The elementary school years are a period when learning across all subject areas is mediated through text, and literacy skills, including decoding and reading comprehension, develop rapidly. Therefore, accurately identifying children’s text comprehen...
The elementary school years are a period when learning across all subject areas is mediated through text, and literacy skills, including decoding and reading comprehension, develop rapidly. Therefore, accurately identifying children’s text comprehension abilities during this period is essential for effectively supporting their subsequent academic progress and overall school adjustment. In actual classroom contexts, however, texts are not presented through a single modality; instead, they appear in multiple formats, such as listening, reading, and combined listening and reading. Moreover, the types of texts that children encounter during this period range widely from literary narratives to expository nonfiction passages. To comprehend such varied texts, children must move beyond extracting surface-level information and engage in processes that integrate multiple cues, draw inferences, and construct a coherent mental representation of the overall meaning. Consequently, the assessment of children’s text comprehension requires an approach that adequately incorporates the multidimensional characteristics of text, including presentation mode, genre, and the cognitive demands of question type.
This study examined text comprehension among elementary school children in Grades 1 through 6 from a multidimensional perspective. A total of 355 children participated in the study, including 53 first graders, 65 second graders, 61 third graders, 66 fourth graders, 57 fifth graders, and 53 sixth graders. Of these participants, 302 were typically developing children and 53 were children with vocabulary delay. Text comprehension was assessed using three presentation modes (listening, listening and reading, reading), two text genres (fiction and nonfiction), and two question types (literal comprehension and inferential comprehension). Because working memory plays a substantial role in text comprehension, participants’ working memory abilities were measured across three domains: phonological working memory, visuospatial working memory, and the episodic buffer. Linear mixed-effects model (LMM) analyses were conducted to examine differences in text comprehension according to presentation mode, genre, and question type across grade levels and groups. To identify which text comprehension conditions significantly discriminated between typically developing children and children with vocabulary delay, hierarchical logistic regression, receiver operating characteristic (ROC) curve analysis, and area under the curve (AUC) analysis were performed. In addition, stepwise multiple regression analyses were conducted separately for the two groups to identify which components of working memory predicted performance under each presentation mode.
First, differences in text comprehension performance across grade levels according to text presentation format, passage type, and question type were analyzed using a linear mixed-effects model (LMM). The findings indicated that students in Grades 2–6 performed significantly better than those in Grade 1, students in Grades 3–6 outperformed those in Grade 2, and Grade 6 students showed significantly higher text comprehension performance than Grade 3 students. In addition, performance under the reading-only condition was significantly lower than under the listening-only and listening-and-reading conditions. Performance on nonfiction texts was significantly lower than on fiction texts, while performance on literal comprehension items was significantly higher than on inferential comprehension items. Performance patterns differed by genre across grade levels. For fiction texts, performance in Grades 2 through 6 exceeded that of Grade 1, and performance in Grades 4 through 6 exceeded that of Grade 2, while no significant differences appeared among Grades 3 through 6. For nonfiction texts, performance in Grades 2 through 6 exceeded that of Grade 1, performance in Grades 3 through 6 exceeded that of Grade 2, and performance in Grades 5 and 6 exceeded that of Grade 3, with no significant differences among Grades 4 through 6. The effect of question type also varied by grade level. For students in Grades 1–4, performance on literal comprehension items was higher than on inferential comprehension items; however, for students in Grades 5 and 6, the difference in performance between the two question types was not statistically significant. Meanwhile, differences in performance across text presentation modes varied by text genre. For fiction texts, performance in the reading-only condition was lower than in the other two conditions (listening and listening and reading), whereas for nonfiction texts, no significant differences were found across text presentation modes. Finally, for both fiction and nonfiction texts, performance on literal comprehension items was higher than on inferential comprehension items, with the difference between the two question types being larger for literary texts. These findings indicate that the development of text comprehension during the elementary years is shaped by interactions among grade level, presentation mode, text genre, and question type. The rapid growth observed in the early grades, the relative weakness in reading-only conditions, the greater difficulty associated with nonfiction texts, and the developmental progression from literal to inferential comprehension align with prior research. Importantly, the present study provides fine-grained evidence using a large sample of Korean elementary school children.
Second, differences in text comprehension performance according to text presentation format, passage type, and question type between the typically developing group and the group with delayed vocabulary development were analyzed using a linear mixed-effects model (LMM). The results showed that the typically developing group performed significantly better on the text comprehension tasks than the group with delayed vocabulary development, and that performance in the reading-only condition was significantly lower than in the listening condition. In addition, performance on fiction texts was higher than on nonfiction texts, and performance on literal comprehension items was higher than on inferential comprehension items. In both fiction and nonfiction texts, performance on literal comprehension items was higher than on inferential comprehension items; however, the difference between the two question types was greater for fiction texts. The finding that the typically developing group demonstrated higher text comprehension performance than the group with delayed vocabulary development confirms that the level of vocabulary development is a key factor influencing overall text comprehension. Notably, in the analysis that included grade level, several interaction effects were significant; however, in the present model that included group as a factor, no interaction effects were statistically significant except for the interaction between passage type and question type. This suggests that the effect of group membership outweighed the effects of presentation mode, text genre, and question type. In other words, group differences according to vocabulary ability accounted for a substantial portion of the variance in performance, reducing the explanatory power of task-condition-specific effects. These findings indicate that the text comprehension difficulties of children with vocabulary delay may not be confined to specific task conditions but instead may reflect consistent challenges across diverse linguistic and cognitive contexts.
Third, to identify the text comprehension conditions that significantly discriminated between typically developing children and children with vocabulary delay, hierarchical logistic regression was conducted using seven text task conditions as independent variables: three presentation modes (listening, listening and reading, reading), two text genres (fiction and nonfiction), and two question types (literal and inferential). All seven conditions significantly increased the explanatory power of the model. ROC and AUC analyses further indicated that the literal comprehension model showed the highest sensitivity, the reading condition showed the highest specificity, and nonfiction texts demonstrated the most balanced classification performance. These findings show that each text comprehension condition serves as a meaningful independent indicator of group differences. The ROC patterns suggest that literal comprehension tasks function effectively as initial screening tools for identifying at-risk children, reading tasks reduce overidentification and selectively detect children with genuine comprehension difficulties, and nonfiction texts provide balanced sensitivity and specificity suitable for educational and clinical decision-making. These findings emphasize the importance of designing assessment and intervention protocols that strategically combine text tasks according to their diagnostic purposes rather than relying on a single task type.
Fourth, the stepwise multiple regression analyses revealed differences between typically developing children and children with vocabulary delay in the working memory components that predicted text comprehension across the three presentation modes. For typically developing children, word list recall predicted listening comprehension; word list recall and nonword repetition predicted listening and reading comprehension; and both matrix tasks and word list recall predicted reading comprehension. For children with vocabulary delay, only word list recall predicted listening comprehension, and no working memory variables significantly predicted performance in the listening and reading condition. These findings demonstrate that different components of working memory contribute to text comprehension in the two groups. The consistent predictive role of word list recall underscores the importance of temporarily maintaining meaning-based information and integrating it with related knowledge retrieved from long-term memory during sentence- and discourse-level processing. The absence of significant predictors for the listening and reading condition among children with vocabulary delay suggests that the dual-modality input may increase processing load, limiting the efficient deployment of working memory resources.
Overall, this study demonstrates that coordination among working memory subsystems plays a critical role in text comprehension and that developmental differences in such coordination help explain the performance gap between typically developing children and children with vocabulary delay. Text comprehension difficulties arise not merely from limited vocabulary knowledge but also from constraints in higher-level processes such as understanding text structure, constructing mental imagery, organizing temporal and spatial information, and engaging in causal reasoning. The study further shows that text comprehension is a multidimensional construct shaped by interactions among grade level, linguistic proficiency, task characteristics, and working memory resources. Moreover, children employ different cognitive strategies depending on presentation mode, text genre, question type, and their working memory profiles, confirming that text comprehension represents a higher-order literacy skill that extends beyond vocabulary. These findings provide empirical support for expanding text comprehension assessment for elementary school children beyond single total scores and toward multidimensional assessment frameworks that consider performance across conditions, text types, and question demands.