Locating agreement in grammar: an argument from agreement in conjunctions

摘要

In Hindi-Urdu, T(ense) can agree with non-overtly case-marked subjects or objects. Despite being controlled by the same head and being sensitive to the same morphological properties of the agreement target, agreement inside conjunction structures reveals differences between subjects and objects: agreement with objects is sensitive to linear proximity, while agreement with subjects is not. This difference shows itself in two sets of conjunction structures: agreement with conjoined subjects and objects, and agreement in Right Node Raising. This difference between agreement with co-ordinated subjects and objects is addressed in terms of two questions: why does object agreement not access the same features as subject agreement, and how does it access the features of the closest NP in the coordinated object. We argue that the answers to these questions show that agreement is largely syntactic, but that post-syntactic processes can be recruited for agreement when syntactic processes have failed to value agreement in the syntax. The inaccessibility of certain features to agreement with objects follows from Bhatt's (2005) proposal that agreement with subjects assigns case, but agreement with objects is agreement with an already case-licensed argument. While T-agreement can access the Φ-features of subjects, case assignment to the object prior to T-agreement deactivates the object's Φ-features so that T can match their features but is not valued by them. Post-syntactic processes use the matching relation between T and the inactive features of objects to retrieve values for T. This process is sensitive to linear proximity.