Fuzzy Sets and Systems 157 (2006) 597 – 603

www.elsevier.com/locate/fss

What is mathematical fuzzy logic夡

Petr Hájek
Institute of Computer Science, Academy of Sciences of the Czech Republic, 182 07 Prague, Czech Republic

Available online 2 November 2005

Abstract
The paper comments on the development and present state of fuzzy logic as a kind (branch) of mathematical logic. It is meant
just as a contribution to the discussion on what fuzzy logic is, not as a systematic presentation of mathematical fuzzy logic.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Fuzzy logic; Mathematical logic; Triangular forms

1. Origin, motivation, task

We mention some milestones of the development of fuzzy logic understood as a branch of mathematical (symbolic)
logic.
Clearly, the story begins by Zadeh’s first paper [40] on fuzzy sets. The term “fuzzy logic” is not used; but Zadeh
mentions Kleene three-valued logic (just in passing).
Goguen’s 1968–1969 paper [17] speaks on logic of inexact concepts but the term “fuzzy logic” occurs there (on p.
359; is this the first occurrence of the term in the literature?) The paper is very general, introduces algebras called closg,
very near to algebras presently called residuated lattices, as algebras of truth functions of connectives for many-valued
logics of inexact concepts, As an example he presents the unit real interval [0, 1] with product and its residuum (Goguen
implication), thus a particular t-norm algebra (not speaking on t-norms). Zadeh has written several papers on fuzzy
logic; an early paper is his “Fuzzy logic and approximate reasoning” [41] from 1975 (reprinted in [30]), where he
uses connectives of Łukasiewicz logic min, max, 1 − x, Łukasiewicz implication—but not strong conjunction. Note
that what we call Łukasiewicz strong (or bold) conjunction or Łukasiewicz t-norm (the t-norm whose residuum is
Łukasiewicz implication) was never explicitly used by Łukasiewicz. The first explicit use of this conjunction in the
context of Zadeh’s fuzzy logic appears to be the paper [16] by Giles. 1
In Zadeh’s understanding, fuzzy logic uses some many-valued logic but works with fuzzy truth values and his
linguistic variables. Zadeh (and the majority of researchers up to today, including the present author) understands fuzzy
logic as truth functional, i.e. having some truth functions for connectives determining the truth value of a compound

夡 The author was partly supported by Grant A100300503 of the Grant Agency of the Academy of Sciences of the Czech Republic and partly by
the Institutional Research Plan AV0Z10300504.
E-mail address: hajek@cs.cas.cz.
1 But Giles refers to papers by Chang and Klaua from mid-sixties using the bold conjunction in studying some generalized set theories over
Łukasiewicz or similar logics.

0165-0114/$ - see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.fss.2005.10.004
598 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603

formula constructed using a connective uniquely from the truth values of the formula(s) to which the connective is
applied. This makes fuzzy logic different from any probability theory (probability is evidently not truth functional). A
general, not necessarily truth-functional approach is possible and was first investigated by Pavelka [36]; but Pavelka
also developed in the quoted series of papers a truth-functional fuzzy logic which can be shown to be just Łukasiewicz
([0, 1]-valued) logic extended by truth constant r̄ for each r ∈ [0, 1]. This approach has been continuously further
developed by V. Novák. For a modern treatment of a non-truth-functional fuzzy logic see [15].
Triangular norms (t-norms) enter the game in early 80’s of the last century. Dubois and Prade [8] wrote: “It is
now well-known that a good model of a fuzzy set-theoretic intersection, or equivalently of a conjunction function in
multivalued logics, is a triangular norm” and they refer to [1,7,38], also to papers from 1980. Gottwald presents in
his 1984 German book [18] left-continuous t-norms and their residua, referring to Pedrycz [37]. Since then t-norms
become “standard” semantics of fuzzy conjunction (cf. e.g. the 1995 monograph [29]).
At this place it is proper to make the distinction between fuzzy logic in broad and narrow sense (the term was coined
by Zadeh), the former being a discipline using the notion of fuzzy logical connectives and other notions of the theory of
fuzzy sets to develop methods of a sort of applied “approximate reasoning” (fuzzy IF-THEN rules, fuzzy controllers,
fuzzy clustering and more or less anything else), whereas the latter discipline (fuzzy logic in the narrow sense) develops
deductive systems of fuzzy logic as a many-valued logic with a comparative notion of truth very much in the style of
classical mathematical logic (propositional and predicate calculi; axiomatization, (in)completeness, complexity, etc.).
It seems reasonable to call this kind of fuzzy logic (in the narrow sense) just mathematical fuzzy logic. The basic
monograph on it is Hájek’s [21] from 1998; Gottwald’s Treatise on many-valued logic [19] contains a part devoted to
(mathematical) fuzzy logic; the monograph [33] is a book on (mathematical) fuzzy logic stressing the approach using
truth constants in the language. I also mention Turunen’s [39]. There is a vast number of papers contributing to this
topic; the references of this paper are by far incomplete.
What is the motivation and the task of mathematical fuzzy logic? Inspired by the “fuzzy logic in broad sense” (which
is often developed by non-logicians) to construct symbolic logical calculi that can serve as foundation for the “broad-
sense”-methods and, moreover, are meaningful as logics of inference under vagueness (or, as Goguen said, logics of
inexact concepts). The reader is invited to take this as our (imprecise, informal) explanation of the term “mathematical
fuzzy logic”; I offer no formal definition. Apparently systems that are t-norm based are of central importance; but
this should be understood in a general sense, open to many alternatives. In the next two sections I survey a notion of
t-norm based mathematical fuzzy logic with its double semantics—standard (with algebras of truth functions given by
some t-norms) and general (working with algebras of truth functions taken from varieties generated by some standard
algebras).

2. t-norm based fuzzy propositional logic

We quickly survey main facts; for a recent detailed survey see [20].

2.1. Continuous t-norms, basic fuzzy predicate logic

As commonly known, a t-norm is a binary operation ∗ on the real unit interval [0, 1] which is commutative, associative,
non-decreasing in each argument and has 0 and 1 as zero and unit element. Its residuum is a binary operation on [0, 1]
defined as x ⇒ y = max{z | x ∗ z
y}. A t-norm has residuum iff it is left continuous. Structure of continuous t-norms
is well understood: roughly, each continuous t-norm is “constructed from copies of” three most important continuous
t-norms called Łukasiewicz, Gödel and product t-norms. (No analogous characterization is known for left continuous t-
norms.) A t-algebra [0, 1]∗ given by a continuous t-norm ∗ is the algebra ([0, 1], ∗, ⇒, 0, 1) with two binary operations
and two constants. Defined operations are (−)x = x ⇒ 0, x ∩ y = x ∗ (x ⇒ y), x ∪ y = ((x ⇒ y) ⇒ y) ∩ ((y ⇒
x) ⇒ x). 2 For each continuous t-norm ∗, x ∩ y is min(x, y) and x ∪ y is max(x, y).
This leads to a propositional calculus with starting connectives & (conjunction), → (implication) and truth con-
stants 0̄, 1̄. Defined connectives are ∧, ∨, ¬ (possibly others). Any t-algebra [0, 1]∗ (∗ continuous) serves as the
algebra of truth functions of connectives, which gives the notion of ∗-tautologies (formulas having ∗-value 1 for any

2 The last identity is called Dummett’s identity.

 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603 599

evaluation of propositional variables by truth values from [0, 1]). There is a simple axiom system defining (together
with the deduction rule modus ponens) the basic fuzzy propositional logic BL (see [21]) which satisfies standard
completeness theorem: a formula
is provable in BL iff for each continuous t-norm ∗,
is a ∗-tautology (proved
in [4]). Formulas satisfying the latter condition are called standard BL-tautologies. Axioms for ∗-tautologies for a
particular ∗ are got as an extension of axioms of BL and for some t-norms (Łukasiewicz, Gödel, product) have a very
simple form.
The reader may ask if there is any natural interpretation of truth-functional semantics of connectives in fuzzy logic:
if the truth degree of
,  is x, y, respectively, what is the truth degree of
& computed as x ∗ y? There is no definitive
answer to this question but there are remarkable interpretations by Paris [34,35] for Gödel, product and Łukasiewicz
and by Cicalese and Mundici [3] for BL using multi-channel Rényi-Ulam games (generalizing previous Mundici’s
interpretation of Łukasiewicz logic using Ulam games with lies).
Up to now we have dealt with standard semantics, having the unit real interval [0, 1] for the set of truth degrees and
some continuous t-norms algebra [0, 1]∗ for the algebra of truth functions of connectives. Let K be a non-empty set
of continuous t-norms (think of the set of all continuous t-norms or a singleton containing your favourite t-norm). A
standard K-tautology is of course a formula being a ∗-tautology for each ∗ in K.
Now let Var(K) be the variety of algebras A = (A, ∗, ⇒, 0A , 1A ) generated by all t-algebras [0, 1]∗ for ∗ in K. It is
easy to see that A ∈ Var(K) iff each standard K-tautology is an A-tautology (has identically the value 1A if computed
using A as the algebra of truth functions). Thus defining a general K-tautology as a formula which is an A-tautology for
each A ∈ K we immediately get that general K-tautologies coincide with standard K-tautologies. What one has to know
is an equivalent algebraical characterization of elements of Var(K) (general K-algebras). For BL (i.e. K consisting of
all continuous t-norms) we get the class of BL-algebras, i.e. residuated lattices A = (A, ∗, ⇒, 0, 1, ∩, ∪) where the
operations ∩ and ∪ are defined from ∗, ⇒ as above, and A is a prelinear residuated lattice, i.e. (A, ∩, ∪, 0, 1) is a
lattice with extremal elements 0, 1, (A, ∗, 1) is a commutative monoid, ⇒ is the residuum of ∗ (i.e. for each x, y, z,
z
x ⇒ y iff x ∗ z
y) and the prelinearity says that for all x, y, (x ⇒ y) ∪ (y ⇒ x) = 1.
Clearly, not each BL-algebra is linearly ordered (think of the direct product of two t-algebras); but each BL-algebra
is a subalgebra of the direct product of some linearly ordered BL-algebras (of BL-chains). This is called the subdirect
representation property. It is important that we are not obliged to work with linearly ordered systems of truth degrees;
but admitting partial order we do not change the logic.
General semantics for particular t-norms is well understood; in particular, for Łukasiewicz it is formed by the variety
of MV-algebras, for Gödel G-algebras (prelinear Heyting algebras) and for product by so-called product algebras (see
[21,19,20] and references thereof). Note that to keep this paper short, I do not discuss questions of strong completeness
(semantic characterization of provability of formulas in theories over a fuzzy logic); see the references just mentioned.
Similarly for results on computational complexity.

2.2. Generalizing BL

The advantage of the logic of continuous t-norms include definability of lattice connectives from conjunction and im-
plication, known structure of continuous t-norms, well-known particular cases (Łukasiewicz, Gödel, product). Clearly
the implication is a logical connective of utmost interest and to have the residuated implication (which has excel-
lent logical properties) we have to work with left-continuous t-norms, thus (full) continuity is not necessary. In their
pioneering paper [10] Esteva and Godo elaborated the logic MTL (monoidal t-norm fuzzy logic) and the correspond-
ing variety of MTL-algebras (just giving up the condition x ∩ y = x ∗ (x ⇒ y) and the corresponding axiom of
BL, whereas adding some natural axioms on ∧). Their logic is very much analogous to BL, but is the logic of left
continuous t-norms in the same meaning as BL is the logic of continuous t-norms (this follows from the results of
[28]). Stronger logics IMTL and MTL generalizing Łukasiewicz and product logic were defined and showed to be
logics of some classes of left continuous t-norms (an analogous generalization of Gödel logic gives just Gödel logic
itself) [11,27].
Let us stress that in MTL-algebras starting operations are ∗, ⇒ and ∩ (since ∩ is not definable from ∗ and ⇒), thus
the starting connectives of MTL are &, ⇒, ∧ (and truth constants 0̄, 1̄); the variety of MTL-algebras is generated by
t-norm algebras [0, 1]∗ where ∗ is a left-continuous t-norm and the operations are ∗, ⇒, ∩; constants are 0 and 1. The
union (supremum) is defined by the same formula as above. Also note at this occasion that in the references given the
class of algebras forming the general semantics of a logic (BL, Ł, . . . , MTL, . . .) is not defined as a variety generated
600 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603

by some set of t-norm algebras but as a class of residuated lattices satisfying some conditions (guaranteeing soundness
w.r.t. a given axiomatization of the logic); showing that it is just the variety generated by the respective set of t-norms
then becomes a result on standard completeness (saying that standard tautologies coincide with general tautologies).
MTL-algebras also have subdirect representation property.
It is rather interesting to investigate fragments of our logics resulting by restricting the language. First one may
study falsity free fragments of BL, MTL etc. just deleting 0̄ from the language of the logic and deleting the constant
0 from the language of the corresponding algebras. The t-norms remain as they have been; but the variety generated
by the respective reduct of t-norm algebras may (and does) contain algebras that are not just reducts of algebras from
the variety with the original richer language. To be concrete: let K be the variety of all continuous t-norms. Consider
the variety generated by the set of algebras ([0, 1], ∗, ⇒, 1) with ∗ in K. You get the variety of basic hoops [12]; the
corresponding algebraic definition is as follows:
A hoop is an algebra A = (A, ∗, ⇒, 1) such that ∗ is a binary commutative operation with the unit element 1 and
⇒ is a binary operation satisfying, for each x, y, z,

x ⇒ x = 1, x ∗ (x ⇒ y) = y ∗ (y ⇒ x), (x ∗ y) ⇒ z = x ⇒ (y ⇒ z).

It follows that ∗ is associative. Order x
y is defined as x ⇒ y = 1. This makes a hoop to a lattice that may and
may not have a least element. A hoop is basic if it satisfies, for all x, y, z,

((x ⇒ y) ⇒ z) ∗ ((y ⇒ x) ⇒ z)
z.

There is a very useful representation of BL-chains using basic hoop chains. The axioms of the hoop logic result from
axioms of BL just deleting the only axiom containing 0̄, namely 0̄ →
(ex falso quodlibet). Since hoop logic does
not have 0̄, it does not have negation; it is just the continuous t-norm logic without negation (and 0̄).
Similarly one can investigate the positive part of MTL; the corresponding algebras are semihoops (see [12]). The
language of semihoops has starting connectives &, →, ∧ (and ∧). ¯ One may go still further and consider t-norm algebras
given by left-continuous t-norm and using only ∗, ⇒, 1. The corresponding variety is the variety of basic quasihoops
[24]; in it the definition x
y iff x ⇒ y 1 gives only a quasiorder (preorder).
For logics of the non-commutative (variants of) left-continuous t-norms see [22,23]. Let us also mention logics based
on two t-norms (Łukasiewicz and product—the logic Ł, see e.g. [9]), extensions of the described logics by hedges,
notably by the so-called Baaz Delta [2] as well as logics with two negations [13].

2.3. Adding truth constants

As mentioned above, Pavelka in his [36] investigated a logic that turns out to be just Łukasiewicz logic with its
standard semantics extended by truth constants r̄ for each r ∈ [0, 1] (r̄ has just the truth value r). Later I suggested
to use only r̄ for rational r ∈ [0, 1] which keeps the language countable and preserves Pavelka’s results. V. Novák
investigates this logic as a logic with evaluated syntax EvŁ (see [32]). Relying on Łukasiewicz logic is explained by
the fact that Łukasiewicz t-norm is the only one whose residuum is continuous and consequently admits a certain
(Pavelka-style) completeness theorem for theories over this logic. (Since I do not discuss theories I do not go into
details; but we meet Pavelka-style completeness in the section on predicate logic.) Here we comment on extending Ł,
G,  by rational truth constants and the corresponding varieties. Thus let L be any of the just mentioned logics. RL
is the extension of the language of L by truth constant r̄ for each rational r ∈ [0, 1] and of the axioms of L by the
following bookkeeping axioms:

(r̄&s̄) ≡ (r ∗ s), (r̄ → s̄) ≡ (r ⇒ s).

Let [0, 1]R L be the corresponding t-norm algebra, now expanded by names r̄ of all rational elements of [0, 1]. Let
Var([0, 1]R L ) be the corresponding variety. For L being Ł, G,  we have standard completeness: For each formula
,

is provable (has a proof) in RL iff
is a [0, 1]R L -tautology. This is equivalent to the statement that Var([0, 1]R L )
is just the class of all RL-algebras, i.e. for Ł all MV-algebras with constants r̄ for rational r ∈ [0, 1] satisfying the
bookkeeping axioms, similarly for G (G-algebras) and  (product algebras). For Ł the result is in [21]; for G and 
the results are very new (see [14,5]). (For theories over RL the situation is different, details are omitted here).
 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603 601

3. Fuzzy t-norm based predicate logics

Admittedly propositional logics are important and interesting but it is only predicate logic (logic with quantifiers and
structured atomic formulas) which is most important (and possibly most interesting). Let L be any of our propositional
logics investigated till now; given by a set K of (left) continuous t-norms one can build a predicate logic L∀ with
“classical” quantifiers ∀, ∃ over it. We shall quickly survey main facts; note that a detailed survey paper on fuzzy
predicate logics [6] is under preparation. Let a predicate language (predicates with arities, object constants) be given.
A standard interpretation of our language is a structure

M = (M, (RP )P predicate , (mc )c constant )

with a non-empty domain M, where for each n-ary P, RP : M n → [0, 1] (n-ary fuzzy relation) and mc ∈ M for each
constant c. Given a t-norm ∗ from K, the value 
∗M,v of a formula
in M (given by ∗ and an evaluation v of object
variables by elements of M) is defined in the obvious (Tarski-style) way, for connectives using the operations given by
∗, and defining the truth value of a formula beginning by ∀ or ∃ as the infimum/supremum of the values of its instances.
(See e.g. [21] for details.) A formula
is a (standard) ∗-tautology if 
∗M,v = 1 for each M and v;
is a standard
K-tautology if it is a ∗-tautology for each ∗ in K. This is the standard semantics of the predicate logic L∀. To define the
general semantics we define, for each A ∈ Var(K), an A-interpretation in the same way as above but RP being A-fuzzy
relations, i.e. RP : M n → A. The definition of 
A M,v is obvious, but need not be total since the suprema/infima in
question need not exist. The interpretation is safe (A-safe) if 
A M,v is defined for each
and v.
A formula
is an A-tautology if 
A M,v = 1 for each safe A-interpretation A and each v;
is a general L∀-
tautology, if it is an A-tautology for each linearly ordered A ∈ Var(K);
is a general L∀− -tautology if it is an
A-tautology for each A ∈ Var(K). The axioms of L∀ are the axioms of L (built from predicate formulas) plus five
axioms for quantifiers: two substitution axioms: (∀x)
(x) →
(t),
(t) → (∃x)
(x) (for t substitutable for x in
),
two axioms on moving quantifiers over implication: (∀x)( →
) → ( → (∀x)
), (∀x)(
→ ) → ((∃x)
→ )
(x not free in ) and last axiom (∀3) : (∀x)( ∨
) → ( ∨ (∀x)
) (x not free in ). Deduction rules are modus ponens
and generalization.
The strong completeness theorem (valid for each of our discussed logics L∀) says that for each theory T over L∀
and each formula
, T proves
over L∀ if
is true in each A-model of T , for each linearly ordered A ∈ Var(K);
an A-model of T being a safe interpretation M in which all elements of T have the A-value 
A M,v = 1 (for all v).
The logic L∀− results from L∀ by deleting the last axiom (∀3); the strong completeness for theories over L∀− results
by replacing L∀ by L∀− and deleting the restriction “linearly ordered”, thus taking models over all A ∈ Var(K). For
some logics the proofs are fully analogous to the proofs for BL∀ (basic fuzzy predicate logic see [21]), for some other
ones additional work is necessary; e.g. for the logics with non-commutative conjunction [26] and also for the logic of
(basic) quasihoops (in which the disjunction ∨ is not available, see [24]), one has to formulate the axiom (∀3) using
only & and → (which turns out to be possible).
Thus general semantics behaves well; what about standard semantics? For Gödel logic G∀ the standard tautologies
coincide with general tautologies and we have the standard strong completeness (provability in a theory over G∀ is
equivalent to truth in all [0, 1]G -models), but this is the only continuous t-norm with standard completeness. For each
other continuous t-norm ∗, standard ∗-tautologies form a proper subset of the set of general ∗-tautologies and the
former set is much more undecidable than the latter, e.g. the set of standard tautologies of BL∀ is not arithmetical in
the sense of arithmetical hierarchy (cf. the survey paper [25]). Surprisingly enough, for MTL∀ we do have standard
completeness, see [31].
For predicate Łukasiewicz logic with rational truth constants (alias Rational Pavelka predicate logic) we have a
result called Pavelka completeness: given a theory T over RŁ∀, define the T-provability degree of
to be |
|T =
sup{r|T ⵫r̄ →
} and the T-truth degree of
to be 
T = inf{
M |M a [0, 1]RŁ -model of T }. The theorem says
that |
|T = 
T (provability degree equals truth degree). But RŁ∀ does not have standard completeness theorem
(since even Ł∀ does not have it). A formula
is a standard RŁ∀-tautology iff for each r < 1, RŁ∀ proves r̄ →
(saying
that the truth value of
is at least r) but it can happen that there is no proof of
itself in RŁ∀. Pavelka completeness
fails for any continuous t-norm ∗ different from Łukasiewicz due to the fact that the corresponding residuum is not
continuous. (This was shown by Pavelka himself in his paper.)
602 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603

4. Conclusion

We have sketched some basic facts on t-norm based fuzzy logics, both propositional and predicate logics. We
consider these logics to be of central importance for mathematical fuzzy logics as discussed above. They are t-
norm based, i.e. their standard semantics is given by (left) continuous t-norms and their residua, but are general,
i.e. develop their general semantics for which the algebras of truth functions of connectives are taken from the va-
riety generated by the corresponding standard algebras. The interplay of both semantics shows the richness of these
logics: general semantic is completely axiomatizable and is a well understood deductive system, whereas the stan-
dard semantics exhibits surprising (and beautiful) features of non-axiomatizability; the axiomatization of the general
semantic is of course sound for the standard semantics, thus it is a powerful deductive system also for it. Math-
ematical t-norm based fuzzy logic understood in the described way is both a very interesting field of research in
“pure” (symbolic, formal) logic and a very useful means of foundational analysis of methods of the fuzzy logic in
broad sense.

References

[1] C. Alsina, E. Trillas, L. Valverde, On non-distributive logical connectives for fuzzy sets theory, Busefal 3 (1980) 8–29.
[2] M. Baaz, Infinite-valued Gödel logics with 0–1-projections and relativizations, in: P. Hájek (Ed.), Gödel’96, Lecture Notes in Logic, vol. 6,
Springer, Berlin, 1996, pp. 23–33.
[3] F. Cicalese, D. Mundici, Recent developments of feedback coding and its relations with many-valued logic, 2005, preprint.
[4] R. Cignoli, F. Esteva, L. Godo, C. Noguera, P. Savický, On product logic with truth constants, Draft, 2005.
[5] R. Cignoli, F. Esteva, L. Godo, A. Torrens, Basic fuzzy logic is the logic of continuous t-norms and their residua, Soft Computing 4 (2000)
106–112.
[6] P. Cintula, P. Hájek, On t-norm based fuzzy predicate logics—a survey, paper in preparation.
[7] D. Dubois, Triangular norms for fuzzy sets, Proc. Second International Seminar on Fuzzy Set Theory Linz, 1980, pp. 39–68.
[8] D. Dubois, H. Prade, A theorem on implication functions defined from triangular t-norms, Stochastica VIII (1984) 267–279.
[9] F. Esteva, L. Godo, Putting together Łukasiewicz and product logic, Mathware and Soft Comput. 6 (1999) 219–234.
[10] F. Esteva, L. Godo, Monoidal t-norm based logic: towards a logic for left-continuous t-norms, Fuzzy Sets and Systems 123 (2001) 271–288.
[11] F. Esteva, L. Godo, J. Gispert, F. Montagna, On the standard completeness of some axiomatic extensions of the monoidal t-norm logic, Studia
Logica 71 (2002) 199–226.
[12] F. Esteva, L. Godo, P. Hájek, F. Montagna, Hoops and fuzzy logic, J. Logic Comput. 13 (2003) 532–555.
[13] F. Esteva, L. Godo, P. Hájek, M. Navara, Residuated fuzzy logics with an involutive negation, Archive for Mathematical Logic, vol. 39, 2000,
pp. 103–124.
[14] F. Esteva, L. Godo, C. Noguera, On rational weak nilpotent minimum logics, Proc. IPMU 2004, Universita Sapienza Perugia, 561–568.
[15] G. Gerla, Fuzzy Logic—Mathematical Tool for Approximate Reasoning, Kluwer Academic Publishers, Dordrecht, 2001.
[16] R. Giles, Lukasiewicz logic and fuzzy theory, Internat. J. Man-Mach. Stud. 8 (1976) 313–327.
[17] J.A. Goguen, The logic of inexact concepts, Synthese 19 (1968–1969) 325–373.
[18] S. Gottwald, Mehrwertige Logik, Eine Einführung in Theorie und Anwendungen, Akademie, Verlag, Berlin, 1984.
[19] S. Gottwald, A treatise on many-valued logic, Research Studies Press, 2000.
[20] S. Gottwald, P. Hájek, Triangular norm based mathematical fuzzy logics, in: Logical, Algebraic, Analytic and Probabilistic Aspects of Triangular
Norms, Elsevier, Amsterdam, 2005.
[21] P. Hájek, Metamathematics of Fuzzy Logic, Kluwer Academic Publishers, Dordrecht, 1998.
[22] P. Hájek, Fuzzy logics with non-commutative conjunctions, J. Logic Comput. 13 (2003) 469–479.
[23] P. Hájek, Observations on non-commutative fuzzy logics, Soft Comput. 8 (2003) 28–43.
[24] P. Hájek, One more variety in fuzzy logic: quasihoops, Technical Report 937, Institut of Computer Science Academic Science Prague, 2005.
[25] P. Hájek, Fuzzy logic and arithmetical hierarchy–a survey, Soft Comput. 9 (2005) 935–941.
[26] P. Hájek, J. Ševčík, On fuzzy predicate calculi with non-commutative conjunction, in: East West Fuzzy Colloquium 2004–Zittau/Görlitz, 2004,
pp. 103–110.
[27] R. Horčík, Algebraic properties of fuzzy logics, Thesis, Czech Technical University Prague, 2005.
[28] S. Jenei, F. Montagna, A proof of standard completeness for Esteva and Godo’s logic MTL, Studia Logica 70, 183–192.
[29] G.J. Klir, B. Yuan, Fuzzy Sets and Fuzzy Logic—Theory and Applications, Prentice Hall PTR, 1995.
[30] J.G. Klir, B. Yuan, (Eds.) Fuzzy Sets, Fuzzy Logic and Fuzzy Systems—Selected Papers by Lotfi A. Zadeh, World Scientific, Singapore, 1996.
[31] F. Montagna, H. Ono, Kripke semantics undecidability and standard completeness for Esteva and Godo’s logic MTL∀, Studia Logica 71 (2002)
227–245.
[32] V. Novák, On the syntactico-semantical completeness of first-order fuzzy logic. Part I (Syntax and semantic), Part II (Main results), Kybernetika
26 (1990) 47–66 134–154.
[33] V. Novák, I. Perfilieva, J. Močkoř, Mathematical Principles of Fuzzy Logic, Kluwer Academic Publishers, Dordrecht, 2000.
 P. Hájek / Fuzzy Sets and Systems 157 (2006) 597 – 603 603

[34] J. Paris, A semantics for fuzzy logic, Soft Comput. 1 (1997) 143–147.
[35] J. Paris, Semantics for fuzzy logic supporting truth functionality, in: Discovering the World with Fuzzy Logic, Physica, Verlag, Wurzburg, 2000,
pp. 82–104.
[36] J. Pavelka, On fuzzy Logic I, II, III, Zeitschr. f. Math. Logik und Grundlagen der Math. 25 (1979) 45–52 119–134, 447–464.
[37] W. Pedrycz, Sterowanie i systemy rozmyte Zeszyty naukowe Politechniki Ślaskiej, Ser. Automatyka, Nr. 70, Gliwice, 1983.
[38] H. Prade, Unions et intersections d’ensembles flous, Busefal 3 (1980) 58–62.
[39] E. Turunen, Mathematics Behind Fuzzy Logic, Physica, Verlag, Wurzburg, 1999.
[40] L. Zadeh, Fuzzy sets. Inform. and Control 8 (1965) 338–353.
[41] L. Zadeh, Fuzzy logic and its application to approximate reasoning, Inform. Process. 74 (1974) 591–594.