Jessica Coon gave a colloquium talk titled “Constraining Agree” at Harvard last Friday, presenting collaborative work with Stefan Keine (UCLA). The abstract is below:
This talk illustrates how a constrained probe mechanism for Agree combined with the feature gluttony system laid out in Coon & Keine 2021 can account for a wide range of complex phenomena, including: (i) hierarchy effects in agreement; (ii) PCC effects involving clitics; and (iii) patterns in which A’-movement is restricted to targeting the closest DP. All of these patterns appear to involve defective intervention, in which an element that is not an eligible target for an operation nonetheless blocks that operation from crossing it. We show that a feature gluttony analysis allows us to understand these patterns without appeal to defective intervention. Instead, problems arise not because agreement fails, but rather because Agree is successful with more than one goal. A probe which has participated in Agree with more than one goal may then create problems down the line for the morphology or syntax. We further show that attested patterns of variation in the systems listed above can be handled without appeal to parametric variation in the basic Agree mechanism (e.g. Multiple Agree, Contiguous Agree, Dynamic Interaction). While the core insights of a feature gluttony system are in principle compatible with any of these types of Agree, we aim to derive attested variation not from the mechanics of Agree, but from independent properties in the languages in question. The result is a more constrained syntax, which makes testable predictions about cross-linguistic variation.