Documentation

Linglib.Phenomena.Quantification.Studies.DavidsonGagne2022

@cite{davidson-gagne-2022} #

Davidson, K. & Gagné, D. (2022). "More is up" for domain restriction in ASL. Semantics & Pragmatics 15, Article 1: 1–52.

Core Argument #

In ASL, vertical height in signing space overtly marks the size of the quantifier domain. When FS(ALL) is signed at neutral height, it quantifies over the contextually supplied domain (e.g., friends who watched the movie). When signed higher, it quantifies over a wider domain (e.g., everyone in the world). This is not a gestural expression of surprise or emphasis — it is a grammatical mechanism that enters through pronominal elements (IX-arc) and composes with quantifiers via a partitive-like restrictor structure.

Key Claims Formalized #

  1. Height = domain size (§5.2): Vertical ordering on loci corresponds to mereological part-hood on domains. Formalized via HeightDDRP as a NestedRestriction, giving all monotonicity theorems free.

  2. Monotonicity contrast (§4): ALL and NONE are ↓MON in the restrictor (smaller domain → easier to satisfy), SOME is ↑MON (larger domain → easier to satisfy). Height affects domain size uniformly, but truth conditions react differently depending on the quantifier's monotonicity.

  3. Height ≠ intensification (§4, after ex. 36): If height were merely emphatic strengthening, then higher SOME should be harder to satisfy (stronger). But higher SOME is easier to satisfy (wider domain for the existential witness). This rules out the emphasis analysis.

  4. Only pronominal structures (§3.3): Bare nouns cannot be height-marked because they lack a pronominal argument position. Height requires IX-arc (overt or incorporated).

  5. Structural parallel to cognitive DDRPs (§5): ASL height (overt, modality-specific) and @cite{ritchie-schiller-2024}'s spatial heuristics (covert, modality-general) are both instances of NestedRestriction — the same abstract nesting structure produces the same formal predictions.

Compositional Grounding #

Truth conditions derive from every_restricted / some_restricted / no_restricted (DomainRestriction.lean), which wrap every_sem / some_sem / no_sem (Quantifier.lean) with a height-indexed domain predicate. Nesting theorems derive from DDRP.every_nesting / DDRP.some_nesting / DDRP.no_nesting, connecting height ordering to restrictor monotonicity.

Not formalized here #

The running example from @cite{davidson-gagne-2022} §1:

Context: "Last night I watched a movie with my friends about vampires.
Afterwards I went to bed and I dreamt that..."

FS(ALL)-neutral BECOME VAMPIRE → 'All of my friends became vampires'
FS(ALL)-high BECOME VAMPIRE → 'All of the people in the world became vampires'

Entities in the vampire scenario.

  • f1f4: friends who watched the movie (contextual domain)
  • p5p6: people in the wider world (outside contextual domain)
Instances For
    @[implicit_reducible]
    Equations
    def DavidsonGagne2022.instReprEntity.repr :
    EntityStd.Format
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      @[reducible, inline]
      Equations
      Instances For
        @[implicit_reducible]
        Equations
        • One or more equations did not get rendered due to their size.
        @[reducible, inline]

        DDRP for the vampire scenario, derived from the height assignment. neutral-domain = {f1, f2, f3, f4}, high-domain = everyone.

        Equations
        Instances For

          World states: who became vampires in the dream.

          Instances For
            @[implicit_reducible]
            Equations
            def DavidsonGagne2022.instReprWorld.repr :
            WorldStd.Format
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              @[reducible, inline]

              Truth of "all [of them] became vampires" under a height-indexed domain. Uses vampireZone e ≤ h directly (definitionally equal to e ∈ vampireDDRP.region h) for Decidable transparency.

              Equations
              Instances For

                D&G's core observation: At neutral height, "all became vampires" is true when only friends became vampires (domain = friends). At high height, it is false — the domain includes everyone, and not everyone became a vampire. Same sentence, different height, different truth value.

                @[reducible, inline]

                Truth of "some [of them] became vampires" under a height-indexed domain.

                Equations
                Instances For

                  @cite{davidson-gagne-2022} §4 (ex. 33–35) discuss NONEsym, showing the same height-domain pattern. NONEsym-neutral = "none of my immediate family", NONEsym-high = "none of my entire family including ancestors, distant relations, etc." NONE is ↓MON in the restrictor, like ALL.

                  @[reducible, inline]

                  Truth of "none [of them] became vampires" under a height-indexed domain.

                  Equations
                  Instances For

                    ⟦every⟧ nesting: truth under a higher height entails truth under any lower height. Smaller domain → fewer entities to check → universal easier. Direct application of DDRP.every_nesting.

                    ⟦some⟧ nesting: truth under a lower height entails truth under any higher height. Larger domain → more witnesses → existential easier. Direct application of DDRP.some_nesting.

                    ⟦no⟧ nesting: truth under a higher height entails truth under any lower height. Same direction as ⟦every⟧ (both ↓MON in the restrictor). Direct application of DDRP.no_nesting.

                    Monotonicity contrast: ALL and SOME react oppositely to height. In the friendsOnly world, ALL is true only under neutral height (↓MON: smaller domain helps), while SOME is true under both heights (↑MON: larger domain never hurts).

                    @cite{davidson-gagne-2022} §4 (after ex. 36) argue that height is not merely emphatic strengthening. The critical evidence comes from existential quantifiers: if height strengthened the utterance, higher SOME should impose stronger truth conditions (harder to satisfy). But higher SOME is easier to satisfy — a wider domain provides more potential witnesses.

                    For universal and negative quantifiers, both strengthening and domain
                    widening make the same prediction (wider domain → stronger statement).
                    But for existentials, they diverge: strengthening predicts harder, domain
                    widening predicts easier. The data favor domain widening. 
                    

                    Height widens existential domain: SOME-high is at least as easy to satisfy as SOME-neutral. If height were intensification, the opposite would hold for existentials. This is the formal refutation.

                    @cite{davidson-gagne-2022} §3.3 (ex. 27): bare nouns like DOG cannot be signed higher to convey domain widening. This is predicted by the pronominal analysis: height modifies IX-arc (a pronoun), not the noun directly. Nouns lack an unsaturated pronominal argument position.

                    Structurally: `HeightDDRP` is a parameter of the pronominal/quantifier
                    denotation, not of the noun denotation. A bare noun `dog : Entity → Bool`
                    has no slot for a `HeightDDRP` argument. Domain restriction enters only
                    via the pronominal restrictor: FS-ALL-of-[IX-arc-height].
                    
                    This is a type-level observation: the type signature of the quantifier
                    (`HeightDDRP Entity → VerticalHeight → ...`) vs. the bare noun
                    (`Entity → Bool`) enforces the restriction structurally. No theorem
                    is needed — the compiler rejects ill-typed compositions. 
                    

                    A height-restricted quantifier: takes a HeightDDRP argument. The pronominal restrictor (IX-arc) provides the domain; the quantifier quantifies over it. Bare nouns (type Entity → Bool) cannot supply the ddrp or h arguments — height is enforced by the type system.

                    Equations
                    Instances For

                      Both @cite{davidson-gagne-2022}'s height-based domain restriction and @cite{ritchie-schiller-2024}'s cognitive DDRP framework instantiate Core.NestedRestriction: a monotone family of predicates indexed by an ordered scale. The scale source differs — physical signing height (ASL, overt) vs. cognitive spatial proximity (spoken language, covert) — but the formal structure is identical.

                      This means the monotonicity theorems proved once in `NestedRestriction`
                      (`forall_nesting`, `exists_nesting`) give correct predictions for BOTH
                      the cognitive heuristic theory (SpatialScale) and the ASL height theory
                      (VerticalHeight) with zero duplication. 
                      

                      Any entity in a height-DDRP region is also in the corresponding spatial-DDRP region when domains are aligned (i.e., the height zone maps into the spatial zone). This shows the two scale systems compose: an entity's signing-height visibility entails its spatial-region visibility whenever the modalities agree on the entity's position.

                      The nesting property is uniform across all NestedRestriction instances: height-DDRP nesting and spatial-DDRP nesting are both instances of the same abstract monotonicity. This is the formal content of the claim that ASL height and cognitive spatial heuristics are "the same structure."

                      @cite{davidson-gagne-2022} §6.2 reports parallel height-domain mappings in other sign languages:

                      **JSL** (ex. 54): "There is/was nobody from a deaf family in my {class /
                      school / (prefecture of) Wakayama}." Height levels @chest < @cheek <
                      @forehead correspond to class ⊂ school ⊂ prefecture.
                      
                      **Nicaraguan Sign Language** (ex. 55): TODO-mid = "all [of us here]" vs
                      TODO-high = "all [of Nicaragua]".
                      
                      These data suggest that the height-domain mapping may be a universal
                      tendency in sign languages, possibly grounded in the MORE IS UP
                      conceptual metaphor. Full formalization awaits dedicated JSL and
                      LSN fragments. 
                      

                      @cite{davidson-gagne-2022} §5.3 derives quantifier truth conditions through the partitive type function ⟦of⟧ = λxλy. y ≤ x:

                      1. IX-arc-h denotes discourse referent z with height presuppositions
                      2. ⟦of⟧ z = `Set.Iic z` = {y | y ≤ z} — the parts of z
                      3. FS-ALL(`Set.Iic z`)(Q) = ∀x ∈ Set.Iic z, Q x
                      
                      The theorems `partitive_neutral_subset` and `partitive_high_superset`
                      (in `Height.lean`) show that the presuppositions encode exactly
                      `Set.Iic z ⊆ C` (neutral) and `C ⊆ Set.Iic z` (high).
                      
                      The DDRP region `vampireDDRP.region h` serves as the computational
                      implementation of `Set.Iic z_h` — see `partitive_ddrp_equiv`
                      for the abstract alignment theorem. The theorems below close the
                      grounding gap by showing that the stipulative truth tables
                      (`fsAllBecomeVampire`, `someBecomeVampire`) agree with the composed
                      meaning using `every_restricted` / `some_restricted` from the
                      theory layer. 
                      

                      FS-ALL composed through the theory-layer quantifier every_restricted. The DDRP region at height h serves as the domain restrictor C, matching the partitive path where Set.Iic z restricts the quantifier to parts of the discourse referent.

                      After the Bool→Prop conversion, fsAllBecomeVampire and fsAllComposed are definitionally equivalent: both express ∀ e, e ∈ region h → .... The grounding theorem below witnesses this structural alignment.

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

                        SOME composed through the theory-layer quantifier some_restricted.

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

                          Grounding theorem: fsAllBecomeVampire agrees with the compositionally-derived fsAllComposed. After the Bool→Prop conversion, both are ∀ e, e ∈ region h → becameVampire w e = true (modulo a vacuous True restrictor), so the proof is structural.

                          Grounding theorem for SOME: someBecomeVampire agrees with some_restricted from the theory layer.