Documentation

Linglib.Phenomena.Questions.Studies.Uegaki2022

Uegaki (2022) @cite{uegaki-2022} #

Question-orientedness and the Semantics of Clausal Complementation. Cham: Springer.

Core Thesis #

All clause-embedding predicates uniformly select for question-type denotations (sets of propositions, type ⟨⟨s,t⟩,t⟩). Proposition-taking is derived via the type-shift p → {p}. This question-oriented theory provides:

  1. A unified semantics for declarative and interrogative complementation
  2. Semantic (not syntactic) explanations for selectional restrictions
  3. Anti-rogativity of believe/hope via triviality (L-analyticity)
  4. Responsiveness of know/be happy via veridicality breaking triviality
  5. A constraint hierarchy: C-distributivity ⟹ {Strawson C-dist, P-to-Q Entailment}

What This Study File Verifies #

Linglib Integration #

This file imports and connects existing infrastructure:

Constraint Hierarchy #

@cite{uegaki-2022} establishes a hierarchy of constraints on clause-embedding predicates. The key result: P-to-Q Entailment is empirically superior to C-distributivity and Strawson C-distributivity.

The Hierarchy (independent weakenings) #

C-distributivity ⟹ Strawson C-distributivity (with trivial presup) C-distributivity ⟹ P-to-Q Entailment (Strawson C-dist and P-to-Q Entailment are independent — neither implies the other)

C-distributivity implies P-to-Q Entailment (re-exported from CDistributivity.lean).

A C-distributive PreferentialPredicate satisfies P-to-Q Entailment (for any fixed comparison class C).

This closes the end-to-end chain for Class 2 and Class 3 NVPs: attitudeBuilder → isCDistributive → IsPtoQEntailing. Class 1 NVPs (non-C-distributive) require individual P-to-Q proofs (e.g., wonder_satisfies_ptoq, daroo_satisfies_ptoq, care_satisfies_ptoq).

NVP Classification (@cite{uegaki-2022} Ch 6) #

ClassC-distValenceTakes Q?Examples
1anyworry, qidai
2negativefear, dread
3positivehope, wish, expect

Veridical preferentials (be happy, be surprised) break triviality despite being C-distributive + positive, because truth adds a world-dependent constraint.

The Triviality Derivation (@cite{uegaki-2022} Ch 6 §6.5) #

Anti-rogativity of positive NVPs is DERIVED, not stipulated:

1. hope has degree-comparison semantics     (definition)
2. → hope is C-distributive                 (degreeComparisonPredicate_isCDistributive)
3. hope has positive valence                 (definition)
4. → hope has TSP                           (positive_hasTSP)
5. hope is non-veridical                     (definition)
6. 2 + 4 + 5 → assertion = presupposition  (hope_triviality_identity)
7. → L-analytic → unacceptable              (@cite{gajewski-2002})

The key step is (6): when the comparison class C equals the question Q, the assertion (∃p ∈ Q. μ(x,p) > θ(Q)) is identical to the presupposition (TSP: ∃p ∈ Q. μ(x,p) > θ(Q)). The meaning is L-analytic — its truth value is completely determined by whether the presupposition holds, with no informative content remaining.

Veridical predicates (be happy) break at step 6: truth requirement means assertion is NOT entailed by TSP (veridical_breaks_triviality).

Triviality identity (re-exported from Preferential.lean): When C = Q, hope's assertion IS its presupposition.

This is the heart of the anti-rogativity derivation. The assertion ∃p ∈ Q. μ(x,p) > θ(Q) is literally the same formula as TSP ∃p ∈ Q. μ(x,p) > θ(Q). The meaning is L-analytic.

hope-whether via Highlighting #

@cite{white-2021} presents attested "hope whether" constructions as apparent counterexamples to the anti-rogativity of positive NVPs.

@cite{uegaki-2022}'s solution: hope is sensitive to the highlighted value of the complement (@cite{pruitt-roelofsen-2011}). Polar interrogatives highlight only the overtly-realized proposition (singleton), so "hope whether p" ≈ "hope that p" — not trivial.

Constituent interrogatives highlight ALL alternatives, preserving triviality.

Cross-Linguistic Predicates (@cite{uegaki-2022} Chs 4, 5, 8) #

The book's key empirical argument for P-to-Q Entailment over C-distributivity comes from cross-linguistic data: predicates that violate C-distributivity but satisfy P-to-Q Entailment.

Table 8.2: Four Constraints × Seven Predicates #

*shknow*knopinioncaremõtlemadaroowondermagtaka
Veridical UniformityNA
C-distributivity
Strawson C-distributivity
P-to-Q Entailment

P-to-Q Entailment correctly rules out only magtaka among attested predicates, while ruling out all three fictitious predicates (*shknow, *knopinion). C-distributivity over-rules-out care, mõtlema, daroo, wonder.

Cross-linguistic embedding datum from @cite{uegaki-2022} Table 8.2.

Fields satisfiesVU, satisfiesCDist, satisfiesStrawsonCDist, satisfiesPtoQ encode whether the predicate satisfies each constraint (not whether the constraint makes a correct prediction). This aligns with the theory-level definitions (IsPtoQEntailing, IsCDistributive, etc.) — e.g., shknow_violates_ptoq proves *shknow does NOT satisfy P-to-Q Entailment, so satisfiesPtoQ = false for *shknow.

The paper's ✓/✗ table encodes "correct prediction" — a different concept. For attested predicates, correct prediction = satisfies. For fictitious predicates, correct prediction = violates (the constraint correctly rules them out). Use correctPrediction to derive the paper's table.

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

      Does a constraint make a correct prediction about this predicate?

      For attested predicates: correct iff the predicate satisfies the constraint (the constraint allows an actually-existing predicate). For fictitious predicates: correct iff the predicate violates the constraint (the constraint rules out a non-existent predicate).

      Equations
      Instances For

        Full Table 8.2 data from @cite{uegaki-2022}, including fictitious predicates.

        Values use "satisfies" semantics:

        • *shknow, *knopinion: violate ALL four constraints (all false)
        • wonder: VU is NA in the paper (VU applies to responsive predicates; wonder is rogative). Encoded as satisfiesVU = true because IsVeridicallyUniform is vacuously satisfied: both IsVeridicalDecl and IsVeridicalInterrog hold vacuously when ⟦wonder⟧({p}↓)(x) is always false, making the biconditional trivially true.
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For

          C-distributivity implies P-to-Q Entailment across all table predicates.

          P-to-Q Entailment is strictly weaker: some attested predicates satisfy P-to-Q but not C-distributivity (care, mõtlema, daroo, wonder).

          C-distributivity implies Strawson C-distributivity across all table predicates.

          All unattested predicates violate P-to-Q Entailment — the constraint correctly rules out every fictitious predicate in the table.

          P-to-Q Entailment allows more attested predicates than C-distributivity.

          P-to-Q makes correct predictions for all predicates except magtaka.

          Connecting to Left Periphery Theory #

          @cite{uegaki-2022}'s question-oriented theory is complementary to the Left Periphery analysis in LeftPeriphery.lean. The question-oriented theory explains why predicates have their selectional restrictions (semantic triviality), while the Left Periphery theory explains how the syntactic structure of embedded clauses varies (CP vs PerspP vs SAP layers).

          The connection: predicates that are anti-rogative under the question-oriented theory (Class 3 NVPs, neg-raising predicates) correspond to uninterrogative in the SelectionClass taxonomy. Predicates that are responsive correspond to responsive. Rogative predicates correspond to rogativePerspP/rogativeSAP.

          NVPClass↔SelectionClass bridge (approximate): the question-oriented theory's NVP classification partially aligns with the Left Periphery's SelectionClass.

          • Class 3 (anti-rogative) → uninterrogative: correct (both predict no interrog)
          • Class 2 (C-dist + negative) → responsive: correct (fear, dread take both)
          • Class 1 (non-C-dist) → responsive: approximate — correct for worry and care (which take both decl + interrog), but incorrect for wonder and daroo, which are better classified as .rogativePerspP. The issue is that .responsive in the Left Periphery entails factivity/knowledge, which wonder and daroo lack.

          This mapping is lossy because NVPClass captures C-dist × valence while SelectionClass captures syntactic selection of left-peripheral layers — orthogonal dimensions that don't reduce to each other.

          Equations
          Instances For

            SelectionClass ↔ P-to-Q Bridge #

            The Left Periphery's SelectionClass taxonomy (syntactic selection) and the question-oriented theory's P-to-Q Entailment (semantic constraint) are consistent: every SelectionClass that admits interrogative embedding is associated with a semantic structure that satisfies P-to-Q.

            Theoretical Gap: daroo #

            The nvpToSelectionClass bridge maps Class 1 (non-C-distributive, takes questions) to .responsive. But .responsive in the Left Periphery sense requires factivity and knowledge entailment — properties daroo lacks. Daroo's behavior is better captured as .rogativePerspP (perspective-layer selection) with an extended semantics that drops the ignorance component. This suggests the SelectionClass taxonomy needs a finer-grained distinction within PerspP-selecting predicates: those with ignorance (wonder) vs without (daroo).

            P-to-Q Entailment is satisfied by all attested interrogative-embedding predicates in Table 8.2 except magtaka (Tagalog), which is the sole counterexample — motivating the search for a still-weaker constraint.

            The SelectionClass taxonomy is consistent with P-to-Q: for each interrogative-embedding SelectionClass, at least one attested predicate in Table 8.2 demonstrates P-to-Q satisfaction.

            • rogativePerspP: wonder satisfies P-to-Q (vacuous singleton)
            • responsive: care, daroo, mõtlema all satisfy P-to-Q

            World-Independence (@cite{uegaki-2022} Ch 6 §6.5.4) #

            The triviality argument relies on non-veridical predicates having world-independent semantics: ⟦x hopes p⟧(w, C) = μ(x,p) > θ(C) contains no world variable. When assertion = presupposition (hope_triviality_identity), this identity holds at ALL worlds — making the meaning L-analytic per @cite{gajewski-2002}.

            L-analyticity: An LF constituent is L-analytic iff its logical skeleton receives the same truth-value under every variable assignment where the denotation is defined. For non-veridical preferentials + interrogative complement, the assertion is always true when TSP is met, regardless of world — L-analytic → ungrammatical.

            Fictitious Predicates (@cite{uegaki-2022} Table 8.2) #

            All three fictitious predicates from the book are formalized and proved to violate P-to-Q Entailment:

            Fear and Negative Preferentials #

            @cite{uegaki-2022} §6.6.2 proposes that fear's assertion direction is flipped relative to hope: where hope asserts Pref(x, p) > θ(C), fear asserts Pref(x, p) < θ(C). The presupposition (TSP) uses the same direction (>) for both.

            This means fear doesn't yield triviality even WITH TSP: the presupposition says "something is preferred above threshold" while the assertion says "something is preferred below threshold" — these are independent conditions, not identical.

            Our formalization simplifies this by treating negative valence as lacking TSP entirely (hasTSP .negative = false). This captures the correct result (fear takes questions) but via a different mechanism. The book's more nuanced analysis — same TSP, flipped assertion — is left as future refinement.