[ABB+23]: QUD manipulation of homogeneity projection

Empirical data from Augurzky, Bonnet, Breheny, Cremers, Ebert, Mayr, Romoli, Steinbach & Sudo (2023), "Putting plural definites into context," Sinn und Bedeutung 27: 19-32.

Empirical contribution

Augurzky et al. extend the Križ & Chemla 2015 paradigm ([kriz-chemla-2015], formalized at Studies/KrizChemla2015.lean) by manipulating the QUD between participants:

• Strict context: QUD targets the strongest reading.
• Lax context: QUD targets the weakest reading.

Across two experiments, they find:

• every is highly QUD-sensitive (low acceptance in strict, high in lax).
• no is essentially QUD-insensitive (low acceptance in both).
• not every patterns with every (high in lax), NOT with no — despite both no and not every being downward-entailing.

The no / not every asymmetry is the central puzzle.

Provenance

This data was previously bundled inside an imprecision-projection file (then Studies/Haslinger2025.lean). Moved here at 0.230.521 — the empirical anchor is Augurzky et al. 2023, not Haslinger. The asymmetry's theoretical explanation in the original file invoked exhaustification logic from [bar-lev-2021] rather than Augurzky's or Haslinger's account; that explanation has been replaced with a statement of the empirical pattern alone, with the rival explanations cited as future-work targets.

structure AugurzkyEtAl2023.QUDManipulationDatum : Type

QUD-manipulation datum for plural-definite acceptance in gap scenarios.

Source: [ABB+23], Experiments 1-2.

The acceptance fields are coded categorically ("low"/"medium"/"high") since the original numerical rates depend on per-experiment baselines and stimulus sets; consult [ABB+23] Tables 1-2 for raw rates.

• operator : Generalizations.HomogeneityProjection.EmbeddingOperator

  The embedding operator

• sentence : String

  Sentence

• strictReading : String

  Strict reading (QUD = strong)

• laxReading : String

  Lax reading (QUD = weak)

• gapScenario : String

  Gap scenario

• strictContextAcceptance : String

  Acceptance rate in STRICT context (gap scenario)

• laxContextAcceptance : String

  Acceptance rate in LAX context (gap scenario)

• contextEffect : Bool

  Is there an interaction (context effect differs by operator)?

@[implicit_reducible]

instance AugurzkyEtAl2023.instReprQUDManipulationDatum : Repr QUDManipulationDatum

Equations

• AugurzkyEtAl2023.instReprQUDManipulationDatum = { reprPrec := AugurzkyEtAl2023.instReprQUDManipulationDatum.repr }

def AugurzkyEtAl2023.instReprQUDManipulationDatum.repr : QUDManipulationDatum → ℕ → Std.Format

Equations

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

def AugurzkyEtAl2023.everyQUDEffect : QUDManipulationDatum

Equations

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

def AugurzkyEtAl2023.noQUDEffect : QUDManipulationDatum

Equations

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

def AugurzkyEtAl2023.notEveryQUDEffect : QUDManipulationDatum

Equations

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

structure AugurzkyEtAl2023.NoNotEveryAsymmetryDatum : Type

The no / not every asymmetry: empirical pattern only.

Both operators are downward-entailing, yet not every permits the weak/non-maximal reading in gap scenarios while no does not. This is the central empirical puzzle of [ABB+23].

Two rival theoretical accounts in the literature (cited as future-work targets, NOT endorsed by this file):

• [BL21]: exhaustification — not every triggers a scalar implicature creating a non-monotonic context where embedded exhaustification can strengthen the embedded plural; no lacks the triggering implicature.
• [Has25] §3.6.2 (Magri effects): the asymmetry follows from how potential p-equivalence and complexity interact with embedding monotonicity; see also Studies/Haslinger2025.lean.

The two accounts make divergent predictions for embedded environments not yet tested experimentally.

• noSentence : String

  no sentence

• notEverySentence : String

  not every sentence

• gapScenario : String

  Gap scenario

• noPermitsWeak : Bool

  no permits weak reading in gap?

• notEveryPermitsWeak : Bool

  not every permits weak reading in gap?

def AugurzkyEtAl2023.instReprNoNotEveryAsymmetryDatum.repr : NoNotEveryAsymmetryDatum → ℕ → Std.Format

Equations

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

@[implicit_reducible]

instance AugurzkyEtAl2023.instReprNoNotEveryAsymmetryDatum : Repr NoNotEveryAsymmetryDatum

Equations

• AugurzkyEtAl2023.instReprNoNotEveryAsymmetryDatum = { reprPrec := AugurzkyEtAl2023.instReprNoNotEveryAsymmetryDatum.repr }

def AugurzkyEtAl2023.noNotEveryAsymmetry : NoNotEveryAsymmetryDatum

Equations

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

def AugurzkyEtAl2023.qudManipulationData : List QUDManipulationDatum

Equations

• AugurzkyEtAl2023.qudManipulationData = [AugurzkyEtAl2023.everyQUDEffect, AugurzkyEtAl2023.noQUDEffect, AugurzkyEtAl2023.notEveryQUDEffect]