We don't Agree (only) upwards
- Thursday 8 October 2020
- Com(parative) Syn(tax) Series
- online via Skype (e-mail email@example.com for the link)
Bjorkman and Zeijlstra (2019) argue that Agree involves two operations: checking and valuation. Checking always happens upwards, in a configuration in which the checker of a feature, carrying an interpretable feature [iF], c-commands the checkee, carrying an uninterpretable feature [uF]. Valuation generally happens downwards, after the valuer has moved to the specifier of the valuee. This makes very clear predictions, which we show are not borne out.
We discuss several configurations involving in situ agreement controllers which have not featured in this debate: subject agreement with in-situ subjects in Matengo and German, object agreement in Sambaa and Liko, and complementiser agreement in Nez Perce. We argue that such φ-agreement phenomena indicate that Bjorkman and Zeijlstra’s (2019) proposal is empirically inadequate. Adding these data to the ongoing debate on the directionality of Agree (see Zeijlstra 2012, Preminger 2013, Bjorkman and Zeijlstra 2014, Preminger and Polinsky 2015) is important because if a given approach is not empirically adequate, the question whether it has conceptual advantages is moot.