On feature sharing and feature transfer
Giuliana Giusti · ARCA (Università Ca' Foscari Venezia) · 2009
Feature sharing is a pervasive property of natural languages which has long been considered quite puzzling. Being the source of redundancy, it is potentially a major problem in the minimalist program which aims to reduce language to a conceptually necessary system (cf. Chomsky 2005). For this reason it has been the focus of so much literature in the last decade that it is impossible to mention even the most influential pieces of work. In this paper, I claim that feature sharing is a non-homogeneous phenomenon and should be analyzed as the result of three different processes Agreement, Concord, and Projection. In current literature, feature transfer is reduced to checking and deleting uninterpretable features (Chomsky 1995). I adopt the mainstream hypothesis that Agreement arises from merger of a formal uninterpretable feature (a probe) and is checked against a constituent (the goal), in a lower specifier, which contains the interpretable counterpart of those features. This triggers movement (copy and re-merge) of the relevant features to obtain a Spec-Head configuration with the probe, resulting in either covert movement (only the feature moves) or overt movement (the moved feature pied-pipes the whole constituent which contains it). I claim that quite differently from Agreement, Concord arises from the first-merger of a modifier underspecified for uninterpretable features. In other words, Concord is directly enhanced by the Spec-Head configuration, it does not