غلامحسین کریمی دوستان

Kurmanji uses two morphological forms of pronouns and R-expressions: direct and oblique. The direct form of first and second pronouns (local pronouns) is fully specified for person and number. In contrast, direct third-person pronouns and R-expressions are unmarked; however, these arguments still exhibit person and number agreement with verbs and number and gender agreement with Ezafe heads. There are two approaches to deal with direct unmarked arguments. In the asymmetric approach, these arguments are assumed to be fully specified for relevant φ-features, which are deleted at PF (Phonological Form) after agreement, whereas in the symmetric approach, they are merged as underspecified for relevant features and acquire their interpretation via agreement. Given the prevalence of unmarked arguments in Kurmanji, the symmetric approach is more economical and minimalistic, requiring fewer theoretical tools and computations. The present study shows that the key distinction between local pronouns and unmarked arguments lies in the specification of relevant φ-features. Direct unmarked arguments enter the derivation underspecified for these features, establishing a relation of ‘subset control’ in agreement and acquiring their interpretation by spreading features from the agreeing head at the LF (Logical Form) interface. This study also reveals that agreement in Ezafe constructions and pro-drop sentences also involve subset control and agreement is the only possible way to license φ-features, such as gender in inanimate arguments and person in R-expressions.