GenHM and Do-Support #
@cite{arregi-pietraszko-2021}
Connects the GenHM formalization to empirical data from SubjectAuxInversion.lean.
Structure #
§1 English terminal strength assignment §2 English GenHM chain configurations for the five do-support contexts §3 Theorems pairing each empirical datum with GenHM predictions §4 The parallelism theorem: do-support uniformity across all five contexts §5 Deriving VMovementParam from GenHM
Central Result #
The parallelism of do-support across negation, SAI, verum focus, tag questions, and VP ellipsis is a DERIVED consequence of GenHM chain structure, not a stipulation about the V-movement parameter. The five contexts involve three distinct structural reasons for chain-splitting — weak intervention, probe displacement, and goal absence — yet all produce the same do-support outcome because spell-out depends only on WHETHER the chain is split.
English terminal strength: Neg and Foc are weak; all others strong.
Equations
Instances For
The five do-support contexts, formalized as GenHM chains. Each chain has:
- A probe (T) and goal (V)
- A split reason encoding WHY the chain is split
- The English strength assignment
The five chains involve three distinct split mechanisms:
- Split-by-Intervention: negation (Neg), verum focus (Foc)
- Split-by-Displacement: SAI (T displaced to C)
- Split-by-Deletion: tag questions, VP ellipsis (goal absent)
Negation chain: T ... Neg ... V
"Sue does not eat fish" — Neg (weak) intervenes between T and V. Split-by-Intervention.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Verum focus chain: T ... Foc ... V
"Sue DOES eat fish" — Foc (weak) intervenes between T and V. Split-by-Intervention.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Question chain (SAI): C ← T ... V
"Where does Sue eat fish?" — T is displaced to C via GenHM(C,T), breaking the T–V chain. The M-value cannot lower to V because T is no longer structurally adjacent to V. Split-by-Displacement.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Tag question chain: T ... [VP absent]
"She likes him, doesn't she?" — VP is anaphoric/absent in the tag. The M-value cannot lower because the goal is not available. Split-by-Deletion.
Equations
- One or more equations did not get rendered due to their size.
Instances For
VP ellipsis chain: T ... [VP deleted]
"She runs faster than he does" — VP is elided at PF. The M-value cannot lower because the goal has been deleted. Split-by-Deletion.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A declarative chain with no split: T ... V
"Sue eats fish" — clear chain, M-value lowers to V (affix hopping).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Well-formedness: intervention chains have genuinely weak interveners.
In declaratives, the M-value lowers to V (affix hopping).
In declaratives with lexical verbs, no do-support is needed.
ex32 "Sue does not eat fish" — negation + lexical V → do-support
ex34 "Sue is not eating fish" — negation + auxiliary → no do-support
ex27 "Where does Sue eat fish?" — question + lexical V → do-support
ex30 "Where is Sue eating fish?" — question + auxiliary → no do-support
ex39 "Sue DOES eat fish" — verum + lexical V → do-support
ex40 "She IS eating fish" — verum + auxiliary → no do-support
ex36 "She likes him, doesn't she?" — tag + lexical V → do-support
ex38 "She runs faster than he does" — VP ellipsis + lexical V → do-support
The Parallelism Theorem (lexical verbs): Do-support is triggered in ALL five contexts with lexical verbs (contentless T), despite three different structural reasons for chain-splitting.
With auxiliaries, do-support is never needed in ANY context.
The abstract parallelism: for ANY two chains with the same split status, the do-support decision is identical. The reason for the split (intervention, displacement, deletion) is irrelevant.
A clear chain (no split) yields the .raises surface pattern.
A split chain yields the .inSitu surface pattern.
The Pollock1989 needsDoSupport function is consistent with GenHM
predictions for lexical verbs across all contexts.
The Pollock1989 needsDoSupport function is consistent with GenHM
predictions for auxiliaries across all contexts.