Department of Linguistics

Research talk by Stefan Kaufmann with response by Kareem Khalifa.

Register to attend virtually:
https://us06web.zoom.us/webinar/register/WN_O2BggnyFRoOVv7KFjwjYJw

Linguistics Colloquium with Colin Philips

"Arguments, Prediction, and Production”

At the start of the pandemic, when we were unable to use some of our favorite comprehension measures (e.g., EEG, eye-tracking), we turned to eliciting speech production online. This turned out to be both rewarding and surprising. First, recording what people say and when they say it works rather well via the internet. Second, the production measures reveal greater sensitivity than we had seen in parallel comprehension studies. Studies on argument role reversals, such as “This is the the waitress that the customer served” show a striking contrast. In EEG and eye-tracking studies, it is commonly found that the verb “served” is processed as if it is unsurprising in a reversal, despite the fact that it is anomalous. This had led us and others to conclude that comprehenders initially ignore argument roles when generating expectations for upcoming words. But in a speeded cloze paradigm, where speakers must complete fragments like “This is the waitress that the customer …” as fast as possible, their responses are overwhelmingly appropriate, even when timing is matched with comprehension studies. This indicates that when speakers self-generate continuations, they are highly sensitive to argument roles. We had assumed that these measures were tapping into the same underlying cognitive processes. The contrast between measures challenges that assumption. This puzzle has set us on an interesting path in which we combine computational modeling and many different experimental probes, to try to get to the bottom of where argument-role insensitivity comes from. This, in turn, pushes us to think more critically about prediction mechanisms in psycholinguistics, and about the relation between speaking and understanding. In a fun twist, it also led us to run studies on the floor of the new Planet Word Museum in Washington DC. This allowed us to test children and seniors. It forced us to make our experiments a LOT more user-friendly. And it taught us that if we do this, then researchers and participants alike have much more fun doing psycholinguistic research.

ECOM Speaker Series will host Alison Springle (Oklahoma University, Philosophy). The title of her talk is "Radicalizing Practical Representations." The event will be held in Family Studies 220, and will be broadcasted via Zoom. For details, contact aliyar.ozercan@uconn.edu