Uegaki & Sudo 2019: The hope-wh puzzle #
[US19] derive the anti-rogativity of non-veridical
preferential predicates (hope, wish) from triviality. With the
degree semantics ⟦x V p⟧ = μ(x,p) > θ(C) ([Vil08]), the
Threshold Significance Presupposition, and clausal distributivity,
combining the predicate with a question whose answers exhaust the
comparison class yields an assertion identical to its presupposition
(hope_question_iff_significance) — an L-analytic meaning in the
sense of [Gaj02], hence ungrammaticality.
Veridical preferentials (be surprised, be happy, be glad,
like, hate) escape and take questions: the truth requirement on
the complement makes the assertion world-dependent, so threshold
significance no longer settles it. veridicalQuestion is the
world-sensitive semantics — still clausally distributive
(veridical_isDistributive) — and veridicality_breaks_triviality
exhibits a model where the presupposition holds, the non-veridical
assertion is (trivially) true, and the veridical assertion is false
because the true answer is not the preferred one.
Triviality for non-veridical preferentials #
With the question's answers drawn from the comparison class, the hope-question assertion entails threshold significance.
Conversely, threshold significance entails the assertion when the comparison class is contained in the question.
When the question's answers are exactly the comparison class, the assertion of hope + question is its presupposition: the meaning is L-analytic ([Gaj02]) — true whenever defined — which is the triviality that makes hope anti-rogative.
Veridical preferentials #
Veridical propositional semantics: ⟦x is happy that p⟧(w, C) requires the complement to be true at the evaluation world.
Equations
- UegakiSudo2019.veridicalProp μ θ C x p w = (w ∈ p ∧ μ x p > θ C)
Instances For
Veridical question semantics: some true answer clears the threshold.
Equations
- UegakiSudo2019.veridicalQuestion μ θ C x Q w = ∃ p ∈ Q, w ∈ p ∧ μ x p > θ C
Instances For
Veridical preferentials are clausally distributive — it is veridicality, not a failure of distributivity, that lets them take questions.
Veridicality breaks the triviality: a model where threshold significance holds and the non-veridical assertion is therefore true, but the veridical assertion is false — the true answer is not the preferred one. Two worlds, the polar question over them, evaluated at the dispreferred world.