LSLT: Rosa Lee (LING) et al.

Planet Cloze: The expanding universe of lexical prediction models

Rosa Lee (LING), Katherine Howitt (LING), London Dixon (LING), Tal Ness (HESP), Masato Nakamura (LING)

Abstract: People actively predict when comprehending sentences in real time. Our group is interested in what kinds of information are used and how quickly they are used to generate expectations in real-time comprehension. One specific example is how people use argument role information to guide their predictions for upcoming verbs. Some of us have been puzzling over this question in particular, because it is a case where we observe a misalignment between different measures (EEG/ERPs, eye-tracking, speeded cloze production) that we had assumed to be tapping into the same underlying mechanism in real-time comprehension. This past summer, we extended our research to the Planet Word Museum in Washington D.C., where we collected data from over 300 museum visitors, including demographic groups that are typically more difficult to recruit, such as school-aged kids, older adults, and non-native speakers from different backgrounds. In our LSLT, we plan to share our experience of adapting a psycholinguistic experiment to a museum setting with a more diverse population, as well as the findings based on the museum data that have pushed us to think about the puzzle from a new perspective.