MIT PhD thesis 2001
This study attempts to understand a variety of cross-clausal dependencies through detailed study of one language, Passamaquoddy (Algonquian). It focusses in particular on three phenomena: successive cyclic wh-movement, wh-scope marking, and raising to object. In exploring these issues I adopt and argue for a recent approach to cyclicity, the phase and Agree theory of Chomsky (1998, 1999).
The successive cyclic nature of wh-movement is shown to be visible in
Passamaquoddy in a phenomenon of agreement with moving operators (wh-phrases, relative operators, and focus operators). This agreement appears on every verb along the path of movement. I argue that the phase theory coupled with a necessary Agree relation between each verb and a moving operator can account for this pattern, as well as a complex pattern of interaction between types of extraction and verbal morphology. Successive-cyclic agreement is also able to decide between competing theories of wh-scope marking, a construction in which short-distance wh-movement takes place but gives rise to a long-distance interpreta-tion, through association with a scope-marking element in a higher position. Agreement is shown to take place even with covert movement in Passamaquoddy, in focus constructions and internally headed relative clauses. In wh-scope marking, agreement indicates that one type of scope marking construction involves covert movement of the lower wh-phrase, but a second type (the less restricted of the two) does not. This fact and others that correlate with the difference in agreement indicate that
Passamaquoddy instantiates both of the leading analyses of wh-scope marking: Direct and Indirect Dependencies (van Riemsdijk 1983, Dayal
Raising to object is shown to involve a dependency of a different kind: one that is clause-bounded in one respect but not in another. Raising to object position does not actually target object position in the higher clause; instead an NP moves just to the edge of the lower clause, where Agree can take place with the higher verb (across a clause boundary). However, raising to object apparently feeds A-movement in the higher clause; but when it does, I show that the "raised" NP must be base-generated at the edge of the lower clause and not moved out of it. The reason (the ban on improper movement) follows from the way features are checked in a cyclic derivation. Data from Japanese are brought in to show the cross-linguistic generality of the principles adduced.