Documentation

Studies.QingEtAl2025

Qing et al. (2025): When Can Non-Veridical Preferential Attitude Predicates Take Questions? #

Qing, Özyıldız, Roelofsen, Romero and Uegaki classify non-veridical preferential predicates by clausal distributivity and evaluative valence. Uegaki and Sudo derive the anti-rogativity of hope from two assumptions, that the predicate is clausally distributive and that it presupposes Threshold Significance. Neither holds in general: worry and Mandarin qidai relate the agent to the question rather than to an answer, and negative predicates such as fear lack the presupposition. Composition with a question is therefore trivial, and the predicate anti-rogative, exactly for the distributive positive class. Apparent uses of hope with a polar question are non-canonical, and the asymmetry between hoping for the radical and for its negation follows from the goals of a hopeful wondering.

Main statements #

Implementation notes #

Triviality quantifies over the degree function and threshold, the non-logical constants of Gajewski's analyticity, with the question as its own comparison class, and the concrete models measure preference in the reals. A relational predicate is a parameter relating the agent to the question, since the paper characterizes the class only as a relation to the question not reducible to one to any answer. The class of a predicate is read off the valence and distributivity its lexical entry records. The paper contrasts positive with negative predicates; a neutral one is taken to lack threshold significance, as a negative one does. The pragmatic derivation compares coming to believe the radical or not by Kratzer's ordering with the event's goals as ordering source; the event summation of the adjunction analysis is not formalized. The Spanish predicates without a Fragment entry, temer and preocupar, are classified here. Rows carry the paper's truth values where it gives them and its felicity judgments otherwise; the attested hope whether examples are recorded as marginal, and temer with a polar question and haipa with a constituent question are recorded but not predicted.

References #

The two escapes from triviality (§3) #

The concrete models measure preference in the reals.

theorem QingEtAl2025.relational_not_isDistributive {W : Type u_1} {E : Type u_2} (R : E → Set (Set W) → W → Prop) {x : E} {Q : Set (Set W)} {w : W} (hR : R x Q w) (h : ∀ p ∈ Q, ¬R x {p} w) :

A predicate that relates an agent to a question without relating her to any of its answers, anxious uncertainty for worry (§3.1.2) or anticipation of resolution for Mandarin qidai (§3.1.1), is not clausally distributive. This is the diagnostic of (23) to (25).

def QingEtAl2025.Trivial {W : Type u_1} {E : Type u_2} (V : (E → W → Set W → ℝ) → (Set (Set W) → ℝ) → Set (Set W) → E → Set (Set W) → W → Prop) (π : (E → W → Set W → ℝ) → (Set (Set W) → ℝ) → E → Set (Set W) → W → Prop) :

Composition of a predicate family with a question is trivial under a presupposition when in every model the presupposition entails the assertion, the question serving as its own comparison class as in Uegaki and Sudo's account. The composition is then true whenever defined, which is Gajewski's analyticity with the degree function and threshold as the non-logical constants.

Equations
  • QingEtAl2025.Trivial V π = ∀ (μ : E → W → Set W → ℝ) (θ : Set (Set W) → ℝ) (x : E) (Q : Set (Set W)) (w : W), π μ θ x Q w → V μ θ Q x Q w
Instances For
    def QingEtAl2025.presupposition {W : Type u_1} {E : Type u_2} :
    Degree.EvaluativeValence → (E → W → Set W → ℝ) → (Set (Set W) → ℝ) → E → Set (Set W) → W → Prop

    A positive preferential predicate presupposes threshold significance of a question. A negative one triggers no such presupposition (§3.2) and presupposes only that the question has an answer in the comparison class.

    Equations
    Instances For

      A distributive positive predicate composed with a question is trivial, as Uegaki and Sudo show.

      Without threshold significance, x fears Q says only that some answer is feared, which a model falsifies.

      theorem QingEtAl2025.not_trivial_relational {W : Type u_1} {E : Type u_2} (v : Degree.EvaluativeValence) (R : E → Set (Set W) → W → Prop) (hR : ∃ (x : E) (Q : Set (Set W)) (w : W), Q.Nonempty ∧ ¬R x Q w) :
      ¬Trivial (fun (x : E → W → Set W → ℝ) (x_1 : Set (Set W) → ℝ) (x_2 : Set (Set W)) => R) (presupposition v)

      A relational predicate is not trivial even under threshold significance, since some agent does not stand in the relation to a question an answer of which clears the threshold.

      The classification (Table 2) #

      The three classes of non-veridical preferential predicates.

      • nonDistributive : PredicateClass

        Class 1 consists of the predicates that are not clausally distributive, of either valence.

      • distributiveNegative : PredicateClass

        Class 2 consists of the distributive negative predicates.

      • distributivePositive : PredicateClass

        Class 3 consists of the distributive positive predicates, the class Uegaki and Sudo describe.

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

          The class of a predicate, read off the valence and distributivity its lexical entry records. A distributive predicate of neutral valence is outside the classification.

          Equations
          Instances For
            def QingEtAl2025.semantics {W : Type u_1} {E : Type u_2} (R : E → Set (Set W) → W → Prop) :
            Bool → (E → W → Set W → ℝ) → (Set (Set W) → ℝ) → Set (Set W) → E → Set (Set W) → W → Prop

            The interrogative use of a preferential predicate is a degree comparison when it is distributive and a relation R to the question otherwise.

            Equations
            Instances For
              theorem QingEtAl2025.trivial_iff_class {W : Type u_1} {E : Type u_2} [Inhabited E] [Inhabited W] (v : Degree.EvaluativeValence) (d : Bool) (R : E → Set (Set W) → W → Prop) (hR : ∃ (x : E) (Q : Set (Set W)) (w : W), Q.Nonempty ∧ ¬R x Q w) :

              Canonical composition with a question is trivial, hence anti-rogative, exactly for the distributive positive class (Table 2).

              The paper's judgments #

              The class of a row's predicate.

              Equations
              Instances For

                The valence of a row's predicate, or of its manner adverb or veridical preferential where the paper's argument turns on valence alone.

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

                  A row holds in its context when it is judged true, where the paper gives a truth value, and when it is not infelicitous otherwise.

                  Equations
                  Instances For
                    @[instance_reducible]
                    instance QingEtAl2025.instDecidableHolds (r : Datum) :
                    Decidable (Holds r)
                    Equations
                    • One or more equations did not get rendered due to their size.

                    An interrogative complement composed canonically, as the predicate's argument.

                    Equations
                    Instances For
                      @[instance_reducible]
                      instance QingEtAl2025.instDecidableArgument (r : Datum) :
                      Decidable (Argument r)
                      Equations

                      Every acceptable interrogative complement is under a class 1 or class 2 predicate.

                      Under canonical composition a class 3 predicate takes no question: the paper marks every such example, hope whether and its counterparts in the four other languages.

                      theorem QingEtAl2025.tsp_rows (r : Datum) :
                      r ∈ Examples.all → r.feature? "polarity" = some "negated" → ∀ v ∈ valence? r, r.judgment = Judgment.acceptable ↔ v = Degree.EvaluativeValence.negative

                      A negated positive preferential is infelicitous where the agent prefers no answer, and a negated negative one is felicitous, so the negative predicates lack threshold significance. The rows are (38) to (41), (44), (49), and (52).

                      theorem QingEtAl2025.nondistributive_rows (r : Datum) :
                      r ∈ Examples.all → r.feature? "truth" = some "false" → r.feature? "clause" = some "declarative" → class? r = some PredicateClass.nonDistributive ∧ ∃ r' ∈ Examples.all, r'.context = r.context ∧ r'.feature? "predicate" = r.feature? "predicate" ∧ r'.feature? "clause" = some "constituent" ∧ r'.feature? "truth" = some "true"

                      Every declarative the paper judges false is under a class 1 predicate and sits beside a constituent question of the same predicate judged true in the same context. This is the non-distributivity diagnostic of (23) to (36).

                      theorem QingEtAl2025.diye_rows (r : Datum) :
                      r ∈ Examples.all → r.feature? "embedding" = some "diye" → r.judgment = Judgment.acceptable

                      Turkish diye clauses combine with um- "hope", kork- "fear", and an intransitive verb alike, in (71), (73), (75), and (76), as adjoined reports of wondering rather than arguments.

                      turkishVerb? s is the Fragment verb that the predicate name s of a Turkish row names.

                      Equations
                      Instances For

                        The paper's glosses segment its Turkish finite verbs into four inflections, the imperfective, its negative, its past, and the perfective past, the last two in the first person.

                        Equations
                        • One or more equations did not get rendered due to their size.
                        Instances For
                          theorem QingEtAl2025.turkish_forms (r : Datum) :
                          r ∈ Examples.all → r.language = "nucl1301" → ∃ v ∈ (r.feature? "predicate").bind turkishVerb?, ∃ sfx ∈ turkishInflections, Turkish.Verb.Licensed (v.suffixes sfx) ∧ ∃ t ∈ r.glossedTokens, Turkish.Phonology.ofSegmented? t.1 = some (v.inflect sfx)

                          The finite verb of every Turkish row is its predicate's Fragment entry under one of the paper's inflections. The finite-verb template licenses the suffix string, and the Fragment's phonology derives its surface form, korkmuyor and umuyordum among them.

                          The verb of (73) is intransitive, so its diye clause is no argument.

                          Non-canonical composition (§4) #

                          theorem QingEtAl2025.highlighted_eq_declarative {W : Type u_1} {E : Type u_2} (μ : E → W → Set W → ℝ) (θ : Set (Set W) → ℝ) (C : Set (Set W)) (x : E) (p : Set W) (w : W) :
                          Preferential.degreeComparison μ θ C x {p} w ↔ p ∈ Preferential.preferred μ θ C x w

                          Composed with the highlighted content of whether p, the singleton of the radical, a degree-comparison predicate means its declarative (70). This first candidate analysis (§4.2) yields the interpretive asymmetry but not the inquisitive implication.

                          def QingEtAl2025.HopefulGoals {V : Type u_3} (φ Bφ : V → Prop) (G : Set (V → Prop)) :

                          The goals of a hopeful event whose agent hopes φ are φ or the agent's believing φ (89).

                          Equations
                          Instances For
                            theorem QingEtAl2025.hopes_radical {V : Type u_3} {Bp : V → Prop} {v₁ v₂ : V} (h₁ : Bp v₁) (h₂ : ¬Bp v₂) :
                            v₁ < v₂

                            When the agent hopes for the radical, with the goal of believing p, the outcome in which they come to believe p is strictly better than the one in which they do not, so the wondering meets Tabatowski's convention on asking whether p.

                            theorem QingEtAl2025.hopes_negation {V : Type u_3} {p Bnp : V → Prop} {v₁ v₂ : V} {G : Set (V → Prop)} (hG : HopefulGoals (fun (v : V) => ¬p v) Bnp G) (hp₁ : p v₁) (hn₁ : ¬Bnp v₁) :
                            ¬v₁ < v₂

                            When the agent hopes for the negation, whichever goal (89) allows, an outcome in which p holds and the agent does not believe ¬p satisfies neither ¬p nor believing ¬p, so coming to believe p is no better than not, and the wondering cannot be hopeful.

                            theorem QingEtAl2025.fears {V : Type u_3} {Bp Bnp : V → Prop} {v₁ v₂ : V} (h₁ : Bp v₁) (h₂ : ¬Bp v₂) (hn₂ : ¬Bnp v₂) :
                            v₁ < v₂

                            Fear constrains no goal (87). The purely epistemic goal of knowing whether p makes coming to believe p strictly better than remaining undecided, whichever answer is feared.

                            theorem QingEtAl2025.asymmetry_rows (r : Datum) :
                            r ∈ Examples.all → ∀ t ∈ r.feature? "target", ∀ v ∈ valence? r, (v = Degree.EvaluativeValence.positive → (Holds r ↔ t = "radical")) ∧ (v = Degree.EvaluativeValence.negative → Holds r)

                            In the polar cases a positive predicate or adverb holds exactly when the agent's attitude targets the radical, and a negative one holds either way. The rows are (45), (50), (62), (63), (75), (76), (78), and (79).