Exhaustivity Presuppositions for Questions @cite{xiang-2022} @cite{dayal-1996} #
Formalizes Dayal's Exhaustivity Presupposition (EP, definition 90) and Xiang's Relativized Exhaustivity (RelExh, definition 91) from:
@cite{xiang-2022}. Relativized Exhaustivity: mention-some and uniqueness. Natural Language Semantics 30:311–362.
Key Definitions #
dayalEP: @cite{dayal-1996}'s Exhaustivity Presupposition — the question presupposes that there exists a strongest true answer (one whose proposition entails all other true answers' propositions).relExh: Xiang's Relativized Exhaustivity — EP is evaluated relative to singleton modal bases. For each accessible world v, restrict the modal base to {v} and check EP there. This weakens EP just enough to license mention-some for ability-can questions while preserving mention-all for non-modal and other modal questions.foQDen: First-order question denotation (◇φ_x interpretation) — the building block for can-questions where the wh-trace scopes below the modal.
Design #
All definitions are polymorphic in W (worlds) and P (individuals/answers),
taking explicit List W universe parameters for decidable computation via
native_decide. No imports beyond basic Lean — this is pure theory.
Propositional Entailment #
Propositional entailment as Bool: p ⊆ q over a finite universe.
propEntails p q worlds = true iff every world where p holds also has q.
Equations
- Theories.Semantics.Questions.Exhaustivity.propEntails p q worlds = worlds.all fun (v : W) => !p v || q v
Instances For
True Answers #
The answers whose denotation is true at world w under modal base mb.
Equations
- Theories.Semantics.Questions.Exhaustivity.trueAnswers qden mb answers w = List.filter (fun (α : P) => qden mb α w) answers
Instances For
First-Order Question Denotation #
First-order question denotation: ◇φ_x interpretation.
foQDen pred mb α w = true iff there exists an accessible world v ∈ mb(w)
where pred(v, α) holds. This is the denotation for can-questions where the
wh-trace takes scope below the modal:
⟦who can VP⟧(mb)(α)(w) = ∃v ∈ mb(w). VP(v)(α)
Equations
- Theories.Semantics.Questions.Exhaustivity.foQDen pred mb α w = (mb w).any fun (v : W) => pred v α
Instances For
Dayal's Exhaustivity Presupposition #
Dayal's Exhaustivity Presupposition (@cite{dayal-1996}; @cite{xiang-2022}, definition 90).
EP holds at w iff there exists a true answer α whose proposition entails the
proposition of every other true answer β:
∃α ∈ True(w). ∀β ∈ True(w). ⟦α⟧ ⊆ ⟦β⟧
When EP holds, α is the strongest true answer — the exhaustive answer to the question. When EP fails, no single answer subsumes all others, so the question has no well-defined exhaustive reading.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Relativized Exhaustivity #
Relativized Exhaustivity (@cite{xiang-2022}, definition 91).
RelExh holds at w iff for every singleton modal base {v} where v ∈ mb(w),
if {v} makes some answer relevant (both true under {v} and true under the full
mb), then EP holds for {v}:
∀v ∈ mb(w). [∃α. qden({v})(α)(w) ∧ qden(mb)(α)(w)] → EP({v})(w)
The key insight: under a singleton modal base {v}, each individual α either determinately can or determinately cannot VP at v. So for each v, there IS a strongest true answer (trivially — the answers don't overlap when the modal base is a singleton). This means RelExh passes for ability-can questions even when the full EP fails due to overlapping answer propositions.
Equations
- One or more equations did not get rendered due to their size.
Instances For
@cite{fox-2018}: Partition by Exhaustification @cite{fox-2018} #
@cite{fox-2018} "Partition by Exhaustification" derives Dayal's EP from the exhaustification operator Exh. The key insight: a question partitions the logical space iff every world has exactly one exhaustified true answer.
The definitions below are Bool-valued analogues of the Prop-valued MC-set/IE
machinery in Exhaustification.Basic, specialized to question cells
for decidable computation via native_decide.
cellMCSetsmirrorsisMCSetfrom ExhaustificationcellIEmirrorsIEfrom ExhaustificationfoxExhmirrorsexhIEfrom ExhaustificationfoxAnsimplements @cite{fox-2018}, definition 35 (answer operator)foxPartitionimplements Schwarzschild's partition test ((38))
All sublists (power set) of a list, preserving order. Used to enumerate candidate MC-sets of cell indices.
Equations
Instances For
Safe cell accessor: returns the cell at index i, or (λ _ => false) for OOB.
Equations
- Theories.Semantics.Questions.Exhaustivity.getCell cells i = cells.getD i fun (x : W) => false
Instances For
Negation consistency: can cells at excluded indices be simultaneously
negated while cell pIdx stays true?
negConsistent cells pIdx excluded worlds = true iff
∃w ∈ worlds. cellspIdx ∧ ∀j ∈ excluded. ¬cellsj
This is the Bool analogue of SetConsistent in Exhaustification,
specialized to the pattern φ ∧ ⋀{¬a : a ∈ excluded}.
Equations
- One or more equations did not get rendered due to their size.
Instances For
MC-sets (maximal consistent sets) of cell indices for cell pIdx.
A subset S of alternative indices (excluding pIdx) is an MC-set iff:
- Negating all cells in
Sis consistent withpIdxbeing true - No proper superset of
Sis also consistent
Bool analogue of isMCSet in Exhaustification.Basic.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Innocently excludable alternatives for cell pIdx: the indices
that appear in every MC-set (intersection of all MC-sets).
Bool analogue of IE in Exhaustification.Basic:
IE_(ALT,φ) = {ψ : ψ belongs to every MC_(ALT,φ)-set}
Equations
- One or more equations did not get rendered due to their size.
Instances For
Fox's exhaustified cell: Exh(Q)(p)(w) = cellspIdx ∧ ∀j ∈ IE. ¬cellsj.
The exhaustified meaning of answer p negates all innocently excludable
alternatives. Bool analogue of exhIE in Exhaustification.Basic.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Non-vacuity: Exh(Q)(p) is satisfiable (true at some world). @cite{fox-2018} requires non-vacuity for QPM.
Equations
- Theories.Semantics.Questions.Exhaustivity.foxNV cells pIdx worlds = worlds.any (Theories.Semantics.Questions.Exhaustivity.foxExh cells pIdx worlds)
Instances For
Count of Exh-true cells at world w: the number of cells whose
exhaustified meaning holds at w. This is NOT Fox's |Ans| (Definition 35)
but a simpler diagnostic: how many exclusive readings are true at w.
When = 0, no cell exclusively holds (overlap); when = 1, one cell
dominates; when > 1, multiple exclusive readings coexist.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Pointwise non-vacuity: every cell true at w has a satisfiable
exhaustified version. This is a necessary condition for Fox's QPM
(Definition 34) but weaker — it checks NV for true cells at a single
world, while Fox's QPM requires both CI and NV globally over the
partition induced by Q.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Fox's answer operator (Definition 35). At world w, finds the unique
cell-identifier (the unique j where foxExh(cells, j, worlds)(w) = true)
and returns the count of Q-members that are true at w AND whose
denotation entails the cell-identifier proposition cells[j].
Fox's (37): Ans(Q)(w) = {q∈Q : w∈q ∧ q ⊆ (ιp∈Q)[Exh(Q,p,w)=1]} The ι picks the unique proposition p whose exhaustification is true at w; the entailment check is q ⊆ p (q entails the proposition p, not Exh(p)).
Returns 0 if no unique cell-identifier exists (zero or multiple Exh-true
cells at w). When foxAns > 1, Fox predicts mention-some (the
cell-identifier is weaker than individual true answers). When = 1,
mention-all (the cell-identifier is the strongest true answer).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Schwarzschild's partition test (@cite{fox-2018}, (38)): do the exhaustified
cells {Exh(Q)(p) : p ∈ Q} partition the world set? Holds iff every world
has exactly one Exh-true cell. When this holds, Fox's QPM (Definition 34)
is satisfied and foxAns is well-defined at every world.
Equations
- One or more equations did not get rendered due to their size.