Documentation

Linglib.Phenomena.ArgumentStructure.Unaccusativity.IslandSensitivity

Unaccusativity and extraction-island sensitivity #

@cite{lu-pan-degen-2025} @cite{storment-2026}

How the unaccusativity classification of manner-of-speaking verbs interacts with their participation in extraction-island effects. This file is the formalizer's synthesis: @cite{storment-2026} doesn't discuss MoS islands, and @cite{lu-pan-degen-2025} doesn't run a syntactic unaccusativity diagnostic. Each paper studies a different property of the same verb class.

The two operations on MoS verbs #

@cite{lu-pan-degen-2025}: wh-extraction from within the complement of an MoS verb is degraded — *Who did Mary whisper that John kissed __? The constraint is discourse-sourced (ameliorated by prosodic focus, unrelated to verb-frame frequency, replicable with say+adverb framing that has no structural change).

@cite{storment-2026}: A-movement of the entire VP containing the quote operator is grammatical — "I'm tired," whispered Mary. The operation is VP-smuggling to Spec,VoiceP, licensed by the verb's non-phase Voice (anticausative).

Compatibility, not contradiction #

The two observations are consistent. They concern different operations:

The asymmetry confirms both analyses simultaneously: MoS islands are discourse-sourced (so they don't block syntactic VP-smuggling that produces QI). If MoS islands were syntactic (PIC, subjacency), they would be predicted to block QI as well — they don't.

§1. The two empirical observations #

mosIslandSources is from @cite{lu-pan-degen-2025} (Phenomena/Islands/ MannerOfSpeaking.lean); derivedQI is the syntactic prediction from @cite{storment-2026} (Theories/Interfaces/SyntaxSemantics/VerbSmuggling.lean). For every MoS verb the Fragment lists as unaccusative, both observations hold simultaneously.

source
def Phenomena.ArgumentStructure.Unaccusativity.IslandSensitivity.mosUnaccusatives :
List Fragments.English.Predicates.Verbal.VerbEntry

The MoS verbs that pass QI (per @cite{storment-2026}) and participate in MoS islands (per @cite{lu-pan-degen-2025}).

Equations
  • One or more equations did not get rendered due to their size.
Instances For
    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.IslandSensitivity.mos_islands_discourse_sourced :
    Islands.MannerOfSpeaking.mosIslandSources = [Islands.IslandSource.discourse]

    @cite{lu-pan-degen-2025}'s finding: MoS islands are discourse- sourced.

    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.IslandSensitivity.mos_unaccusatives_license_qi (v : Fragments.English.Predicates.Verbal.VerbEntry) :
    v ∈ mosUnaccusatives → v.derivedQI = true

    @cite{storment-2026}'s prediction: every MoS unaccusative licenses QI. Quantified version of the per-verb theorems.

    §2. Compatibility theorem #

    Stating both observations together: each MoS verb in the list is extraction-island-sensitive (sub-extraction blocked) AND smuggling- licensed (whole-VP A-movement OK). Different operations, different sources, both true.

    source
    theorem Phenomena.ArgumentStructure.Unaccusativity.IslandSensitivity.mos_extraction_smuggling_asymmetry :
    Islands.MannerOfSpeaking.mosIslandSources = [Islands.IslandSource.discourse] ∧ ∀ v ∈ mosUnaccusatives, v.derivedQI = true

    §3. The diagnostic value #

    Were MoS islands syntactically sourced (e.g., PIC, subjacency), they would be predicted to block VP-smuggling as well — A-movement of an opaque domain is no easier than wh-movement out of one. The fact that QI is grammatical is evidence that MoS islands are not syntactic, independent of @cite{lu-pan-degen-2025}'s direct experimental tests. The two literatures converge on the same conclusion.