Documentation

Linglib.Phenomena.Polarity.Studies.Ladusaw1979

Ladusaw (1979): Polarity Sensitivity as Inherent Scope Relations #

@cite{ladusaw-1979}

Ladusaw's dissertation established the foundational generalization connecting NPI licensing to downward entailingness (DE). The core claim:

Weak NPIs are licensed in downward-entailing contexts.

This file bridges the GQ monotonicity proofs from Core.Quantification and Semantics.Quantification.Quantifier to the NPI licensing classification indexed by Typology.PolarityItem.LicensingContext, making the DE ↔ NPI connection formally explicit.

Key connections #

  1. Scope DE → weak NPI licensing: ScopeDownwardMono licenses weak NPIs in the scope of a quantifier (e.g., "No student saw anyone").
  2. Restrictor DE → weak NPI licensing: RestrictorDownwardMono licenses weak NPIs in the restrictor (e.g., "Everyone who saw anyone was questioned").
  3. Left anti-additivity → strong NPI licensing: LeftAntiAdditive licenses strong NPIs (e.g., "Nobody lifted a finger"). Mere DE ("few") is insufficient for strong NPIs.

The monotonicity-based licensing strength of a context. @cite{ladusaw-1979}: DE licenses weak NPIs. @cite{zwarts-1998}: anti-additive licenses strong NPIs.

A coarsening of Core.NaturalLogic.DEStrength that collapses the .antiMorphic and .antiAdditive cases (Ladusaw treats them identically: both license strong NPIs).

Instances For
    @[implicit_reducible]
    Equations
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      Classify NPI licensing contexts by their monotonicity-based strength, derived from Semantics.Polarity.Licensing.contextProperties. The Ladusaw classification is a coarsening of the Icard signature lattice.

      Equations
      Instances For

        The core Ladusaw generalization: scope-DE quantifiers license weak NPIs in their scope. Formally: ScopeDownwardMono q implies that the scope of q is a weak-NPI-licensing environment.

        Verified instances:

        • no_scope_down: "No student saw anyone" ✓
        • few_scope_down: "Few students saw anyone" ✓
        Instances For

          The restrictor Ladusaw generalization: restrictor-DE quantifiers license weak NPIs in their restrictor.

          Verified instances:

          • every_restrictor_down: "Everyone who saw anyone was questioned" ✓
          • no_restrictor_down: "No one who saw anyone was questioned" ✓
          Instances For

            "No" licenses weak NPIs in restrictor via restrictor-↓ monotonicity.

            Equations
            Instances For

              @cite{zwarts-1998}: anti-additive contexts license strong NPIs. LeftAntiAdditive q means the restrictor of q is anti-additive, licensing strong NPIs like "lift a finger" and "in years".

              Verified instances:

              • every_laa: "Everyone who ever lifted a finger..." ✓
              • no_laa: "Nobody lifted a finger" ✓

              Counter-example: "few" is merely DE, not anti-additive:

              • *"Few people lifted a finger to help" ✗
              Instances For

                "Every" is left-anti-additive → licenses strong NPIs in restrictor. P&W Prop 13: the restrictor of "every" is anti-additive (not just DE).

                Equations
                Instances For

                  "No" is left-anti-additive → licenses strong NPIs in restrictor.

                  Equations
                  Instances For

                    Scope-level anti-additivity also licenses strong NPIs. RightAntiAdditive q means the scope of q is anti-additive. "Nobody lifted a finger" is licensed by scope-level AA of no.

                    Instances For

                      "No" is right-anti-additive → licenses strong NPIs in scope. "Nobody lifted a finger" / "Nobody saw anyone".

                      Equations
                      Instances For

                        "Few" is DE but NOT right-anti-additive in scope: few(R, S∨S') ≠ few(R,S) ∧ few(R,S') in general. This is why *"Few people lifted a finger" is bad — strong NPIs need AA.

                        Counterexample: R = {john, mary, pizza}, S = {john}, S' = {mary}.

                        • few(R, S∨S') = (2 < 1) = false
                        • few(R, S) = (1 < 2) = true, few(R, S') = (1 < 2) = true
                        • true ∧ true ≠ false

                        The Ladusaw hierarchy: AA ⊂ DE ⊂ NV (nonveridical). Strong NPIs need AA; weak NPIs need DE; "any" in questions needs NV. "Few" is DE but not AA, explaining the licensing contrast: ✓ "Few students saw anyone" (weak NPI in DE) vs ✗ "*Few people lifted a finger" (strong NPI needs AA).