This study investigates the impact of word predictability on processing difficulty in different reading contexts, including information seeking and repeated processing. While general predictions about the relationship between word predictability and processing time hold, specific contexts reveal discrepancies with surprisal theory, indicating potential misalignments between human cognitive mechanisms and existing language models.

This study explores whether eye movements during reading can effectively predict reading comprehension, utilizing large-scale eyetracking data and focusing on single-question assessments from text passages. The findings indicate that while the task is challenging, eye movements provide valuable signals for predicting comprehension, with evaluations conducted across various models and reading contexts, and all data and code will be made publicly available.

This study investigates how eye movements differ between first and repeated readings in L1 English readers, focusing on linguistic properties and various reading contexts. The findings reveal significant facilitation effects during repeated reading, influenced by reading regime, intervening text, and task relevance, while also highlighting individual differences related to reading speed but not proficiency, offering insights for future models of reading eye movements.

This study proposes an incremental language model that better captures how the brain processes auditory context over long timescales by summarizing information incrementally, matching neural activities in the neural networks.