In Greek, coordinated nominals of mismatched gender (e.g. masculine and feminine) trigger masculine resolved agreement when the conjuncts denote humans but neuter agreement when the conjuncts denote inanimates. In this talk, we offer a unified, novel perspective on grammatical gender features and coordination resolution that derives this generalization and related facts through a synthesis of i) a feature geometry (à la Harley and Ritter 2002) in which [fem] is a dependent of [masc], with [neut] as a default interpretation of the organizing node [class], and ii) a dual-feature system that distinguishes between interpretable and uninterpretable features (along the lines of Smith 2015, 2021) and determines the options for coordination resolution, the result of which ultimately feeds a single output for PF realization. In addition to reflecting patterns of markedness in the language and capturing 'core' facts from various environments with resolved agreement, our system correctly predicts the behavior of previously underappreciated patterns, such as resolved agreement with conjuncts of mismatched animacy, including those in which one conjunct is a fixed-gender human-denoting nominal. We further demonstrate how our system correctly predicts both gender agreement outside of coordination resolution in Greek, such as with Closest Conjunct Agreement and Backward Control, and systematically derives resolution patterns beyond Greek for other three-gender languages from minor variations to the feature geometry, which we illustrate for Bosnian/Croatian/Serbian and Icelandic. Our findings have implications for the representation of markedness and animacy in gender systems and for the status of coordination resolution in the grammar.
