The Role of Functional Features in the Derivational Procedure: A New Account of the EPP-Effects, Case and Agreement

In this paper, I propose an alternative account of feature checking/valuation based solely on categorial features. The main assumption is that lexical heads have an exclusively lexical feature matrix, whereas functional heads have an exclusively functional feature matrix. The next assumption is that only lexical features allow strictly local feature checking/valuation (i. e. under Agree); functional features, on the other hand, require a syntactic operation (Merge or Move) to check/value their functional features. Interaction between functional heads (Probes) and lexical ones (Goals) is made possible only via the mediating functional heads c-selecting the lexical ones (e. g. D selects N/NP, v selects V/VP). The mediating functional heads provide lexical heads/categories with a functional layer indispensable for "communication" with the functional Probes (as in Chomsky 1999: 9). The analysis proposed here accounts for the facts traditionally ascribed to the Extended Projection Principle (EPP). I show on the basis of English and Icelandic data that the EPP-effects result from categorial functional feature checking and cannot be reduced to either Case or agreement; rather, Case and agreement are intertwined in the processes of categorial feature checking/valuation and constitute their output, but crucially not their goal.