. 299-300). Concerning 3, the logic we are interested
in in this paper, Suszko dened a two-valued semantics for 3 in [29] (we refer the
reader to Tsujis paper for a discussion of Suszkos thesis and Malinowskis objections
to it). On the other hand, the Bivaluation semantics (or Bivalent semantics) are
the semantics dened by da Costa and his school for paraconsistent C-systems (see [7]
and references therein). Now, let us stress that both Suszkos reduction method and
Bivaluation semantics use non-truth functional two-valued valuations, a fact which
distinguishes them sharply from Dunns semantics with gaps and gluts, which is truth
functional.

We have just explained to what extent our results depend upon Dunns work, but
a not a lesser debt is owed to Ross Brady, who in his nice and clear paper [5] (see also
[6], Ch. 9) showed how to use two-valued interpretations with either gaps or gluts (or
both) for axiomatizing three-valued or four-valued logics. In the cited paper, Brady
axiomatizes in particular the three-valued logic RM3, the four valued logic BN4 and
indicates how 3 in the rst sense (i) recorded above could be axiomatized by extending
BN4 and using under-determined interpretations.
As it has been pointed out above, we shall axiomatize 3 by extending Routley
and Meyers basic positive logic B+ . This suggested to try and dene a Routley-Meyer
ternary relational semantics for this axiomatization as well as for the truth-preserving
and well-determined 3, respectively, in (ii) and (iii) above. These aims were accomplished and are recorded in [23]. However, we cannot include them here unless in pain
of enlarging even more a now large paper. Anyway, let us say that the aforementioned
result (that in [23]) has some interest because it relates 3 (in the three senses), and
similar systems, to relevant logics.
Let us now explain how the paper is organized. In 2, the logic 3(B + ) is dened
and some of its theorems to be used later in the paper are proved (on the label 3(B + ) ,
see 2, below). In 3, we prove a number of facts about prime theories, consistent prime
theories, and complete prime theories. It is especially important what is dubbed The
Fundamental Theorem which allows us to dene the dual two-valued interpretations
referred to above. The facts proved in this section are used in the completeness proofs
in later sections. In 4 under(over)-determined valuations for 3 in the sense of (i)
are dened. Next, it is shown how to extend them to dene under(over)-determined
interpretations. And then, these interpretations are shown to be isomorphic to standard interpretations dened upon the three-valued matrices M3. In 5, it is proved
that 3(B + ) is complete w.r.t. (equivalently) under(over)-interpretations and M3interpretations. In the last section of the paper, 6, consequence relations which are
coextensive to respectively (ii) and (iii) above are dened by using under and overdetermined interpretations. Then, once the appropriate proof-theoretical relations are

dened, strong-completeness theorems (completeness w.r.t. deducibility) are proved.

Finally, it is shown that the well-determined ukasiewicz logics are paraconsistent. We
have included an appendix displaying some matrices used in proving the independence
of the axioms of 3(B + ) added to B+ .

2 The logic 3(B + )

In this section, the logic 3(B + ) is axiomatized and some of its theorems to be used
later in the paper are proved. The label 3(B + ) stands for an axiomatization of 3
built upon B+ where by 3 it is understood the set of three-valued valid formulas as
dened in [15] and B+ is the logic dened in [20].
The propositional language consists of a denumerable set of propositional variables
and the connectives < (conditional), 2 (conjunction), j (disjunction) and (negation).
The biconditional () and the set of ws are dened in the customary way. D, E, F,
etc. are metalinguistic variables for ws.
As it is well-known, Routley and Meyers basic positive logic B+ can be axiomatized
as follows (cf. [20], [25] or [4]).

Axioms:

A1. D < D
A2. (D 2 E) < D / (D 2 E) < E
A3. [(D < E) 2 (D < F)] < [D < (E 2 F)]
A4. D < (D j E) / E < (D j E)
A5. [(D < F) 2 (E < F)] < [(D j E) < F]
A6. [D 2 (E j F)] < [(D 2 E) j (D 2 F)]

Rules:

Modus Ponens (MP): From D < E and D to infer E.

Adjunction (Adj): From D and E to infer D 2 E
Su!xing (Suf): From D < E to infer (E < F) < (D < F)
Prexing (Pref): From E < F to infer (D < E) < (D < F)

Then, the logic 3(B + ) is axiomatized by adding the following axioms to B+ :

A7. D < (E < D)
A8. (D < E) < (E < D)
A9. (D < E) < (E < D)
A10. (D 2 E) < [E j (D < E)]
A11. (D j E) j (D < E)

The notions of proof and theorem are dened in the usual way.

Proposition 1 (On the axiomatization of 3(B + ) )

Given B+ , A7-A11 are independent from each other.

Proof See Appendix 1.

Therefore, 3(B + ) is well-axiomatized w.r.t. B+ . Next, we record some useful theorems of 3(B + ) . We shall provide a proof sketch for each one of them. Regarding these
sketches, we remark the following: (a) by B+ we usually refer to any standard lattice
property of 2 and j provable in B+ ; (b) applications of MP or Adj are not annotated;
(c) by An (Tn) we refer to the inferential use of An (Tn) or, simply, to an instance of

An (Tn). Finally, see Remark 3 below on the labels rVEQ and rEFQ.

Rule VEQ (rVEQ): From D to infer E < D

A7

Rule Transitivity (Trans): From D < E and E < F to infer D < F

Suf

T1. D < D

A1, A8

T2. D < D

A1, A9

T3. (D < E) < (E < D)

Rule Contraposition (Con): From D < E to infer E < D
T4. D < (D < E)

A8, T1, Pref, Trans

T3
A7, A9, Trans

Rule EFQ (rEFQ): From D to infer D < E

T5. D < (D < E)

T4
T4, T1, Suf, Trans

T6. (D j E)  (D 2 E)

A4, Con, Adj, A3; A2, A8, Adj, A5

T7. (D 2 E)  (D j E)

A4, A9, Adj, A3; A2, Con, Adj, A5.

T8. (D < E) < E

T9. (D < E) < D

T10. (D 2 E) < [D j (D < E)]

Proof (Proof sketch) By A10 and B+ ,

1. (D 2 E) < [D j (E < D)]

By 1, T1 and B+ ,
2. (D 2 E) < [D j (E < D)]
By T3,
3. (E < D) < (D < E)

A7, Con
T5, A9

By 2, 3 and B+
4. (D 2 E) < [D j (D < E)]

T11. [(D < E) 2 E] < (D j E)

Proof (Proof sketch) By A10 and Con,

1. [E j (D < E)] < (D 2 E)

By T6, T7, T1, T2 and B+ ,

2. [(D < E) 2 E] < (D j E)

T12. [(D < E) 2 D] < (D j E)

Proof (Proof sketch) Similar to that of T11 using now T10 instead of A10.

T13. [(D < E) 2 (D 2 E)] < F

Proof (Proof sketch) By rVEQ and A11,

1. F < [(D j E) j (D < E)]

By A9
2. [(D j E) j (D < E)] < F
By T6, T1 and B+ ,
3. [(D < E) 2 (D 2 E)] < F
The next theorem is a lemma leaning on which the important theorem T15, whose
corresponding rule (i.e., from D 2 D to infer E j E) is dubbed safety in [10], is
proved.

10

T14. (D 2 D) < [(D j E) < (E j E)]

Proof (Proof sketch) By A4 and Pref,

1. (D < E) < [D < (E j E)]

By T4 and B+ ,
2. (D 2 D) < (D < E)
By 1, 2 and Trans,
3. (D 2 D) < [D < (E j E)]
By A5,

4. {[D < (E j E)] 2 [E < (E j E)]} < [(D j E) < (E j E)]

By A4 and rVEQ,
5. [D < (E j E)] < [E < (E j E)]
By A1,
6. [D < (E j E)] < [D < (E j E)]
By 4, 5, 6 and B+ ,

7. [D < (E j E)] < [(D j E) < (E j E)]

Finally, by 3, 7 and Trans

8. (D 2 D) < [(D j E) < (E j E)]

T15. (D 2 D) < (E j E)

11

Proof (Proof sketch) By A4 and B+ ,

1. (D 2 D) < (D j E)
By T14 and 1,

2. (D 2 D) < {[(D j E) < (E j E)] 2 (D j E)}

By T12,

3. {[(D j E) < (E j E)] 2 (D j E)} < [(D j E) j (E j E)]

By 2, 3 and Trans,

4. (D 2 D) < [(D j E) j (E j E)]

By T6 and B+ ,
5. (D j E) < (E j E)
By 5 and B+ ,
6. [(D j E) j (E j E)] < (E j E)
By 4, 6 and Trans,
7. (D 2 D) < (E j E)

Remark 2 (On some characteristic theorems of 3(B + ) ) Notice that T11 and T12 are
weak versions of the contraposition and modus ponens axioms

conax. [(D < E) 2 E] < D

mpax. [(D < E) 2 D] < E
not derivable in 3(B + ) . And that A10 and T10 are weak versions of (D 2 E) <
(D < E) that is not 3(B + ) -derivable either. The non-derivability of these theses in

12

3(B + ) follows from the completeness of 3(B + ) w.r.t. the 3-valued matrices M3 (cf.
Corollary 3). On the other hand, T13 is a signicative theorem (cf. Remark 11).

Remark 3 (On the rules rVEQ and rEFQ) The label VEQ abbreviates Verum e
quodlibet (A true proposition follows from any proposition) and EFQ, E falso
quodlibet (Any proposition follows from a false proposition). We shall discuss the
rule ECQ (E contradictione quodlibet: Any proposition follows from a contradiction)
rECQ. From D 2 D to infer E
in Section 6 below. We also remark that rVEQ is not infrequently dubbed Rule K
whereas rECQ is named explosion by some paraconsistent logicians.

3 Theories. The Fundamental Theorem

In this section we shall prove some facts about dierent classes of 3(B + ) -theories.
Among these facts the Fundamental Theorem is included. This theorem will make
it possible to dually and equivalently interpret 3 by either under-determined or by
over-determined two-valued interpretations. We begin by recalling some denitions.

Denition 1 (Theories. Consistency)

1. A theory is a set of formulas closed under adjunction and provable 3(B + ) -implication.
That is, d is a theory if whenever D, E M d, then D 2 E M d; and if whenever D < E
is a theorem of 3(B + ) and D M d, then E M d.
2. A theory is prime if whenever D j E M d, then D M d or E M d.
3. A theory is regular i all theorems of 3(B + ) belong to it.
4. A theory is empty i no w belong to it.
5. A theory is w-inconsistent (inconsistent in a weak sense) i for some theorem D of
3(B + ) , D M d. Then, d is w-consistent (consistent in a weak sense) i d is not
w-inconsistent (cf. [22] on the label w-consistency).

13

6. A theory d is a-inconsistent (inconsistent in an absolute sense) i d is trivial, i.e., i

every w belongs to it. Then a theory is a-consistent (consistent in an absolute sense)
i it is not a-inconsistent.
7. A theory d is sc-inconsistent (inconsistent according to the standard concept) i for
some w D, D 2 D M d. Then, d is sc-consistent (consistent according to the standard
concept) i d is not sc-inconsistent.
8. A theory d is complete i for every w D, D M d or D M d.
Next, we record three useful propositions.
Proposition 2 (Regularity and non-emptyness) Let d be a theory. Then, d is
regular i d is non-empty.
Proof Immediate by rVEQ.
Proposition 3 (w-consistency and a-consistency) Let d be a theory. Then, d is
w-consistent i d is a-consistent.
Proof Immediate by rEFQ.
By Proposition 3, w-consistency and a-consistency are coextensive concepts in the
context of the present paper. Concerning the relationship between w-consistency and
sc-consistency, we note the following.
Remark 4 (On w-consistency and sc-consistency) Let d be a non-empty sc-consistent
theory. Then, d is w-consistent as well. For, suppose D M d, D being a theorem. As d
is regular (Proposition 2), D M d and then D 2 D M d, contradicting the sc-consistency
of d. On the other hand, we shall discuss the question if any w-consistent theory is
sc-inconsistent in the last section of the paper.
Proposition 4 (Primeness) Let d be a theory and D a w such that D M
@ d. Then,
there is a prime, w-consistent theory { such that d \ { and D M
@ {.
Proof By using, for example, Zorns lemma, d is extended to a maximal theory { such
that D M
@ {. Then, it is easy to show { is prime (see, e.g., [27]). Finally, by Proposition
3 { is w-consistent.

14

Next, we shall prove three lemmas on the properties of prime theories, prime scconsistent theories and prime, w-consistent, complete theories. These theories shall
play in the completeness proofs (cf. Section 5 and Section 6) the role that, say, Henkin
sets (maximal consistent sets) play in a standard completeness proof of classical logic.
Lemma 1 (Properties of prime theories) Let d be a prime theory. Then, for all
ws D, E,
1. D M d i D M d
2.
(a) D 2 E M d i D M d and E M d
(b) (D 2 E) M d i D M d or E M d
3.
(a) D j E M d i D M d or E M d
(b) (D j E) M d i D M d and E M d
Proof 1, by T1 and T2; 2a, by A2 and the fact that d is closed under adjunction; 2b,
by T7 and the fact that d is prime; 3a, by A4 and the fact that d is prime; 3b, by T6
and the fact that d is closed under adjunction.
In addition to the properties recorded in the preceding lemma, prime sc-consistent
theories have the following properties concerning the conditional.
Lemma 2 (Properties of prime sc-consistent theories) Let d be a prime scconsistent theory. Then, for all ws D, E,

1. D < E M d i D M d or E M d or (D M
@ d and E M
@ d)
2. (D < E) M d i D M d and E M d
Proof
1.

15

(a) Suppose D < E M d. Further, suppose D M d. By T12, D j E M d. Suppose, on

the other hand, E M d. By T11, D j E M d. As d is prime, either D M d or
E M d. Therefore, if either D M d or else E M d, then either D M d or E M d, as
required.
(b) Suppose:
i. D M d. By T5, D < E M d=
ii. E M d. By A7, D < E M d.
iii. D M
@ d and E M
@ d. By A11, (D j E) j (D < E), whence by the primeness of
d, D < E M d.
2.
(a) Suppose (D < E) M d. By T9, D M d; by T8, E M d.
(b) Suppose D M d and E M d. Then, D 2 E M d and, as d is sc-consistent, E M
@ d.
By A10, E j (D < E) M d. Finally, (D < E) M d by the primeness of d.
In addition to the properties recorded in Lemma 1, prime, w-consistent complete
theories have the following properties concerning the conditional.
Lemma 3 (Properties of prime, w-consistent, complete theories) Let d be a
prime, w-consistent, complete theory. Then, for all ws D and E,

1. D < E M d i D M d or E M d
2. (D < E) M d i (D M d and E M
@ d) or (D M
@ d and E M d)
Proof
1.
(a) Suppose D < E M d. If D M
@ d, then D M d because d is complete. Then, by T12,
D j E M d and so, E M d by the primeness of d. If, on the other hand, E M
@ d,
then as d is complete, E M d. Then, by T11, D j E M d and so, D M d by the
primeness of d.

16

(b) Suppose:
i. D M d. By T5, D < E M d.
ii. E M d. By A7, D < E M d.
2.
(a) Suppose (D < E) M d. By T9 and T8, D M d and E M d. Suppose now D M d
and E M d. Then, D 2 E M d. By T13, [(D < E) 2 (D 2 E)] < (F < F). So,
(F < F) M d contradicting the w-consistency of d. Therefore, either D M
@ d or
@ d) or (D M
@ d and E M d) as was to
EM
@ d. Consequently, either (D M d and E M
be proved.
(b) Suppose:
i. D M d and E M
@ d. As d is complete, E M d. Thus, D 2 E M d. By A10,
E j (D < E) M d. So, (D < E) M d by the primeness of d.
ii. D M
@ d and E M d. Proof similar to that of (i) by using now T10.
In the sequel we shall use the Routley operator W (cf. [27] and references therein)
in order to dene the W-images of prime theories, as in relevant logics.
Denition 2 (W-images of prime theories) Let d be a prime theory. The set dW is
@ d}.
dened as follows: dW = {D | D M
Then, we prove a couple of lemmas on the relationship between prime theories and
their W-images.
Lemma 4 (Primeness of W-images) If d is a prime theory, then dW is a prime theory
as well. Moreover, for any w D> D M dW i D M
@ d.
Proof (Cf. e.g. [27]).
1. dW is closed under 3(B + ) -implication: by Con.
2. dW is closed under adjunction: by (D 2 E) < (D j E) (T7, from left to right) and
the fact that d is a prime theory.
3. dW is prime: by (D 2 E) < (D j E) (T6 from right to left) and the fact that d is a
theory.

17

4. D M dW i D M
@ d: by T1 and T2.
Lemma 5 (Regularity, consistency and W-images) Let d be a prime theory. Then,
1. d is regular i dW is w-consistent.
2. d is w-consistent i dW is regular.
3. d is sc-consistent i dW is complete.
4. d is complete i dW is sc-consistent.
Proof By Denition 1, Denition 2 and Lemma 4. Let us prove, as a way of an example,
a couple of cases:

1.
(a) Suppose d is regular. If dW is not w-consistent, then for some theorem D, D M dW.
By Lemma 4, D M
@ d, contradicting the regularity of d. The converse of 1a is proved
similarly.
3.
(a) Suppose d is sc-consistent. If dW is not complete, then D M
@ dW and D M
@ dW for
some w D. By Denition 2 and Lemma 4, D M d and D M d, i.e., D 2 D M d,
contradicting the sc-consistency of d. The converse of 3a is proved similarly.
Finally, we prove a proposition preceding The Fundamental Theorem.

Proposition 5 (sc-consistency or completeness) Let d be a prime theory. If d is

sc-inconsistent, then d is complete.
Proof Immediate by Denition 1 and T15.
Theorem 1 (The fundamental theorem) Let d be a non-empty theory such that
for some w D, D M
@ d. Then, there is a regular, prime, sc-consistent theory { such
that D M
@ {; and a regular, prime, w-consistent, complete theory | such that D M |.

18

Proof Assume the hypothesis of Theorem 1. By Proposition 2 and Proposition 4, there

is a regular, prime, w-consistent theory } such that d \ } and D M
@ }. By Lemma 4
and Lemma 5, }W is a regular, prime and w-consistent theory such that D M }W. If }
is sc-consistent, then } and }W are the required { and | in the statement of Theorem
1, respectively. So, suppose } is sc-inconsistent. Then, } is complete by Proposition
5. And D M }, because D M
@ }. On the other hand, }W is sc-consistent by Lemma 5.
Finally, D M
@ }W because D M }. Consequently, if } is sc-inconsistent, }W and } are,
respectively, the required { and | in the statement of Theorem 1.
In the next section under-determined and over-determined two-valued interpretations for 3(B + ) are dened.

4 Under-determined and over-determined two-valued interpretations for

3(B + )
We begin by recalling ukasiewiczs three-valued matrices (cf. [15], [33]):
Denition 3 (ukasiewiczs matrices) ukasiewiczs three-valued matrices M3
are the following (1 is the only designated value and 0 $

1
2

$ 1):

<

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

1
2

Then, we set:
Denition 4 (3-valuations, 3-interpretations, 3-validity) Let S3 be the
set {0> 12 > 1}. An 3-valuation, y, is a function from P (the set of all propositional variables) to S3 . Now, let y be an 3-valuation. An 3-interpretation, L> is the extension
of y to all ws according to the truth tables in Denition 3. Finally, a w D is 3-valid
(in symbols, 3 D) i L(D) = 1 for all 3-valuations y.
Next, under-determined valuations are dened.

19

Denition 5 (U-valuations) Let V be the set {W> I }. A u-valuation (under-determined

valuation), y, is a function assigning a proper subset of V to each propositional variable. That is, y is a u-valuation i for each propositional variable s, y(s) = {W } or
y(s) = {I } or y(s) = .
Then, u-valuations are extended to u-interpretations as shown in Denition 6.
Denition 6 (U-interpretations) Let y be a u-valuation. A u-interpretation, L, is
a function assigning a proper subset of V to all propositional variables and all ws
according to the following conditions: for each propositional variable s and ws D, E,

1. L(s) = y(s)
2a. W M L(D) i I M L(D)
2b. I M L(D) i W M L(D)
3a. W M L(D 2 E) i W M L(D) and W M L(E)
3b. I M L(D 2 E) i I M L(D) or I M L(E)
4a. W M L(D j E) i W M L(D) or W M L(E)
4b. I M L(D j E) i I M L(D) and I M L(E)
5a. W M L(D < E) i I M L(D) or W M L(E) or (W M
@ L(D) and I M
@ L(E))
5b. I M L(D < E) i W M L(D) and I M L(E)
Before dening validity, we point out a remark that may be useful in working with
the present semantics.
Remark 5 (On the determination of the value of ws according to u-interpretations)
Let L be a u-interpretation. Notice that if W M L(D) (I M L(D)), then I M
@ L(D)
(W M
@ L(D)), but that the converse does not follow generally: D can be assigned neither
{W } nor {I }.
Next, u-validity, and then, over-determined interpretations are dened.
Denition 7 (U-validity) A formula D is u-valid (in symbols, u D) i W M L(D)
for all u-valuations y.

20

Denition 8 (O-valuations) Let V be the set {W> I }. An o-valuation (over-determined

valuation), y, is a function assigning a non-empty subset of V to each propositional variable. That is, y is an o-valuation i for each propositional variable s, y(s) = {W } or
y(s) = {I } or y(s) = {W> I }.
Then, o-valuations are extended to o-interpretations as follows.
Denition 9 (O-interpretations) Let y be an o-valuation. An o-interpretation, L,
is a function assigning a non-empty subset of V to all propositional variables and all
ws according to the following conditions: 1, 2a, 2b, 3a, 3b, 4a and 4b are dened
exactly as in Denition 8. Then, 5a and 5b are as follows, for any ws D, E:

5a. W M L(D < E) i I M L(D) or W M L(E)

@ L(E)) or (I M
@ L(D) and I M L(E))
5b. I M L(D < E) i (W M L(D) and W M
Remark 6 (On the determination of the value of ws according to o-interpretations)
Notice that if W M
@ L(D), (I M
@ L(D)), then I M L(D) (W M L(D)), but that W M L(D)
(I M L(D)) does not imply that I M
@ L(D) (W M
@ L(D)) (cf. Remark 5).
Validity is (dierently from u-validity) dened as follows (cf. Denition 7):
Denition 10 (O-validity) A formula D is o-valid (in symbols, o D) i for all
o-valuations I M
@ L(D).
Remark 7 (Duality of u-interpretations and o-interpretations) As pointed out in the
introduction, u-interpretations and o-interpretations are dual in the sense that  is dual
to the set V = {W> I }.
Now, we shall put in correspondence u-interpretations with 3-interpretations.
Denition 11 (Corresponding u-interpretations (3-interpretations) to 3interpretations (u-interpretations)) Let L3 be an 3-interpretation extending
the 3-valuation y3 . Then, dene a u-valuation yu as follows: for each propositional

21

variable s, set:

1. yu (s) = {W } i y3 (s) = 1
2. yu (s) =  i y3 (s) =

1
2

3. yu (s) = {I } i y3 (s) = 0
Next we extend yu to a u-interpretation Lu according to clauses 1-5 in Denition
6. It is said that yu and Lu are the corresponding u-valuation and u-interpretation to
y3 and L3 , respectively.
On the other hand, suppose given a u-interpretation Lu extending a u-valuation
yu . The 3-interpretation L3 and 3-valuation y3 (upon which L3 is dened) corresponding to Lu and yu , respectively, are dened in a similar way. Therefore, given
a u-interpretation (3-interpretation) it is always possible to dene the corresponding
3-interpretation (u-interpretation).
Lemma 6 (Isomorphism of u-interpretations and 3-interpretations) Let L3
(Lu ) be an 3-interpretation (u-interpretation) and Lu (L3 ) its corresponding
u-interpretation (3-interpretation) as dened in Denition 11. Then, for each w D,
it is proved

1. Lu (D) = {W } i L 3 (D) = 1
2. Lu (D) =  i L3 (D) =

1
2

3. Lu (D) = {I } i L3 (D) = 0
Proof By an easy induction on the length of D (the proof is left to the reader).
Theorem 2 (Coextensiveness of u-validity and 3-validity) For each w D,
u D i 3 D.
Proof Immediate by Denition 4, Denition 7 and Lemma 6.
Now, as it has been the case with u-validity, we will show that o-validity and 3validity are coextensive concepts.

22

Denition 12 (Corresponding o-interpretations (3-interpretations) to 3interpretations (o-interpretations)) Let L3 be an 3-interpretation extending

the 3-valuation y3 . Then, dene a o-valuation yo as follows: for each propositional
variable set:

1. yo (s) = {W } i y3 (s) = 1
2. yo (s) = {W> I } i y3 (s) =

1
2

3. yo (s) = {I } i y3 (s) = 0
Next we extend yo to an o-interpretation Lo according to clauses 1-5 in Denition
9. It is said that yo and Lo are the corresponding o-valuation and o-interpretation to
y3 and L3 , respectively.
On the other hand, given an o-interpretation Lo extending an o-valuation yo , the
corresponding 3-interpretation L3 and 3-valuation y3 are dened in a similar way.
Therefore, given an o-interpretation (L3-interpretation) it is always possible to dene
the corresponding 3-interpretation (o-interpretation).
Now, similarly as in the case of u-interpretations, it is proved:
Lemma 7 (Isomorphism of o-interpretations and 3-interpretations) Let L3
(Lo ) be an 3-interpretation (o-interpretation) and Lo (L3 ) its corresponding o-interpretation
(3-interpretation) as dened in Denition 12. Then, for each w D, it is proved
1
2
1
2. I M Lo (D) i L 3 (D) = 0 or L 3 (D) =
2
1. W M Lo (D) i L3 (D) = 1 or L3 (D) =

Proof By an easy induction on the length of D (the proof is left to the reader).
Theorem 3 (Coextensiveness of o-validity and 3-validity) For each w D,
o D i 3 D.
Proof Immediate by Denition 4, Denition 10 and Lemma 7.

23

Therefore, it follows from Theorem 2 and Theorem 3 that u-validity and o-validity
constitute equivalent alternative dual (in the sense of Remark 7) interpretations of
3-validity. We record the fact in the following corollary.
Corollary 1 (Coextensiveness of u-validity and o-validity) For any w D, u D
i o D.
Proof Immediate by Theorem 2 and Theorem 3.
In the next section it is proved that 3(B + ) is sound and complete w.r.t. 3-validity.
Now, given Corollary 1, it can be accomplished either via u-validity or via o-validity.
We shall follow the rst path and will sketch how to proceed along the second one.

5 Completeness of 3(B + )
We shall prove that 3(B + ) axiomatizes the set of matrices M3. Actually, we shall
prove that a w D is a theorem of 3(B + ) i D is u-valid, whence by Theorem 2 it
follows that D is 3-valid i D is a theorem of 3(B + ) . In other words, it will be proved
that 3(B + ) is an alternative axiomatization (to that of Wajsberg [35], for example)
of ukasiewicz three-valued logic 3 understood as the set of all the three-valued valid
formulas according to M3 (cf. Denition 3). On the other hand, and at the same time,
it is proved that ukasiewicz logic 3 is sound and complete w.r.t. u-validity whence
by Corollary 1 3(B + ) is sound and complete also w.r.t o-validity.
Theorem 4 (Soundness of 3(B + ) w.r.t. 3-validity) For any w D, if `3(B

+)

D, then 3 D.
Proof It is easy to check that all axioms of 3(B + ) are 3-valid and that the rules of
derivation of 3(B + ) preserve 3-validity (the easiest way it to use any matrix tester
such as, e.g., MaTest (cf. [12])).
Then, we have:
Corollary 2 (Soundness of 3(B + ) w.r.t. u-validity) For any w D, if ` 3(B
then u D.

+)

D,

24

Proof Immediate by Theorem 2 and Theorem 4.

Next, we proceed into proving completeness. Firstly, we dene interpretations by
using prime theories.
Denition 13 (yT -valuations and LT -interpretations) Let S be, as above, the
set {W> I } and T

be a prime, sc-consistent theory. Now, we dene the valuation yT

as follows: for each propositional variable s, set:

1. W M yT (s) i s M T
2. I M yT (s) i s M T
Then, we extend this valuation yT to an interpretation LT for all ws according
to clauses 1-5 in Denition 6.
We note the following:
Remark 8 (LT is a u-interpretation) Notice that yT is a u-valuation: given that T is
sc-consistent, each propositional variable s is assigned either {W }, {I } or the empty
set, but not {W> I }. Then, the fact that LT is a u-interpretation follows by Denition
13.
Now, we have:
Theorem 5 (Interpretation by using sc-consistent theories) Let yT

and LT

be as in Denition 13. Then, for each w D, W M LT (D) i D M T and I M LT (D)

i D M T .
Proof Induction on the length of D. (The clauses cited in points 2-5 below refer to
clauses in Denition 6. H.I. abbreviates hypothesis of induction).
1. D is a propositional variable. By clauses 1 and 2 in Denition 13.
2. D is of the form E.
(a) W M LT (E) i (clause 2a) I M LT (E) i (H.I.) E M T .

25

(b) I M LT (E) i (clause 2b) W M LT (E) i (H.I.) E M T i (Lemma 1(1)) E M T .

3. D is of the form E 2 F
(a) W M LT (E 2 F) i (clause 3a) W M LT (E) and W M LT (F) i (H.I.) E M T and
F M T i (Lemma 1(2a)) E 2 F M T .
(b) I M LT (E 2 F) i (clause 3b) I M LT (E) or I M LT (F) i (H.I.) E M T or
F M T i (Lemma 1(2b)) (E 2 F) M T .
4. D is of the form E j F= Similar as 3 by using clause 4a, clause 4b and Lemma 1(3a and
3b).
5. D is of the form D < F
@ LT (E) and
(a) W M LT (E < F) i (clause 5a) I M LT (E) or W M LT (F) or (W M
@ T and F M
@ T ) i (Lemma 2(1))
I M
@ LT (F)) i (H.I.) E M T or F M T (or E M
E <F MT.
(b) I M LT (E < F) i (clause 5b) W M LT (E) and I M LT (F) i (H.I.) E M T and
F M T i (Lemma 2(2)) (E < F) M T .
We can now prove the completeness theorem.
Theorem 6 (Completeness of 3(B + ) w.r.t. u-validity) For any w D, if u D,
then `3(B

+)

D.

Proof We will prove the contrapositive of the claim. Suppose that D is not a theorem
of 3(B + ) , and consider 3(B + ) as the set of its theorems. As 3(B + ) is a regular theory
without D, by Theorem 1, there is a prime, regular, sc-consistent theory T such that
DM
@ T . By Denition 13 and Theorem 5, W induces a u-interpretation LT of all ws
@ LT (D), i.e., 2u D, as was to be
such that for any w E, LT (E) i E M T . So, W M
proved.
Corollary 3 (Soundness and completeness of 3(B + ) w.r.t. 3-validity) For
any w D, 3 D i `3 (B

+)

D.

Proof Immediate by Theorem 4, Theorem 6 and Theorem 2.

26

Finally, we note the following corollaries.

Corollary 4 (Soundness and completeness of 3(B + ) w.r.t. o-validity) For
any w D, o D i `3(B

+)

D.

Proof Immediate by Theorem 3 and Corollary 3.

We have proved the completeness of 3(B + ) by using u-interpretations. But, as it
was remarked above, this can also be done dually by relying on o-interpretations.
So, let us sketch a direct proof of Corollary 4 from left to right.
Corollary 5 (Completeness of 3(B + ) w.r.t. o-validity) For any w D, if o D
then `3 (B

+)

D.

Proof (Proof sketch)

1. We dene yT -valuations and LT -interpretations similarly as in Denition 13 but now
referred to a prime, w-consistent, complete theory T . Notice that yT is now an ovaluation: given that T is prime and complete each propositional variable is assigned
either {W } or {I } (or {W> I }: although T is w-consistent, it is not necessarily scconsistent. We shall refer to this question in the last section of the paper). Therefore,
LT is an o-interpretation, by Denition 9.
2. We prove that LT (as dened in point 1 above) is such that for each w D, W M LT (D)
i D M T and I M LT (D) i D M T . The proof is similar to that of Theorem 5 using
now Lemma 3 in the conditional case.
3. We prove completeness by contraposition similarly as in Theorem 6. We now use the
second part of Theorem 1: if D is not a theorem of 3(B + ) , then there is a prime,
w-consistent, complete theory such that D M T , whence, I M LT (D), i.e., 2o D as was
to be proved.
In conclusion, ukasiewicz 3, understood as the set of all three-valued valid formulas according to M3 (cf. Denition 3) can be interpreted as a two-valued logic,
the third truth-value being in its turn interpreted equivalently as under-determined or
over-determined w.r.t. truth and falsity classically understood.

27

6 On strong completeness and paraconsistency

Let us begin by recalling the following denition:
Denition 14 (3-interpretation of sets of ws) Let  be a set of ws and L3
an 3-interpretation. Then,

L3 ( ) = inf{L3 (D) : D M  }
(cf. Denition 3).
Now, as was noted in the Introduction, there are essentially two ways of dening
a consequence relation in many-valued logics: truth-preserving and well-determined
consequence relations which, in the case of 3, are given in Denition 15 and Denition
16 (the labels are after [37], 13).
Denition 15 (Degree of truth preserving consequence relation) Let  be a
set of ws and D a w. Then,  
3 D i L3 ( ) $ L3 (D) for each 3-valuation y
(cf. Denition 4).
Denition 16 (Truth-preserving consequence relation) Let  be a set of ws
and D a w. Then,  13 D i if L3 ( ) = 1, then L3 (D) = 1 for each 3-valuation y
(cf. Denition 4).
1
We shall refer by the symbols 
3 and 3 to the relations dened in Denition 15

and Denition 16, respectively.

These two ways of understanding the notion of semantical consequence in manyvalued logics are not in general equivalent and it is the case with 3, as it is shown
below.
1
Proposition 6 (Relationship between 
3 and 3 ) For any set of ws  and
1
any w D, if  
3 D, then  3 D. But the converse does not hold.
1
Proof It is immediate that the relation 
3 entails the relation 3 . Regarding the

converse, {D < E> D} 13 E holds for any ws D, E. However, consider propositional

28

variables sl , sm and a 3-interpretation L3 such that L3 (sl ) =

1
2

and L3 (sm ) = 0.

Then, {sl < sm > sl } 2

3 sm for this 3-interpretation L3 .

Remark 9 (Incidental remark on Gdel logics) Gdel logics are a remarkable case in
which the two alternative ways of dening semantical consequence are equivalent (cf.
[2] or [3]).

We recall that, according to [37], 13, we can label truth preserving three-valued
ukasiewicz logic the logic determined by 13 and well-determined three-valued
ukasiewicz logic that determined by 
3 . The aim of this section is: (1) to dene
under-determined and over-determined consequence relations corresponding to 
3
and 13 ; (2) to investigate to which proof-theoretical consequence relations 
3 and
13 correspond, and, thus, to establish (two dierent) completeness theorems w.r.t.
deducibility; (3) to investigate in which sense, if any, the 3-valued logic 3 can be
considered a paraconsistent logic.
We begin by establishing isomorphisms of u-interpretations and o-interpretations
with 3-interpretations.
But, we rst set the following denition:

Denition 17 (U-interpretations and o-interpretations of sets of ws) Let L

be an arbitrary u-interpretation or o-interpretation. Then, for any set  of ws,

1. W M L( ) i D M  (W M L(D)
2. I M L( ) i !D M  (I M L(D))
Then, it is proved:

Proposition 7 (Isomorphism of 3-interpretations and u-interpretations of

sets of ws) Let  be a set of ws, L 3 (Lu ), an 3-interpretation (u-interpretation)

29

and Lu (L3 ) its corresponding u-interpretation (3-interpretation). Then, we have:

1. Lu ( ) = {W } i L 3 ( ) = 1
2. Lu ( ) =  i L3 ( ) =

1
2

3. Lu ( ) = {I } i L3 ( ) = 0
Proof Immediate by Lemma 6 and Denition 14 and Denition 17.
Proposition 8 (Isomorphism of 3-interpretations and o-interpretations of
sets of ws) Let  be a set of ws, L 3 (Lo ), an 3-interpretation (o-interpretation),
and Lo (L 3 ) its corresponding o-interpretation (3-interpretation). Then, we have:
1
2
1
2. I M Lo ( ) i L3 ( ) = 0 or L 3 ( ) =
2

1. W M Lo ( ) i L3 ( ) = 1 or L3 ( ) =

Proof Immediate by Lemma 7, Denition 14 and Denition 17.

We can now dene under-determined and over-determined relations corresponding
to 13 .
Denition 18 (Under-determined 1u -relations) For any set of ws  and any
w D,  1u D i if W M L( ), then W M L(D) for all u-valuations y.
Denition 19 (Over-determined 1o -relations) For any set of ws  and any w
@ L( ), then I M
@ L(D) for all o-valuations y.
D,  1o D i if I M
Then, we immediately have:
Proposition 9 (Coextensiveness of 13 , 1u and 1o ) For any set of ws  and
any w D,  13 D i  1u D i  1o D.
Proof (a)  13 D i  1u D: immediate by Denition16, Denition 18 and Proposition 7; (b)  13 D i  1o D: immediate by Denition 16, Denition 19 and
Proposition 8.

30

Next, we shall establish which is the proof-theoretical consequence relation corresponding to 13 . Consider the following denition:
Denition 20 (Proof-theoretical consequence relation. First sense:
a-derivability) Let  be a set of ws and D be a w. Then,  `d3 D (D is a-derivable
from  in 3) i there is a nite sequence of ws E1 > ===> Ep such that Ep is D and
for each l (1 $ l $ p) one of the following is the case: (1) El M  ; (2) El is an axiom
of 3(B + ) ; (3) El is the result of applying any of the rules Adj, MP, Suf or Pref to one
or two of the preceding ws in the sequence.
We shall refer by `d3 to the relation dened in Denition 20.
Now, in order to prove completeness, we set the following:
Denition 21 (The set of consequences of a set of ws. First sense: ad

derivability) Let  be a set of ws. The set F q (the set of all ws a-derivable
from  ) is dened as follows:
d

F q = {D :  `d3 D}
Then, it is proved:
d

Proposition 10 (F q is a regular theory) For any set of ws  , F q is a regular

theory.
d

Proof It is trivial that F q is closed under adjunction and modus ponens. Then, by
Corollary 3, all theorems of 3(B + ) are derivable from A1-A11 and the rules Adj, MP,
d

Suf and Pref. Thus, for any theorem D of 3(B + ) ,  `d3 D. So, D M F q , i.e., F q is
d

regular. Finally, F q is closed under 3(B + ) -implication, given that it is regular and
closed under modus ponens.
Proposition 11 (The rule rECQ) For any ws D, E, D 2 D `d 3 E.
Proof By the theorem (D 2 D) < [(D 2 D) < E] and the fact that by Denition 21
d

and Proposition 10, F q (D 2 D) is closed under MP and under 3(B + ) -implication

(cf. Denition 1).

31

Now, we have:

Theorem 7 (Strong soundness and completeness w.r.t. `d3 ) For any set of
ws  and any w D,  `d3 D i  13 D i  1u D i  1o D.
Proof (a) We have to show: if  `d3 D, then  13 D. The proof is by induction on the
length of the proof of D from  . Now, by using the set of matrices M3 (Denition 3) it
is immediately shown that A1-A11 are 3-valid and that the rules preserve 3-validity
(notice that the theses su!xing and prexing, i.e., (D < E) < [(E < F) < (D < F)]
and (E < F) < [(D < E) < (D < F)] are 3-valid. The tester developed in [12]
can be useful). (b) Now, we prove the converse, that is: if  13 D, then  `d3 D.
@
Suppose that there is a set of ws  and a w D such that  0d3 D. Then, D M
d

F q . So, by Proposition 4, there is a (regular) prime w-consistent theory  such that

d

@ . Now, suppose  is sc-inconsistent. By Proposition 5,  is then

F q \  and D M
complete. Consequently, D M , and, on the other hand, there is an o-interpretation
L such that W M L () and I M L (D) (cf. Corollary 5). Now,  is sc-consistent:
d

otherwise, D M F q because F q is closed by rECQ (Denition 21 and Proposition

11). Moreover, I M
@ L ( ). For suppose I M L (E) for some E M  (cf. Denition
17). Now, I M L (E) i E M . So, E M  , which contradicts the sc-consistency
of  . Therefore, I M
@ L ( ) and, as shown above, I M L (D). Consequently,  21o D
(cf. Denition 19). So, if  is sc-inconsistent, then  21o D. But suppose on the other
hand that  is sc-consistent. By Theorem 5, there is a u-interpretation L such that
d

@ L (D). As  \ F q \ , W M L ( ). Therefore,  21u D (cf.

W M L () and W M
Denition 18). In consequence, if  013 D, either  21o D or else  21u D. So, by
Proposition 9,  213 D follows. Consequently, by (a) and (b) above, we have  `13 D
i  13 D, whence by using again Proposition 9, Theorem 7 immediately follows.

Next, we shall investigate to which proof-theoretical relation 

3 corresponds.
Firstly, we dene under-determined and over-determined relations corresponding to

3 .

32

Denition 22 (Under-determined 
u -relations) For any set of ws  and any
@ L( ) and I M
@ L(D)) for all u-valuations
w D,  
u D i I M L( ) or W M L(D)or (W M
y.
Denition 23 (Over-determined 
o -relations) For any set of ws  and any w
@ L( ) or W M L(D)) and (I M L( ) or I M
@ L(D)) for all o-valuations
D,  
o D i (W M
y.


Then, we prove two propositions on the coextensiveness of 
3 with u and o ,

respectively.

Proposition 12 (Coextensiveness of 
3 and u ) For any set of ws  and any

w D,  
3 D i  u D.


Proof (a) We have to prove if  
3 D, then  u D. Suppose  3 D for a set

of ws  and a w D and let Lu be an arbitrary u-interpretation. We have to show:

@ Lu (D) and I M
@ Lu (D)). Now, let L3 be the 3I M Lu ( ) or W M Lu (D) or (W M
interpretation corresponding to Lu (cf. Denition 11). If I M Lu ( ), then case (a) is
proved. So, suppose I M
@ Lu ( ). By Proposition 7 either L3 ( ) = 1 or L3 ( ) =

1
2.

Suppose L3 ( ) = 1. By hypothesis, L3 (D) = 1 and, so, W M Lu (D), by Lemma 6.

Suppose, on the other hand, L3 ( ) = 12 . By hypothesis, L3 (D) = 1 or L3 (D) = 12 . If
@ Lu ( ) and I M
@ Lu (D)
L3 (D) = 1, then W M Lu (D) (Lemma 6). If L3 (D) = 12 , then W M
by Proposition 7 and Lemma 6. (b) Now, we prove the converse, that is, if  
u D,

then  
3 D. Suppose  u D for a set of ws  and a w D and let L3 be an

arbitrary L3 -interpretation. We have to show L3 ( ) $ L3 (D). By Denition 22, for

@ Lu ( ) and
any interpretation Lu , we have either I M Lu ( ) or W M Lu (D) or (W M
I M
@ Lu (D)). We prove that  
3 D follows from each one of these alternatives (we
shall employ Proposition 7 and Lemma 6 throughout the proof without mentioning
them). Let Lu be the corresponding u-interpretation to L3 .
1. I M Lu ( ). Then, L3 ( ) = 0. So, L3 ( ) $ L3 (D) for any w D.
2. W M Lu (D). Then, L3 (D) = 1. So, L3 ( ) $ L3 (D) for any set of ws  .

33

3. W M
@ Lu ( ) and I M
@ Lu (D)=Then, I M Lu ( ) or I M
@ Lu ( ) and W M Lu (D) or W M
@ Lu (D).
Suppose I M Lu ( ). Then, L3 ( ) = 0 and so, for any w D, L3 ( ) $ L3 (D).
Suppose, on the other hand, I M
@ Lu ( ). Then, Lu ( ) = , and so, L3 ( ) =

1
2.

Now,

@ Lu (D). If W M Lu (D), then L3 (D) = 1, whence L3 ( ) $ L3 (D). If

W M Lu (D) or W M
W M
@ Lu (D), then Lu (D) = , and so, L3 (D) = 12 , whence L3 ( ) $ L3 (D).

Proposition 13 (Coextensiveness of 
3 and o ) For any set of ws  and any

w D,  
3 D i  o D.

Proof Similar to that of Proposition 12 (it is left to the reader).

Now, after dening the rule 3-implication, another proof-theoretical relation is
presented.
Denition 24 (Rule 3-implication) For any ws D, E, the rule 3-implication
is:
r. imp. From D and `3 D < E to infer E.
Denition 25 (Proof-theoretical consequence relation. Second sense:
b-derivability) Let  be a set of ws and D a w. Then  `e3 D (D is b-derivable
from  in 3) i there is a nite sequence of ws E1 > ===> Ep such that Ep is D and
for each l (1 $ l $ p) one of the following is the case: (1) El M  ; (2) El is a theorem
of 3; (3) El is the result of applying Adj to two previous formulas in the sequence;
(4) El is by 3-implication from a previous formula in the sequence.
Denition 26 (The set of consequences of a set of ws. Second sense: be

derivability) Let  be a set of ws. The set F q (the set of all ws b-derivable
from  ) is dened as follows:
e

F q = {D :  `e3 D}
Then, we immediately have:
e

Proposition 14 (F q is a regular theory) Let  be a set of ws. Then, F q is

a regular theory (cf. Denition 1).

34

Proof It is immediate.
Next, it is proved
Theorem 8 (Strong soundness and completeness w.r.t. `e3 ) For any set of


ws  and any w D,  `e3 D i  
3 D i  u D i  o D.

Proof (a) We have to show: if  `e3 D, then  

3 D. The proof is by induction on
the length of the proof of D from  . If D M  or D is by Adj., the proof is trivial;
and it is immediate if D is a theorem of 3 (by Theorem 4). So, suppose that D
is by r. imp. Then,  `e3 E, E < D being a theorem of 3. Let now L3 be an
arbitrary 3-interpretation. By hypothesis of induction, L3 ( ) $ L3 (E); by Theorem
4, L3 (E < D) = 1. Now, there are three possibilities to consider:
1. L3 (E) = 1. Then, L3 (D) = 1 according to the set of matrices M3 (cf. Denition 3).
So, L3 ( ) $ L3 (D).
2. L3 (E) = 0. Then, clearly, L3 ( ) $ L3 (D).
3. L3 (E) =

1
2.

Then, L3 (D) = 1 or L3 (D) =

1
2

according to the set of matrices M3

(cf. Denition 3). So, L3 ( ) $ L3 (D)

e
(b) Now we prove the converse, that is: if  
3 D, then  `3 D. Suppose there

is a set of ws  and a w D such that  0e3 D. Similarly as in the proof of Theorem
e

@ .
7, we have a (regular) prime w-consistent theory  such that F q \  and D M
Suppose now that  is sc-inconsistent. Then,  is complete (Proposition 5). So, there
@ L (D) (cf. the proof of Corollary
is an o-interpretation L such that W M L (), W M
e

5). As  \ F q \ , W M L ( ). So,  2
o D (cf. Denition 23). Suppose, on the
other hand, that  is sc-consistent. By Theorem 5, there is a u-interpretation L such
@ L (D). As  is sc-consistent, there is no E M  such that
that W M L () and W M
e

@ L (). Now, as  \ F q \ , W M L ( )
L (E) = I (otherwise E M ). So, I M
@ L (D),  2
and, as shown above, I M
@ L ( ), whence, together with W M
u D follows

(cf. Denition 22). Therefore, if  2e3 D, then either  2
o D or  2u D. So, by


Proposition 9,  2
3 D. Consequently, by (a) and (b) we have  `3 D i  3 D,

whence by using again Proposition 9, Theorem 8 immediately follows.

35

Remark 10 (On rECQ and strong completeness) Notice the essential role that rule
rECQ plays in the proof of Theorem 7 in order to establish that  is sc-consistent: a
similar conclusion cannot be reached in the case of Theorem 8.
Let us sum up. We have considered ukasiewicz three-valued logic 3 from three
dierent points of view:
1. 3: as the set of the formulas valid according to the set of matrices M3 (cf.
Denition 3).
2. 3a: as the logic determined by the truth-preserving relation 13 or, equivalently,
that determined by the proof-theoretical relation `d3 .
3. 3b: as the logic determined by the degree of truth-preserving relation 
3 or,
equivalently, that determined by the proof-theoretical relation `e3 .
3a is labelled truth preserving ukasiewicz logic 3 and 3b well-determined
ukasiewicz logic 3 by Wojcicki in [37], p. 42. Each of 3, 3a and 3b have been
provided with equivalent and dual two-valued semantics with either under-determined
or over-determined valuations. Moreover, as it was pointed out in the introduction, a
Routley and Meyer ternary relational semantics for 3, 3a and 3b can be provided.
(cf. [23]). But we do not have room here to discuss the matter. We shall instead end
the paper with some notes on 3a and 3b and paraconsistency.
As it is well known, the notion of a paraconsistent logic can be rendered as follows
(cf. [7] or [21]).
Denition 27 (Paraconsistent logics) Let represent a consequence relation (may
it be dened either semantically or proof-theoretically). And let S be the logic determined by it. Then, S is paraconsistent if the rule ECQ
r.ECQ D 2 D E (for any D, E)
is not a rule of S.
Then, we have:

36

Proposition 15 (3a and 3b and paraconsistency) The logic 3b is paraconsistent but 3a is not.
Proof (a) 3a is not paraconsistent: immediate by Proposition 11. (b) 3b is paraconsistent. Consider the lwk and pwk propositional variables sl and sp and an 3interpretation L3 such that L3 (sl ) =
L3 (sp ). So,

sl 2 sl 2
3

1
2

and L3 (sp ) = 0. Then, L3 (sl 2 sl ) A

sp and, consequently, r.ECQ is not 

3 -valid.

Now, given that it is decidable if any w is a theorem of 3 (use, e.g., MaTest, cf.
[12]), we have a strong paraconsistent logic, 3b, useful in cases when some propositions
are under-determined or over-determined w.r.t. its truth-value.
On the other hand, sometimes a logic is dened to be paraconsistent if at least one
of its sc-inconsistent theories is not trivial. Now, in this sense we note the following:
Proposition 16 (Sc-inconsistent theories that are a-consistent) There are regular, prime, w-consistent, complete theories that are sc-inconsistent (cf. Denition 1).
Proof Let sl and sp be the lwk and pwk propositional variables, and consider the
set | = {E : `3 D & `3 [D 2 (sl 2 sl )] < E}. It is easy to prove that | is a
theory. Moreover, it is regular (by A2), but | is sc-inconsistent: sl 2 sl M |. Anyway,
| is a-consistent: [(sl < sl ) 2 (sl 2 sl )] < sp is falsied by M3 (cf. Denition
3) according to any 3-interpretation L3 such that L3 (sl ) =

1
2

and L3 (sp ) = 0.

@ |. Now, by
So, 03 [(sl < sl ) 2 (sl 2 sl )] < sp by Theorem 4, and then, sp M
Proposition 4, there is a regular, prime, w-consistent theory { such that | \ { and
@ {. As sl 2sl M |, { is sc-inconsistent. Finally, { is complete as it is sc-inconsistent
sp M
(Proposition 5).
In a similar way, a regular, prime, w-consistent, complete theory can be built upon
e

F q(sl 2 sl ).
The remarks referred to above are the following:
Remark 11 (On the collapse of theories into triviality) By T13, if a theory contains a
negated conditional together with its consequent and the negation of the antecedent,

37

this theory is trivial. Also, if a theory contains the negation of a theorem, it is trivial
(by T5).
Remark 12 (On a general denition of 3b) In [36], p. 204, 3a is dened as follows:

3a = (L, FqW dxw3 , MP)

where L is the algebra of the propositional formulas and W dxw3 is the set of formulas
validated by M3 (cf. Denition 3). In a parallel way, 3b could be dened as follows:

3b = (L, FqW dxw3 , r. imp., Adj)

where r. imp- and Adj are the rules 3-implication and adjunction (cf. Denition 24
and recall that MP does not preserve 
3 -validity: cf. Proposition 6).
Remark 13 (Paraconsistency of all well-determined ukasiewicz logics) It is a corollary
of Proposition 15 that each one of the well-determined ukasiewicz logics 3,..., $
(cf. [37], 13) is paraconsistent.

Appendix 1. Independence of A7-A11 w.r.t. B+

All matrices that follow are such that:
1. A total order 0 $ 1> $> ===> $ q is dened on the set of truth values V = {0> 1> ===> q}.
2. For all d> e M V, d j e and d 2 e are understood as pd{{d> e} and plq{d> e},
respectively.
3. Designated values are starred.
Each matrix veries B+ plus four of the ve axioms A7-A11 whereas falsifying the
fth one. Each one of the ve matrices is the simplest one supporting its respective
claim. These matrices have been found by using MaGIC, the matrix generator developed by John Slaney (see [28]). In case a tester is needed, the reader may use that in
[12].

38

Matrix 1. Independence of A7:

<

*1

*2

Falsies A7 (y(D) = 2, y(E) = 1).

Matrix 2. Independence of A8:

<

*1

Falsies A8 (y(D) = 0, y(E) = 1).

Matrix 3. Independence of A9:

<

*1

Falsies A9 (y(D) = 0, y(E) = 1).

Matrix 4. Independence of A10:

<

*2

Falsies A10 (y(D) = 2, y(E) = 0)

Matrix 5. Independence of A11:

39

<

*3

Falsies A11 (y(D) = 2, y(E) = 1)

Acknowledgements - Work supported by research project FFI2011-28494 nanced by the

Spanish Ministry of Economy and Competitiviness.
- G. Robles is supported by Program Ramn y Cajal of the Spanish Ministry of Economy and
Competitiveness.
- We sincerely thank two anonymous referees of the JPL for their comments and suggestions
on a previous version of this paper.

References
1. Avron, A. (1991). Natural 3-Valued Logics. Characterization and Proof Theory. Journal
of Symbolic Logic, 56, 276-294.
2. Baaz, M., Zach, R. (1998). Compact Propositional Gdel Logics. 28th International Symposium on Multiple-valued Logic Proceedings (Fukuoka, Japan), 108-113.
3. Baaz, M., Preining, N, Zach, R. (2007). First-Order Gdel Logics. Annals of Pure and
Applied Logic, 147, 23-47.
4. Bimb, K., Dunn, J. M. (2008). Generalized Galois Logics. Relational Semantics of Nonclassical Logic Calculi, (CSLI Publications, Stanford, CA).
5. Brady, R. (1982). Completeness Proofs for the Systems RM3 and BN4. Logique et Analyse,
9-32.
6. Brady, R. (Ed.) (2003). Relevant Logics and their Rivals, vol. II, (Ashgate).
7. Carnielli, W., Coniglio, M., Marcos, J. (2007). Logics of Formal Inconsistency. (In D.
Gabbay and F. Guenthner (Eds.), Handbook of Philosophical Logic, vol. 14, Springer),
1-93.
8. Dunn, J. M. (1966). The Algebra of Intensional Logics. Doctoral dissertation, University
of Pittsburg. (Ann Arbor, University Microlms).
9. Dunn, J. M. (1976). Intuitive semantics for rst-degree entailments and coupled trees.
Philosophical Studies, 29, 149-168.

40
10. Dunn, J. M. (2000). Partiality and its Dual. Studia Logica, 65, 5-40.
11. Gabbay, D. and Guenthner, F. (Eds.), Handbook of Philosophical Logic, 2nd Edition,
Springer.
12. Gonzlez, C. (2012). MaTest. Available at http://ceguel.es/matest. (Last access
26/11/2012).
13. Hhnle, R. (2002). Advanced Many-Valued Logics. (In D. Gabbay and F. Guenthner (Eds.),
Handbook of Philosophical Logic, vol. 2, Springer), 297-395.
14. Karpenko, A. S. (2006). ukasiewicz Logics and Prime Numbers. (Luniver Press, Beckington).
15. ukasiewicz, J. (1920). On three-valued logic. (In J. ukasiewicz (1970)), 87-88.
16. ukasiewicz, J. (1970). Selected works. (North-Holland, Amsterdam).
17. ukasiewicz, J. and Tarski, A. (1930). Investigations into the sentential calculus. (In
ukasiewicz (1970)), 131-152.
18. Malinowski, G. (1993). Many-valued Logics. (Clarendon Press, Oxford).
19. McCall, S. (1967). Polish Logic: 1920-1939. (Oxford University Press).
20. Meyer, R. K., Routley, R. (1972). Algebraic analysis of entailment I. Logique et Analyse,
15, 407-428.
21. Priest, G. (2002). Paraconsistent Logic. (In D. Gabbay and F. Guenthner (Eds.), Handbook
of Philosophical Logic, vol. 6, Springer), 287-393.
22. Robles. G., Mndez, J. M. (2008). The basic constructive logic for a weak sense of consistency. Journal of Logic Language and Information, 17/1, 89-107.
23. Robles. G., Mndez, J. M. (Manuscript). A Routley-Meyer semantics for truth-preserving
and well-determined ukasiewiczs 3-valued logics.
24. Rose, A., Rosser, J. B. (1958). Fragments of many-valued statement calculi. Transactions
of the American Mathematical Society, 87, 1-53.
25. Routley, R., Meyer, R. K. (1972). Semantics of Entailment III. Journal of Philosophical
Logic, 1, 192-208.
26. Routley, R., Routley, V. (1972). Semantics of rst-degree entailment. Nos, 1, 335-359.
27. Routley, R. Meyer, R. K., Plumwood, V., Brady R. T. (1982a). Relevant Logics and their
Rivals, vol. 1. (Atascadero, CA: Ridgeview Publishing Co.).
28. Slaney, J. (1995). MaGIC, Matrix Generator for Implication Connectives: Version 2.1, Notes and Guide. (Canberra: Australian National University). URL:
http://users.rsise.anu.edu.au/~jks
29. Suszko, R. (1975). Remarks on ukasiewiczs three-valued logics. Bulletin of the Section
of Logic, 4, 87-90.

41
30. Tokarz, M. (1974). A method of axiomatization of ukasiewicz logics. Studia Logica, 33,
333-338.
31. Tsuji, M. (1998). Many-valued logic and Suszko thesis revisited. Studia Logica, 60, 299-309.
32. Tuziak, R. (1988). An axiomatization of the nite-valued ukasiewicz calculus. Studia
Logica, 47, 49-55.
33. Urquhart, A. (2002). Basic Many-Valued Logic. (In D. Gabbay and F. Guenthner (Eds.),
Handbook of Philosophical Logic, vol. 2, Springer), 249-295.
34. Van Fraasen, B. (1969). Facts and Tautological Entailments. The Journal of Philosophy,
67, 477-487.
35. Wajsberg, M. (1931). Axiomatization of the 3-valued propositional calculus. (In S. McCall
(1967)), 264-284.
36. Wojcicki, R. (1974). The Logics Stronger than ukasiewiczz Three-Valued Sentential Calculus - The Notion of Degree of Maximality versus the Notion of Degree of Completeness.
Studia Logica, 33, 201-214.
37. Wojcicki, R. (1984). Lectures on Propositional Calculi. Ossolineum.

&RQWDFWGHWDLOV

CONTACT DETAILS
Gemma Robles
Dpto. de Psicologa, Sociologa y Filosofa
Universidad de Len
Campus de Vegazana, s/n, 24071, Len, Espaa.
URL: http://grobv.unileon.es
E-mail: gemmarobles@gmail.com
Francisco Salto
Dpto. de Psicologa, Sociologa y Filosofa
Universidad de Len
Campus de Vegazana, s/n, 24071, Len, Espaa.
URL: http://www3.unileon.es/personal/wwdfcfsa/web/html/
E-mail: francisco.salto@unileon.es
Jos M. Mndez
Universidad de Salamanca
Campus Unamuno, Edicio FES
37007, Salamanca, Spain.
URL: http://web.usal.es/~sefus
E-mail: sefus@usal.es