Speaker: Danfeng Wu (MIT)
Title: Stripping and but
Time: Thursday, April 16th, 12:30pm – 2pm
Abstract: Standard analyses of clausal ellipsis (sluicing, fragment answers, and stripping) involve movement (of the remnant, the element that survives ellipsis, out of the ellipsis site) and deletion. For example, in the stripping example in (1), the remnant Chris moves out of the ellipsis site and the TP gets deleted:
(1) Pat left, (but) not Chris_i [TP 〈t_i left〉].
While movements in general respect island constraints, clausal ellipsis is known to be able to evade islands despite involving movement (especially famous is the island-insensitivity of sluicing). There have been different analyses for why sluicing and fragment answers may be island-insensitive (Merchant (2004), Griffiths and Liptak (2014) and Barros et al. (2014)). Because these analyses were proposed for clausal ellipsis in general, they should extend to stripping as well. The first part of this talk evaluates these three analyses with novel data from stripping. The data are consistent with Barros et al., but not with the other two accounts, thus supporting Barros et al.’s analysis (all movements are island-sensitive, and apparent island evasion is due to another parse that does not actually involve any island-violating movement) over the others.
The second part of the talk starts from the novel observation that while stripping without but can apparently evade islands (complex NP island in (2) and left branch island in (4)), stripping with but cannot (3) & (5).
(2) They hired someone who speaks French yesterday, not German.
(3) *They hired someone who speaks French yesterday, but not German.
(4) They bought a blue car, not green.
(5) *They bought a blue car, but not green.
I argue that it is the presence of but that causes the ungrammaticality. But in English is lexically ambiguous, and the meanings relevant to (2)-(5) are counterexpectational but (which has the implication that generally, if the first conjunct holds, the second conjunct does not) and semantic opposition but (which does not have this implication) (see e.g. Winter & Rimon (1994), Jasinskaja & Zeevat (2008, 2009), Toosarvandani (2014)). I argue that these two buts have different syntax too. Specifically, counterexpectational but (but not semantic opposition but) bans stripping, and semantic opposition but (but not counterexpectational but) requires parallel conjuncts.