Quotative Inversion as Smuggling #
@cite{storment-2026}
Formalizes @cite{storment-2026}: the smuggling derivation of quotative inversion in English and Setswana.
Storment's central claims #
- §2 + §3: In QI clauses (
"Hello," said Mary), the VP containing the quote moves to a position above the agentive AGENT, which remains in-situ in Spec,vP. Evidence: agreement, parasitic gaps, raising, conjoint/disjoint marking in Setswana. - §4: This is smuggling (@cite{collins-2005}): the VP moves to Spec,VoiceP, making the theme accessible to T⁰ for Case licensing. The smuggling projection is identified as VoiceP, which is not the external-argument-introducing head (departing from @cite{kratzer-1996}/Pylkkänen).
- §5: The transitivity constraint — QI is blocked when multiple DPs compete for Case licensing — falls out from the Case-licensing configuration after smuggling.
- §6: Locative inversion shares the same mechanism. QI and LI are both nonactive inverse-voice constructions.
Cross-paper meta-bridges (live elsewhere per CLAUDE.md convention) #
The following comparisons are the formalizer's synthesis, not Storment's claims. They live in topic-named files alongside this study:
../Unaccusativity/ProtoRoles.lean— Dowty proto-role profiles vs. Storment's QI-based unaccusativity classification (the MoS divergence)../Unaccusativity/VerbClasses.lean— Levin verb classes vs. the QI classification (alignment + the §37.3 within-class split)../Unaccusativity/IslandSensitivity.lean— @cite{storment-2026}'s smuggling-licenses-QI prediction vs. @cite{lu-pan-degen-2025}'s discourse-sourced MoS island finding (compatibility, not contradiction)
§1 + §2. Lexical annotations and QI data #
Per @cite{storment-2026}, every MoS verb passes the QI diagnostic and
is classified unaccusative; the canonical communication verbs speak/
talk fail QI and are unergative. Quantified theorems collapse the
per-verb pattern; specific instances are recoverable by fin_cases.
MoS verbs annotated unaccusative on the basis of the QI diagnostic.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Canonical unergative communication verbs that fail QI.
Equations
Instances For
§3. TransitivityClass derivation #
Maps a VerbCore to its three-way transitivity classification used by
the auxiliary-selection system (Phenomena/AuxiliaryVerbs/Selection.lean).
Stays in this study file because TransitivityClass lives in Phenomena/
and cannot be imported by Theories/.
Derive TransitivityClass from VerbCore fields.
Equations
- One or more equations did not get rendered due to their size.
Instances For
§4. Voice bridge #
VerbCore.voiceFor (defined in Theories/Interfaces/SyntaxSemantics/ VerbSmuggling.lean) maps unaccusative→non-thematic Voice and
unergative→agentive Voice. Per @cite{storment-2026}'s §4.3, the Voice
head is the smuggling projection (not the external-argument introducer
of @cite{kratzer-1996}); permitting smuggling is equivalent to being
non-phase, which is equivalent to not introducing an external argument.
§5. Auxiliary selection bridge #
In split-auxiliary languages (Italian, French, German), unaccusatives select be and unergatives select have.
§6. Levin §37.3 mannerOfSpeaking class membership #
Pure data — the divergence and within-class split analysis lives in
../Unaccusativity/VerbClasses.lean.
§8. Smuggling derivation of QI #
VerbCore.derivedQI (defined in Theories/Interfaces/SyntaxSemantics/ VerbSmuggling.lean) derives QI licensing from two independently
motivated properties: (1) Voice is non-phase (= unaccusative);
(2) verb has a complement (the quote).
These two properties are then verified against the empirical QI
diagnostic data: every MoS unaccusative with a complement is correctly
predicted to license QI; agentive speak/talk is correctly predicted
to block QI; unaccusative arrive (no complement) is correctly
predicted not to license QI (it requires LI, not QI).
arrive is unaccusative but has no complement: doesn't license QI.
This is correct — *"arrived Mary" requires a fronted locative
(LI), not a fronted quote (QI).
Consistency: each (verb, QI-datum) pair has its empirical result
matching derivedQI. Pairs the diagnostic data in
Unaccusativity/Data.lean with the smuggling prediction.
§9. QI ∥ LI distributional contrasts (Storment §6) #
Storment §6: QI and LI share the smuggling mechanism but differ in
their inputs (quote vs. locative PP) and distribution. Both are subject
to the transitivity constraint (§5). The shared inverse-voice family
membership is captured by Minimalist.qiCanonical and liCanonical in
Theories/Syntax/Minimalism/Movement/InverseVoice.lean.
whisper passes QI but is only marginal in LI — same mechanism (smuggling), different inputs.
The transitivity constraint (§5): QI is blocked with multiple DP
arguments (using warn per Storment eq. 125, naturally
ditransitive); QI is fine with a quote + PP goal.
LI categorically blocks pronominal subjects; QI merely degrades them.
LI blocks transitive verbs, just as QI does.
Unified smuggling analysis (§6): LI with arrive works because
arrive projects non-thematic Voice, permitting VP-smuggling — the
same mechanism that licenses QI.
§11 + §12. The QI derivation (Storment §3 + §4) #
The smuggling derivation assigns each major constituent to a structural position. Each position predicts observable consequences tested against the §3 structural-evidence data.
Quote vs. quotative operator (Storment §3.5, eq. 103). The quote
itself is not in the syntactic derivation — it may be totally absent
from QI clauses (Says me!). What sits in Spec,TP is a null quotative
operator (the THEME), bound by a Discourse⁰[QUOT] head in DiscourseP.
The fields below distinguish the operator's landing site (Spec,TP) from
the quote's binding head (DiscourseP).
Structural position in the QI derivation.
- specTP : QIPosition
- specVoiceP : QIPosition
- specvP : QIPosition
- discourseQUOT : QIPosition
Instances For
Equations
- Phenomena.ArgumentStructure.Unaccusativity.Bridge.instDecidableEqQIPosition x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
The QI derivation assigns four major constituents to structural
positions. Note that quoteBinder is the binding head in
DiscourseP, not the quote itself (which is not in the syntax —
Storment §3.5).
- themePosition : QIPosition
- agentPosition : QIPosition
- vpPosition : QIPosition
- quoteBinder : QIPosition
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.
- Phenomena.ArgumentStructure.Unaccusativity.Bridge.instBEqQIDerivation.beq x✝¹ x✝ = false
Instances For
Storment's smuggling derivation of QI.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Each structural position predicts an observable property. We verify each prediction against the §3 data. The bridge theorems below pair the position assignment (Storment's claim) with the empirical observation (also Storment's claim).
Theme-as-operator in Spec,TP → agreement can track agent via defective circumvention (§3.1).
Theme-as-operator is phi-deficient → Setswana SM surfaces as default SM17 (§3.1).
Theme-as-operator A-moves → cannot license parasitic gaps (§3.2).
Theme-as-operator A-moves → compatible with subject-to-subject raising (§3.3).
Agent in Spec,vP → Setswana disjoint morpheme blocked (§3.4).
Quote (separately from operator) bound by Discourse⁰[QUOT] → can split around verb + agent, need not be grammatical (§3.5).
VP smuggling predicts: VP-internal material (complements) precedes Agent; vP-external material (adjuncts) follows Agent (§2).
§13. Inverse-voice family membership #
QI is one instance of the inverse-voice family (§4.3 + §6 + §7). The
canonical instance lives in Theories/Syntax/Minimalism/Movement/ InverseVoice.lean; here we just affirm membership.
§14. Defective circumvention derives the agreement contrast #
Storment §3.1.4 (eq. 59): the difference between Setswana QI agreement
(always SM17 default) and English QI agreement (optionally tracks the
postverbal agent) reduces to a single parameter — whether the probe T⁰
is allowed to re-probe past the defective quotative-theme operator. The
defective-circumvention operation is in Theories/Syntax/Minimalism/ Probing/DefectiveCircumvention.lean; here we wire it to the QI
agreement data.
The theorems abstract over the precise feature bundles and feature-
compatibility predicate — Storment's substantive claim is that the
operation is the same and only the allowReprobe parameter varies.
The same defective-probing situation produces Setswana's obligatory default agreement (no re-probe) and English's optional agent-tracking agreement (re-probe with compatible features) — a single Bool parameter accounts for the cross-linguistic split.
Storment's English-specific prediction: a 1st/2nd person agent
(whose phi-features clash with the defective theme's [3]) cannot
license re-probe — *"What do we do now?" ask we. The derivation
crashes on feature incompatibility (eq. 46, page 14).