News & Events

Note: This lecture is in a different room than usual: 325 Genesee Hall.

Abstract

Spoken language unfolds over time, with information relevant to understanding a speaker’s utterance distributed throughout the signal. Suppose Alice is talking to her friend Bob and says, "I don’t mind peas, but I hate wasps." The word "wasps" suggests that Alice intended to say "bees" rather than "peas"—how does Bob reconcile these two pieces of conflicting information? In this talk, I will explore the cognitive mechanisms underlying human listeners' ability to represent past spoken language input and integrate it with future information. I'll start by reviewing my PhD work at Rochester, which suggested that listeners are able to maintain graded representations of past speech in memory. That is, Bob retains information about how "peas"-like or "bees"-like the initial word was; later, he can refer back to that information for integration with the current input he's receiving. Next, I will discuss recent work showing that listeners' ability to integrate these sources of information has few limits, being invariant both to the relative ordering of cues and the temporal distance between them. The final portion of my talk will consider why listeners can maintain detailed information about past input when human memory is limited. Ongoing work in my lab suggests that listeners can calibrate the amount of information they retain about previous input given current demands. I will close by considering the broader implications of this work for the cognitive science of language.

Event Details

Language Science Lectures include light refreshments of bagels and coffee.

Livestream

This talk will be livestreamed. Register to attend via Zoom.