Predictive inference in reading comprehension: What can biased judgments inform us about information processing in the brain?

How is information processed in the brain? The answer to this question is still far from being resolved. Many of the elements involved in information processing are known but understanding the whole process is a complex problem. In this work, we attempt to uncover some facets of this complexity in a process unique to human behavior, namely reading comprehension. Reading comprehension, an important task in daily life, requires the engagement of bottom-up and top-down processes in a dynamical way to speed-up processing of incoming information and make decisions about current and future expectations about the meaning of the text. Studies on lexical or syntactic ambiguity resolutions have shown that in these instances a slowing down in information integration is measured, while information processing is faster when reading underspecified text. This behavior points to different levels of information integration by the comprehender that are influenced by the complexity of the text and the time demands (i.e., memory constraints) to efficiently integrate new information that is consistent with current text and world knowledge of the reader. In this work, we conduct self-paced reading experiments that address how the various types of these behaviors interact in a single text. The text is prepared in such a way to include gender ambiguities of two characters in the story, and local information about one that affects the processing of future information about the second character. This provides a window on how information is being reorganized in the brain between local and global information about the text. It is shown how prediction and inference in reading comprehension interact with each other following a principle of least effort driven by the lowering of entropy in the working memory of the reader. The studies are designed to explore how learning is being manipulated by working memory to reach a coherent picture of the whole text. Therefore, the interaction can also be understood in terms of top-down versus bottom-up processes in the brain. Gender bias is exploited as a tool in these experiments to untangle the various processes involved in a task as simple as understanding a short paragraph. Bias, which is stored in long-term memory where beliefs reside, can be reversed by the context of the acquired knowledge. This shows that short-term learning is very flexible. This latter process is an instance of error correction coding.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Educational Psychology

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2020

Author dc:creator

• Rebei, Adnan

Contributors dc:contributor

• Mestre, Jose
• Christianson, Kiel
• Dell, Gary
• Hummel, John

• psycholinguistics, inference, information, brain

Statement dc:rights

• Copyright 2020 Adnan Rebei

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/108100

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/108100