Exfoliation: towards a derivational theory of clause size
David Pesetsky
April 2019

Across the languages of the world, we repeatedly find that the extraction of the subject of an embedded clause correlates with a reduction in the size of that clause. This generalization suggests a novel unification of two questions that are not generally considered to represent parts of the same puzzle:
  1. Why do non-finite clauses exist in the first place, and why do the properties of subject position in nonfinite clauses differ so often from their counterparts in finite clauses?
  2. What accounts for the obligatory absence in so many languages of the normal declarative complementizer when the subject is Ā-extracted — so-called "complementizer-trace effects"?
Common accounts of the special properties of the subject position of non-finite clauses take for granted the fact that such clauses exist in the first place, viewing this as a consequence of free lexical choice — and attribute the special properties of their subjects to deficiencies related to case, agreement or tense. Such accounts put nonfiniteness in the explanatory driver's seat and view the special properties of the subjects of nonfinite clauses as derivative. I argue for the opposite logic: that nonfinite clauses start out full and finite, and are rendered nonfinite as a by-product of subject extraction — which triggers a rule of Exfoliation that peels away outer layers of the clause so the subject ends up occupying its edge. This proposal is essentially a 21st-century revival of an idea from the earliest days of generative grammar: that infinitival clauses are derivationally created from clauses that start their lives full and finite, as a consequence (not a cause) of processes such as raising and control.

On this view, complementizer-trace effects are just the consequence of a shallower application of Exfoliation that leaves tense and agreement untouched, but peels away the CP layer, and perhaps some others — once again, in order to place the subject at the edge of the clause, so that it can be extracted. Though infinitivization is usually studied in the context of A-movement and complementizer-trace effects in the context of Ā-movement, instances of Ā-movement-triggered infinitivization and A-movement-triggered complementizer-trace effects are also found.

This is very much a draft, posted to solicit comments.

2019-02-09: a few typos fixed, some footnotes relevant to the literature on cartography added
2019-02-20: typo corrections only
2019-02-23: added a short section on specifiers of CP, other small fixes
2019-03-02 typo corrections throughout; small additions to the hyperraising section concerning Serbian data (fn 62 etc.)
2019-03-16: a number of the missing sections identified on p.2 of earlier versions have been added, most notably discussion in several places of comparison with Müller's proposals for structure removal — including a short discussion of passive (adapting Müller's analysis) in the final section.
2019-03-24: footnotes about Heavy NP Shift added, trees fixed.
2019-04-09: footnote about derivational timing and nominative objects added, other small fixes.

Format: [ pdf ]
Reference: lingbuzz/004440
(please use that when you cite this article)
Published in: draft - comments welcome
keywords: infinitive, raising, complementizer, complementizer-trace phenomena, that-trace, syntax
previous versions: v10 [March 2019]
v9 [March 2019]
v8 [March 2019]
v7 [March 2019]
v6 [February 2019]
v5 [February 2019]
v4 [February 2019]
v3 [February 2019]
v2 [February 2019]
v1 [February 2019]
Downloaded:848 times


[ edit this article | back to article list ]