Humans are extremely good at understanding sentences, even very complex ones which they have never heard or read before. We understand thousands of sentences every day, and we do so quickly, effortlessly, and more often than not, correctly. How do we achieve this tremendous feat?
In the Sentence Processing Lab, we study the mechanisms which enable people to understand language in real time. We investigate questions such as how readers and listeners use syntactic structures to extract meaning, how they create dependencies between distant elements in the sentence and beyond, and how they use lexical and structural prediction to anticipate upcoming material. A major objective of our research is to understand the processing strategies of Hebrew comprehenders, and the ways in which they are shaped by the grammar of the language.
We use a variety of methods: self-paced reading, event-related potentials (ERPs), priming, acceptability judgements, and more.
Here are some of the projects that we are currently working on:
Elucidating Human Language Processing with Large Language Models (TAD Center Grant with Prof. Jonathan Berant)
In this project we compare human sentence comprehension to the "comprehension" of large language models, investigating whether processing factors that hinder understanding in humans, for example similarity-based interference, have a similar effect in LLMs.
Mechanisms of encoding and encoding interference during sentence processing (Israel Science Foundation)
This project aims to further our understanding of encoding mechanisms during sentence processing, their susceptibility to interference, and their effect on comprehension.
Encoding and retrieval in sentence processing errors: Comparing Hebrew and English (Binational Science Foundation - National Science Foundation grant)
In this collaboration with Prof. Brian Dillon (University of Massachussets), we investigate the processing of sentences with agreement errors, in Hebrew and in English, in order to understand what causes processing failures – memory encoding or retrieval – and what role the language’s grammar plays in these processes.