Ling-F. The Many Failures of Agree* (Advanced!)(Omer Preminger, University of Maryland)
This course is advanced. Some background in generative syntactic theory is required.
In this course, we will explore the two core empirical domains that Chomsky’s (2000, 2001) Agree system was originally intended to account for: (i) syntactic relations involving valuation of PERSON, NUMBER, and/or GENDER/NOUN-CLASS features; and (ii) the assignment of (abstract) case to nominals. We will see a variety of arguments that the Agree system, as formulated, is not an adequate model for either of these two domains.
Next, we will explore what modifications are necessary in order to turn this system into an empirically adequate one. We will see that the crucial change required is that Agree (or whatever operation we replace it with) must be allowed to fail, and that this failure cannot be cause for ungrammaticality (or a "crash").
We will then explore several important consequences and extensions of these results. The first is the outright falsification of Chomsky’s (2001) “Strong Minimalist Thesis.” The second is the existence of privative features values in syntax. Specifically, the fact that traditional categories like ‘3rd person’, ‘singular’, and ‘nominative’ are not feature values unto themselves, but represent the outright absence of feature values of the relevant kind. Finally, we will see how this tolerated-failure logic extends to empirical domains beyond case and agreement.
The arguments presented will be based on primary data from Kaqchikel (Mayan); Zulu (Bantu); Sakha (Turkic); Icelandic (North Germanic); Tsez (NE Caucasian); and Basque.