Amy Rose Deal (ardeal@berkeley.edu)
This course starts with the observation that agreement sometimes shows a `more than you bargained for’ behavior: a probe that is specified to pick up a certain type of feature nevertheless picks up other features that lie between it and the feature that it seeks. “Overagreement” of this type raises a foundational question about how Agree works: is Agree really driven by the need to correct a formal deficiency of the probe, stated in terms of uninterpretable or unvalued features? I will argue that the answer is no, and lay out a theory of agreement in which Agree instead takes place in order to create featural redundancies. On this model, probes are specified for two types of information: interaction features (the features that the probe will copy back) and satisfaction features (the features that will cause the probe to stop probing). In developing this theory, we will build connections to several related research programs (e.g. Bejar and Rezac on Cyclic Agree, Preminger on failed agreement, many authors on clitics vs. agreement) and seek answers to questions such as the following:
- How does morphology handle the output of syntax?
- How do portmanteaux arise?
- What forces some languages to front all of their wh-words?
- What is the difference for syntactic representation between a D with phi-features vs. a head in the verbal projection that has gotten phi-features from a D(P)?
- Why does scrambling have so little effect on agreement?