Document Type : Research Paper

Authors

• Hamed Zakeri ¹
• Sahar Bahrami-Khorshid ²
• Maryam Danaye Tous ³
• Masoud Ghayoomi ⁴

¹ Department of Linguistics, Faculty of Humanities, Tarbiat Modares University, Tehran, Iran

² Assistant Professor of Linguistics, TMU

³ Department of English Language and Literature, Faculty of Literature and Humanities, University of Guilan, Rasht, Iran

⁴ Department of Dialectology, Institute for Humanities and Cultural Studies, Tehran, Iran

Abstract

Text comprehension and recall are among the most fundamental cognitive processes, and explaining their mechanisms can be effective both in the development of cognitive science theories and in the design of intelligent computational systems. Written texts generally fall into two main categories - narrative and expository - each activating different patterns of mental processing: narratives, relying on temporality, causality, and character-centeredness, are closely related to episodic memory, while expository texts, with their logical and conceptual organization, are more dependent on semantic memory. The present study, with the aim of examining these two text types simultaneously, empirically tested two well-known models of text understanding - the Landscape model and the Event-Indexing model - and, alongside them, evaluated the effectiveness of a revised model designed on the basis of integrating conceptual activation, event structuring, and positional weighting. For this purpose, three Persian texts with different structures were selected. Fifty-one ninth-grade students, after reading the texts, participated in a free recall task, and their memory outputs were recorded and coded sentence by sentence. The three cognitive models were then implemented, and their normalized outputs (0–1) were compared with the actual data of the participants. Statistical analyses included accuracy, precision, recall, and F1 score, along with Pearson correlation and paired t-tests. The findings showed that the Landscape model was more effective in representing key concepts and semantic relations, whereas the Event-Indexing model better captured the temporal and causal coherence of narrative texts. However, neither model alone was sufficient to cover the full pattern of human recall. In contrast, the revised model, by integrating the three cognitive mechanisms and incorporating the serial position effect, provided the highest level of correlation with behavioral data and demonstrated a significant advantage over the two classical models in terms of the F1 measure. These results indicate that employing multilayered cognitive mechanisms can offer a more accurate and realistic prediction of textual memory. Beyond its theoretical importance, this outcome also opens new perspectives for designing adaptive educational tools, cognitive rehabilitation programs, and natural language processing systems grounded in human memory patterns.

Keywords

• Text recall
• Landscape model
• Event-Indexing model
• Revised model
• Computational text analysis

Main Subjects

• Click subject name to see related articles.