ABSTRACT

Here I show that an investigation of the minutiae of Agree can lead to a better understanding of symmetries and asymmetries in instances of multiple agree (in both the A-and A-bar domain). Crucial to this account is the decomposition of Agree into suboperations (Match and Value), the pervasive role of minimality as a locality condition, and the idea that the computational system is strongly derivational in character. Of particular relevance here is the claim that case and agreement features must be regarded as features with a parallel function. Specically, agreement features are analyzed as interpretable on the Goal (NP) but uninterpretable on the Probe (T/v). The reverse situation obtains for Case: It is an interpretable feature on the Probe (corresponding to the Tense/Aspect feature), but uninterpretable on the Goal (NP).