出版社：Department of Linguistics, University of Toronto
摘要：This thesis explores the dependency-forming featural operation Agree and its interaction with the structure-building operation Merge in the local, derivational framework of Chomsky (2000). Chapter II develops a theory of cyclicity where individual instances of Merge can change the search-space of a probe. Such dynamic cyclicity is motivated from cyclic displacement phenomena where search-space expands downwards from one cycle to the next through evacuation of interveners, and the ergative displacement phenomenon in Basque where searchspace expands upwards through cyclic construction of the phrase marker. Chapter III shows how Merge composes with Agree in the derivation to give both movement and copy-raising. The Merge step is constrained by the Agree step because Agree transmits the variable names needed to interpret movement in Heim & Kratzer (1998). Differences between movement and copy-raising reduce to the Binding Theory and copy-deletion in the result, forcing copies in local domains and pronouns in non-local domains. English tough-movement exemplifies copy-raising whose pronoun tail is an A'-operator. Chapter IV explores the ontology of Agree, departing from Chomsky (2000) where Agree in weak agreement languages manipulates features rather than atoms and has no further consequences. It shows that in strong agreement languages Agree apparently does satisfy the EPP and is visible to the Binding Theory. It demonstrates that the weak/strong distinction does iii not reflect an ontological difference in Agree; rather, strong agreement languages make available expletives with φ-features (overtly manifested in Czech), yielding more visibility upon basegeneration than expletives of weak agreement languages. The ontological ramifications of Agree manipulating subatomic features but interacting with the structure-building Merge are discussed. Chapter V investigates what happens to the goal under Agree, from the standpoint of Case and deactivation. It proposes a theory where Agree affects the goal by assigning it a shell consisting of the category of the probe, interpreted as Case. The hypothesis derives the fact that deactivation can be partial, that shells ("Cases") can be stacked, and that the same category can assign different Case depending on how its φ-probe has been affected earlier in the derivation.
