Łukasiewicz Logic

Origin. Jan Łukasiewicz introduced three-valued logic (1920), then infinitely-valued (1930). Foundation for fuzzy logic. MV-algebras as algebraic semantics (Chang, 1958). One of the main many-valued logics. Motivated by future contingents.

Models. Truth values in [0,1]. Classical: {0, 1}. Łukasiewicz: [0,1] (or finite subsets). Connectives: ¬x = 1-x, x ⊕ y = min(1, x+y). Designated: value 1 (or > 0 for tautologies). Models vagueness and degrees of truth.

Formalism.

Connectives:

• ¬x = 1 - x (negation)
• x ⊕ y = min(1, x + y) (strong disjunction)
• x ⊗ y = max(0, x + y - 1) (strong conjunction)
• x → y = min(1, 1 - x + y) (implication)
• x ∧ y = min(x, y) (weak conjunction/lattice meet)
• x ∨ y = max(x, y) (weak disjunction/lattice join)

Three-valued: Values: 0, ½, 1. Third value: "indeterminate" or "possible."

Infinite-valued: Values: rational or real [0,1]. Tautology: evaluates to 1 for all valuations.

Axioms (Łukasiewicz):

1. φ → (ψ → φ)
2. (φ → ψ) → ((ψ → χ) → (φ → χ))
3. (¬φ → ¬ψ) → (ψ → φ)
4. ((φ → ψ) → ψ) → ((ψ → φ) → φ)

MV-algebras: Algebraic semantics: (A, ⊕, ¬, 0). Categorically equivalent to certain lattice-ordered groups.

Symbols.

Metatheory. Rose and Rosser (1958) proved the Łukasiewicz axioms complete for the standard [0,1] semantics; Chang (1958, 1959) gave the algebraic proof, showing every MV-algebra is a subdirect product of MV-chains and that the equations valid in [0,1] are exactly those valid in all MV-algebras, so Ł∞ is sound and complete both for MV-algebras and for the single standard algebra. McNaughton's theorem (1951) identifies the definable n-ary functions of Ł∞ exactly with the continuous piecewise-linear functions [0,1]ⁿ → [0,1] whose pieces have integer coefficients. Mundici's Γ functor (1986) is a categorical equivalence between MV-algebras and abelian lattice-ordered groups with strong unit, which transports the group theory to the logic. Mundici (1987) settled the complexity: tautologyhood in Ł∞ is coNP-complete, the same class as classical validity; each finite-valued Łₙ is decidable by evaluation over its finite matrix. First-order Ł∞ is not recursively axiomatizable — Scarpellini (1962) showed the standard-model validities are not recursively enumerable, and that set is Π₂-complete.

Applies to. Fuzzy logic and control. Vagueness modeling. Approximate reasoning. Future contingents (philosophical). Hardware with undefined states. Default reasoning. Partial information.

Limitations. The classical laws hold only in their strong forms: φ ⊕ ¬φ and ¬(φ ⊗ ¬φ) are valid, but φ ∨ ¬φ takes value max(x, 1−x) and φ ∧ ¬φ takes min(x, 1−x), so excluded middle and non-contradiction fail for the lattice connectives. Strong conjunction is not idempotent, so contraction fails and only a local deduction theorem is available: Γ, φ ⊢ ψ iff Γ ⊢ φⁿ → ψ for some n, with the n depending on the derivation. By McNaughton's characterization the definable functions all have integer coefficients, so Ł∞ cannot define x/2 or any constant other than 0 and 1 — truth constants must be added by hand, which is what Rational Pavelka logic does. Ł∞ is explosive rather than paraconsistent: no valuation designates both φ and ¬φ. Truth-functionality fixes the degree of a compound from the degrees of its parts, so borderline sorites cases receive determinate intermediate values and higher-order vagueness is untouched. First-order Ł∞ has no complete recursive axiomatization at all.

© 2026 Lingenic LLC