Ellipsis as Catena-Targeting #
@cite{osborne-2019} @cite{osborne-gross-2012}
Formalizes @cite{osborne-2019}'s thesis: ellipsis targets catenae, not constituents. This is the primary empirical motivation for the catena concept — it explains why VP ellipsis, gapping, pseudogapping, sluicing, stripping, and fragment answers all elide connected-but-not-necessarily-complete-subtree material.
Key Insight #
In DG, the "VP" is NOT a complete subtree (constituent) because the verb's subject is also its dependent. This means even VP ellipsis targets a catena, not a constituent — making catenae essential.
Bridges #
- →
Catena.lean: usesisCatena,isConstituentfor proofs
Ellipsis types in English (@cite{osborne-2019}, Ch 12–13).
- vpEllipsis : EllipsisType
- gapping : EllipsisType
- pseudogapping : EllipsisType
- stripping : EllipsisType
- sluicing : EllipsisType
- fragmentAnswer : EllipsisType
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
VP Ellipsis: "She will cook dinner, and he will too" Pre-ellipsis second clause: cook(0) → he(1:nsubj), will(2:aux), dinner(3:obj). Elided = {0, 3} (cook + dinner): connected via obj → catena. Subtree of cook = {0,1,2,3} ≠ {0,3} → NOT constituent. @cite{osborne-2019}.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Elided nodes in VP ellipsis: {cook, dinner} = {0, 3}.
Equations
Instances For
Gapping: "Fred eats beans and Jim rice" Pre-ellipsis second clause: eats(0) → Jim(1:nsubj), rice(2:obj). Elided = {0} (eats only): singleton → catena. Subtree of eats = {0,1,2} ≠ {0} → NOT constituent. @cite{osborne-2019}.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Elided nodes in gapping: {eats} = {0}.
Equations
Instances For
Pseudogapping: "She has helped him, and he has her too" Pre-ellipsis: helped(0) → he(1:nsubj), has(2:aux), her(3:obj), too(4:advmod). Elided = {0} (helped only): singleton → catena, NOT constituent. @cite{osborne-2019}.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Elided nodes in pseudogapping: {helped} = {0}.
Equations
Instances For
Sluicing: "She helped someone, but I don't know who" Embedded clause (pre-ellipsis): helped(0) → she(1:nsubj), someone(2:obj). Elided = {0, 1} (helped + she): connected via nsubj → catena. Subtree of helped = {0,1,2} ≠ {0,1} → NOT constituent. @cite{osborne-2019}.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Elided nodes in sluicing: {helped, she} = {0, 1}.
Equations
Instances For
Fragment answer: "Who helped? — Him." Full answer: helped(0) → him(1:obj). Elided = {0} (helped): singleton → catena, NOT constituent. @cite{osborne-2019}.
Equations
- DepGrammar.Ellipsis.fragmentTree = { words := [Word.mk' "helped" UD.UPOS.VERB, Word.mk' "him" UD.UPOS.PRON], deps := [{ headIdx := 0, depIdx := 1, depType := UD.DepRel.obj }], rootIdx := 0 }
Instances For
Elided nodes in fragment answer: {helped} = {0}.
Equations
Instances For
VP ellipsis elided material is a catena.
VP ellipsis elided material is NOT a constituent (not a complete subtree).
Gapping elided material is a catena.
Gapping elided material is NOT a constituent.
Pseudogapping elided material is a catena.
Pseudogapping elided material is NOT a constituent.
Sluicing elided material is a catena.
Sluicing elided material is NOT a constituent.
Fragment answer elided material is a catena.
Fragment answer elided material is NOT a constituent.
@cite{osborne-2019}: All five types of ellipsis target catenae. Verified for all example trees.
@cite{osborne-2019}: ALL five ellipsis types target non-constituent catenae. This is the empirical advantage of catenae over constituents: a constituent-based theory cannot capture these ellipsis patterns.