@cite{ritchie-schiller-2024} — Default Domain Restriction Possibilities #
@cite{ritchie-schiller-2024} @cite{cutting-vishton-1995} @cite{baker-jara-ettinger-saxe-tenenbaum-2017} @cite{clark-1996} @cite{stalnaker-2002}
Ritchie, H. & Schiller, K. (2024). Default Domain Restriction Possibilities. Semantics & Pragmatics 17, Article 13: 1–49.
The Argument #
When a speaker says "Every bottle is empty" at a dinner party, the hearer restricts the quantifier domain to contextually relevant bottles — not all bottles in the universe (R&S §1, ex. 3). Ritchie & Schiller argue that existing accounts fail to explain why certain restrictions are defaults:
- Rational pragmatic (§2.1): RSA/Gricean reasoning doesn't explain default status
- Discourse-structural (§2.2): QUD-based accounts are too demanding
- Intentionalist (§2.3): speaker-intention accounts are too unconstrained
Their positive proposal (§3): cognitive heuristics — perceptual availability, salience, and manipulability — generate a structured set of default domain restriction possibilities (DDRPs). These are grounded in spatial cognition, where nested spatial regions provide a natural ordering on candidate restrictions.
Scenario #
We construct an illustrative scenario (not from the paper) with 4 entities at increasing spatial distances and 3 world states, then verify key formal consequences of the DDRP framework for both ⟦every⟧ (↓MON) and ⟦some⟧ (↑MON).
Compositional Grounding #
Truth conditions derive from every_restricted / some_restricted
(DomainRestriction.lean), which wrap every_sem / some_sem (Quantifier.lean)
with a domain restrictor predicate. Nesting theorems derive from
DDRP.every_nesting / DDRP.some_nesting, connecting the nested spatial
regions to restrictor monotonicity.
RSA Connection #
While R&S argue against RSA as explaining default status (§2.1), DDRPs are
compatible with RSA as the selection mechanism: the listener reasons over
candidate DDRPs (= Latent in RSAConfig) to infer which domain restriction
the speaker intended. We demonstrate this connection in §7.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Peripersonal space: entities within arm's reach (b1, b2 on the table).
Equations
- Phenomena.Quantification.Studies.RitchieSchiller2024.near Phenomena.Quantification.Studies.RitchieSchiller2024.Entity.b1 = true
- Phenomena.Quantification.Studies.RitchieSchiller2024.near Phenomena.Quantification.Studies.RitchieSchiller2024.Entity.b2 = true
- Phenomena.Quantification.Studies.RitchieSchiller2024.near x✝ = false
Instances For
Action space: entities accessible by locomotion (b1, b2, b3).
Equations
Instances For
DDRP for the bottle scenario. Near ⊆ mid ⊆ vista (= unrestricted in this indoor scene).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Equations
- One or more equations did not get rendered due to their size.
- Phenomena.Quantification.Studies.RitchieSchiller2024.emptyIn Phenomena.Quantification.Studies.RitchieSchiller2024.World.allEmpty x✝ = true
- Phenomena.Quantification.Studies.RitchieSchiller2024.emptyIn Phenomena.Quantification.Studies.RitchieSchiller2024.World.midEmpty x✝ = true
- Phenomena.Quantification.Studies.RitchieSchiller2024.emptyIn Phenomena.Quantification.Studies.RitchieSchiller2024.World.nearEmpty x✝ = false
Instances For
All entities are bottles in this scenario (trivial restrictor).
Instances For
Truth of "every bottle is empty" under a given spatial domain restriction.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Truth of "some bottle is empty" under a given spatial domain restriction.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Proximal default: In the proximal world, only peripersonal restriction makes "every bottle is empty" true. The listener must infer the speaker intended the proximal domain — no other DDRP candidate works. This illustrates R&S's claim (§3.1) that perceptual availability biases hearers toward proximal domains: when only one candidate restriction works, pragmatic selection is forced.
↓MON/↑MON contrast: ⟦every⟧ and ⟦some⟧ react oppositely to domain restriction. In the proximal world, ⟦every⟧ is true only under peripersonal restriction (↓MON: smaller domain helps), while ⟦some⟧ is true under all restrictions (↑MON: larger domain never hurts).
⟦every⟧ nesting: truth under any scale entails truth under any smaller scale.
Uses the general DDRP.every_nesting theorem parameterized by the ordering.
Transitive nesting: vista → peripersonal (2-step composition).
⟦some⟧ nesting (reversed): truth under any scale entails truth under any larger scale. The ↑MON dual of ⟦every⟧ nesting.
Transitive ⟦some⟧ nesting: peripersonal → vista (2-step composition).
Utterance type for the RSA model.
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Literal meaning under a given DDRP scale.
Equations
- One or more equations did not get rendered due to their size.
Instances For
RSAConfig instantiation with DDRPs as the latent variable.
The listener reasons over which spatial scale the speaker intended when
uttering a quantified sentence. Latent = SpatialScale: L1 marginalizes
over candidate domain restrictions to infer the speaker's intended domain.
While R&S argue against RSA as explaining default status (§2.1), their DDRPs are fully compatible with RSA as the selection mechanism once the candidate set is fixed by cognitive heuristics. This models the pragmatic step: given the DDRP candidates, which one did the speaker intend?
Equations
- One or more equations did not get rendered due to their size.
Instances For
L0 correctly reflects literal semantics: when ⟦every⟧ is true under a given scale, L0's score is positive.
L0 correctly reflects literal semantics: when ⟦every⟧ is false under a given scale, L0's score is zero.
Connects DDRPs to @cite{baker-jara-ettinger-saxe-tenenbaum-2017}'s BToM architecture and @cite{stalnaker-2002}'s common ground.
Perception-generated DDRPs: A spatial scene induces a DDRP via
sceneToDDRP. Monotonicity follows from transitivity of ≤ onSpatialScale.BToM instantiation:
RSAConfig.toBToMgives a BToM model; the bridge theoremL1_eq_btom_worldMarginalproves L1 IS BToM world-marginal.Common-ground constraint: When the scene is common ground, speaker and hearer derive the same DDRP (@cite{clark-1996} on joint attention).
Perfect-perception collapse: Under perfect perception, the perception-generated DDRP equals the hand-written one.
Imperfect perception: Different perceptual access → different DDRPs, motivating R&S's requirement of perceptual co-presence.
A spatial scene: each entity occupies a spatial zone.
Equations
Instances For
Entities perceivable at a given scale threshold: those whose zone ≤ threshold.
Equations
- Phenomena.Quantification.Studies.RitchieSchiller2024.perceivable scene threshold e = decide (scene e ≤ threshold)
Instances For
A spatial scene induces a DDRP. Region s contains entities in zone ≤ s.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The dinner-party scene: b1,b2 peripersonal, b3 action, b4 vista.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The RSA-BToM bridge applies to the domain restriction RSA config.
The actual scene at each world.
Equations
Instances For
A DDRP is grounded in common ground when the spatial scene is common knowledge among discourse participants.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The dinner scene is common ground.
Under perfect perception, the perception-generated DDRP yields the same truth conditions as the hand-written one.
SpatialScale is a mental state (speaker's private choice).
Equations
Instances For
The spatial scene is shared (perceptual co-presence).
Equations
Instances For
Full latent classification for the domain restriction model.
Equations
- One or more equations did not get rendered due to their size.
Instances For
An alternative scene where b3 is behind a partition.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Different spatial scenes yield different DDRPs.
Without perceptual co-presence, domain-restricted quantifiers can receive different truth values.