Gong 2022 @cite{gong-2022} #
Case in Wholesale Late Merger: Evidence from Mongolian Scrambling. Linguistic Inquiry, Early Access.
Core claim #
Condition C reconstruction effects in Mongolian scrambling are controlled by case assignment, not by the A/A-bar distinction or the special status of subject binders. Wholesale Late Merger (WLM) can bleed Condition C iff the movement chain has a case position above the pronoun binder.
Mongolian hybrid case system #
- ACC = dependent case (assigned by competition between two NPs in the same phase; @cite{baker-vinokurova-2010})
- NOM = assigned by finite T via Agree
- DAT = nonstructural (inherent)
Key empirical patterns #
Clause-internal scrambling:
- SS over IO (DAT) binder: no Condition C effect (WLM bleeds)
- IS over IO binder: no Condition C effect
- SS/IS over Subject (NOM) binder: obligatory Condition C reconstruction
Clause-external scrambling (LDS):
- LDS of ACC OBJ, matrix DAT binder: no obligatory Condition C effect
- LDS of ACC OBJ, matrix NOM (Subject) binder: obligatory Condition C
PP-scrambling:
- Always shows obligatory Condition C reconstruction, regardless of binder position (PPs lack the DP structure required for WLM)
Negative result #
The A/A-bar distinction does not predict these patterns. Mongolian SS and IS both behave like A-movement in terms of anaphor binding and WCO amelioration, yet they diverge in Condition C reconstruction when the binder changes between IO and Subject. The @cite{frank-lee-rambow-1996} Subject Binding Generalization also fails for Mongolian: scrambling over a subject binder can bleed Condition C in LDS when a dependent ACC position is available in the matrix clause.
A scrambling scenario encoding the empirical data from @cite{gong-2022}. Each scenario records the scrambling type, the binder's grammatical role, and whether Condition C reconstruction is observed.
- label : String
- scrambleType : Fragments.Mongolian.Case.ScrambleType
- binderRole : Fragments.Mongolian.Case.BinderRole
- reconstructs : Bool
true= obligatory Condition C reconstruction observed.false= no Condition C effect (WLM bleeds reconstruction).
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Gong2022.instReprScrambleScenario = { reprPrec := Gong2022.instReprScrambleScenario.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
(18b) SS over IO: DO containing R-expression Cemeg scrambled over DAT pronoun binder tuund. No Condition C violation. 'Cemeg's book, (the) teacher gave (to) her.'
Equations
- Gong2022.ss_io = { label := "SS over IO (18b)", scrambleType := Fragments.Mongolian.Case.ScrambleType.SS, binderRole := Fragments.Mongolian.Case.BinderRole.io, reconstructs := false }
Instances For
(18a) SS over IO, base order: DAT pronoun tuund c-commands R-expression Cemeg inside the ACC DO. Condition C is violated. '*The teacher gave her Cemeg's book.'
Equations
- One or more equations did not get rendered due to their size.
Instances For
(19b) IS over IO: DO scrambled past IO and subject. IO binder is DAT. No Condition C violation. 'Cemeg's book, (the) teacher gave (to) her.'
Equations
- Gong2022.is_io = { label := "IS over IO (19b)", scrambleType := Fragments.Mongolian.Case.ScrambleType.IS, binderRole := Fragments.Mongolian.Case.BinderRole.io, reconstructs := false }
Instances For
(20b) IS with Subject binding DO, transitive: Subject pronoun ter (NOM) binds R-expression Cemeg inside DO. Obligatory reconstruction. '*Cemeg's book, she tore.'
Equations
- One or more equations did not get rendered due to their size.
Instances For
(21b) IS with Subject binding DO, ditransitive: Subject pronoun ter (NOM) binds R-expression Cemeg inside DO. Obligatory reconstruction. '*Cemeg's book, she gave to Bat.'
Equations
- One or more equations did not get rendered due to their size.
Instances For
(41) LDS of ACC OBJ with matrix DAT binder: no obligatory Condition C. Embedded ACC object scrambled to matrix clause; matrix dative argument tuund is the binder. '?Bat's essay, Zaya said to him that the teacher read.'
Equations
- One or more equations did not get rendered due to their size.
Instances For
(40) LDS of ACC OBJ with matrix Subject binder: obligatory Condition C. Embedded ACC object scrambled to matrix clause; matrix Subject ter (NOM) is the binder. '*Bat's essay, he said that the teacher read.'
Equations
- One or more equations did not get rendered due to their size.
Instances For
WLM correctly predicts SS over IO: no reconstruction.
WLM correctly predicts IS over IO: no reconstruction.
WLM correctly predicts IS over Subject (transitive): reconstruction.
WLM correctly predicts IS over Subject (ditransitive): reconstruction.
WLM correctly predicts LDS with DAT binder: no reconstruction.
WLM correctly predicts LDS with Subject binder: reconstruction.
The A/A-bar distinction does not predict Mongolian reconstruction. SS over IO and SS over Subject involve the same scrambling type, but differ in Condition C reconstruction. An A/A-bar account would assign the same reconstruction prediction to both, since both involve the same kind of movement. Case-based WLM correctly captures the contrast by looking at case positions, not movement type.
This connects to @cite{keine-2020}'s probe profiles: even if the scrambling probe is classified as A or A-bar, its classification is constant across scenarios that differ in reconstruction behavior.
The Subject Binding Generalization (@cite{frank-lee-rambow-1996}) fails for Mongolian. That generalization predicts that scrambling over a subject binder always forces reconstruction. But in LDS, dependent ACC can be assigned in the matrix clause, allowing WLM to bleed Condition C even with a (matrix non-subject) binder. The generalization is too strong: what matters is case positions, not the subject/non-subject status of the binder per se. The correlation with subjects is an epiphenomenon of the fact that subjects typically occupy the highest case position.
The dependent case algorithm determines WLM availability:
dependentAccusative succeeding at Spec,VP (above IO) is exactly
what makes the case position available for late merger.
This connects the Mongolian fragment's WLM predictions to the
theory-layer dependent case algorithm in DependentCase.lean,
rather than stipulating case positions independently.
PP-scrambling always reconstructs, unlike DP-scrambling. @cite{gong-2022} section 6.2: PPs lack the determiner + NP restrictor structure required for WLM, so Condition C reconstruction is obligatory regardless of the binder's position. This contrast between DP- and PP-scrambling is a further prediction of the WLM account.
LDS involves successive-cyclic movement through the embedded CP edge.
The embedded ACC object moves to Spec,CP of the embedded clause (phase edge escape hatch per PIC), then to the matrix clause. The CP edge does NOT provide a case position (C passes features to T via Feature Inheritance, @cite{chomsky-2008}). Case availability in the matrix clause depends on whether a dependent case competitor exists.
This derives the LDS chain positions from phase theory rather than
stipulating them directly in casePositionsAbove.
Equations
Instances For
LDS chain predictions agree with the direct casePositionsAbove
predictions from the Mongolian fragment. The CP edge contributes
nothing (no case); only the matrix position matters.
The CP edge alone — without a matrix case position — never bleeds Condition C, regardless of the binder's height.