How narrative content supports reading immersion: an empirical analysis of digital reading traces of detective fiction

Purpose: This study examines how narrative content in detective fiction is associated with indicators of reading immersion, using narrative transportation theory to connect textual topics with readers’ emotional and cognitive responses. Design/methodology/approach: The study analyses 27,999 reader-selected excerpts from 100 highly ranked detective novels on Douban, posted between February 2011 and January 2024. Latent Dirichlet allocation (LDA) was used to identify narrative topics, and a BERT classifier assigned excerpts to the resulting categories. Sentiment dictionaries and the Text Mind system generated emotional and cognitive indicators. Non-parametric tests examined temporal trends and differences across subtopics. Findings: The excerpts formed three narrative dimensions (plot, humanity, and society) and seven subtopics. Plot-related content received consistently high attention; attention to humanity-level content declined over time, whereas attention to society-level content increased. Positive and negative sentiment indicators both rose, and the seven subtopics showed distinct emotional and cognitive profiles. These patterns inform a three-stage interpretive model linking content accessibility, emotional and cognitive engagement, and intellectual extension. Originality: The study moves beyond reviews and self-reports by treating reader-selected excerpts as digital traces of textual salience and by integrating topic modelling, supervised classification, sentiment analysis, and cognitive-semantic analysis. Contribution to the field of digital humanities: It provides a content-sensitive framework and a reproducible computational workflow for investigating how literary form, reader selection, and interpretive response interact at scale.