Abstract
<p>Previous evaluations of school textbooks have predominantly focused on aggregate overall averages, thereby overlooking critical intra-textbook heterogeneity. This study aimed to identify latent profiles of cognitive involvement and assess the instability of instructional design within Iranian senior high school sociology textbooks. Utilizing a secondary data analysis of datasets derived from the Comparative Adjudication Protocol (CAP)—encompassing 51 lessons across 4 core textbooks (Sociology 1, Sociology 2, Sociology 3, and Civic Identity)—cognitive involvement indices for text, images, and questions were calculated using the William Romey (1968) method. To quantify internal heterogeneity, the Coefficient of Variation (CV) and Levene's test were utilized, while K-means clustering with Euclidean distance (Hair et al., 2010) was employed to uncover latent profiles.The findings revealed that the Sociology 2 textbook exhibited the highest textual irregularity (CV = 122.45%), whereas the Civic Identity textbook possessed the most uniform text structure (CV = 36.52%). Levene's test confirmed no statistically significant difference in variances across the textbooks for any component (p &gt; 0.05), indicating a more structurally homogeneous design across sociology textbooks compared to history textbooks. However, cluster analysis uncovered three distinct structural profiles, revealing that 35.3% of the analyzed lessons fell into a "fully passive" cluster, while 21.6% were classified as "hyper-active." These findings demonstrate that while sociology textbooks exhibit lower overall heterogeneity than history textbooks, a persistent question-text disjuncture remains—where end-of-chapter tasks demand high analytical engagement while the textual content remains predominantly passive. Consequently, future curricular interventions must prioritize structural alignment between narrative content and assessment tasks, alongside variance reduction across grade levels.</p>