Cognitive Diagnostic Modeling of Reading Difficulties in Secondary School Students: Attribute Diagnosis and Profile Analysis within a Reading Literacy Framework

Abstract

This study focused on reading literacy within the broader domain of literacy and conceptualized secondary school students’ perceived reading difficulties in terms of six attributes: decoding, fluency, vocabulary and background knowledge, literal comprehension, inference and integration, and reading anxiety. A self-report scale was analyzed using the generalized DINA (G-DINA) model to estimate attribute-level patterns of perceived reading difficulties and individual reading profiles and to examine their relationships with external criterion variables. Data were obtained from 1,359 middle and high school students across eight regions in Korea. The model showed adequate fit (RMSEA₂= .018, SRMSR = .076), with classification accuracy of .870 at the test level and .962–.986 at the attribute level. Mean EAP-based probabilities of attribute possession were highest for inference and integration (.244), followed by vocabulary and background knowledge (.193), literal comprehension (.169), fluency (.130), decoding (.097), and reading anxiety (.095), indicating relatively greater perceived difficulties in comprehension-related than in word-recognition-related processes. The MAP profile with no perceived difficulty across all six attributes (000000) accounted for 74.0% of the sample. When the MAP profiles were regrouped according to the reading difficulty types proposed by Spear-Swerling (2016), the specific reading comprehension difficulties (SRCD) type accounted for 13.8% and the mixed reading difficulties (MRD) type for 11.1% of the sample. Associations with external criteria varied across attributes: decoding was significantly associated with the dyslexia risk index, vocabulary and background knowledge with listening comprehension difficulties and academic difficulties, and inference and integration with listening preference; reading anxiety was also associated with several cognitive and academic criteria. In contrast, the attributes explained little variance in oral reading time (R² = .026). These findings suggest that applying cognitive diagnostic modeling to self-reported perceptions of reading difficulties can provide screening and diagnostic information about the specific domains and combinations of difficulties perceived by individual students.