Eye-tracking reveals the cognitive processes underlying the construction of mental models while reading. While there are machine learning models that predict comprehension using eye movements from entire texts, the specific contribution of theoretically critical text spans remains unclear. This study examines whether eye-movement patterns on these segments reflect successful information encoding and integration into coherent situation models when readers anticipate, but do not know, subsequent comprehension questions. This study identifies which eye-movement measures on critical text spans best predict comprehension. It examines if there is a difference in predictive relevance among different eye movement measures depending on the mental model required to answer specific question types. One hundred participants are recruited at Saarland University in Germany. Eligible participants are adult native speakers of German with normal or corrected-to-normal vision. Individuals may have one or more first languages. This study examines reading comprehension using eye-tracking during natural, silent reading of texts. Employing a within-subjects design, comprehension question type is manipulated across three levels (local, bridging, and global), each targeting a distinct level of text representation as conceptualized by Kintsch (1998). Critical text spans were annotated for each question.
Brasser et al. (Wed,) studied this question.