Doctoral Dissertation Research: Hierarchical relations in real-time sentence comprehension

About This Grant

This project investigates how language comprehenders use syntactic information during real-time language comprehension. Syntactic information refers to the abstract, hierarchical organization of linguistic elements within a sentence, and has long been recognized as central to the interpretation of human language. For example, in a sentence such as “The neighbors’ kids playfully teased each other,” the interpretation of “each other” must be linked to “kids”, not “neighbors”, because of the syntactic position each of these occupies. However, relatively little is known about how such information is used in real-time language comprehension. This project tests the use of syntactic relations in real-time linguistic processing. The project employs eye-tracking methodology, a highly time-sensitive technique, to monitor the time course of online comprehension. The researchers will determine the relative importance of hierarchical syntactic organization and simple linguistic features like number in guiding attention in the earliest moments of language comprehension. In addition to training a graduate student, the findings from this project will provide insight into the cognitive mechanisms involved in language comprehension. Moreover, these findings will inform the continued development of computational language models, which continue to struggle with the hierarchical organization of human language despite achieving impressive performance on many linguistic tasks. This project explores how hierarchical relations are represented and used to guide retrieval within the working memory systems that support language comprehension. These questions are examined in two eye-tracking experiments investigating the time course of anaphor comprehension. Anaphors, such as reflexives (himself, herself) and reciprocals (each other), are elements whose meaning is dependent on their referent, which must stand in a hierarchically relevant position relative to the anaphor. The first experiment aims to determine the mechanism used to encode hierarchical information in short-term or working memory representations. Two theoretical hypotheses will be evaluated: that comprehenders keep hierarchically relevant items in active attention, rendering them immediately available in language comprehension, or that comprehenders are able to specifically reactivate only hierarchically accessible items when needed for interpretation. These hypotheses will be tested using a visual world study that allows for direct investigation into the time course of referent access. In this task, comprehenders will listen to sentences while viewing a visual display containing relevant referents, and their eye movements on the display will be recorded. The second experiment investigates whether hierarchical relations constrain predictive processes. This study will employ an eye-tracking-while-reading paradigm, where participants read sentences silently while their eye movements are recorded. If hierarchical relations are rapidly employed, hierarchically unavailable referents should not influence reading behavior, whereas hierarchically available ones should. Together, these two experiments aim to shed light on the mental representations and operations that language comprehenders rely on when using hierarchical information to constrain interpretation in real-time processing. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.