Developing a hierarchical cognitive second language (L2) reading model using multi-dimensional fuzzy Delphi method and interpretive structural modeling

Many secondary students learn to read academic texts in a non-native language. However, classroom guidance and diagnostic tests often rely on flat skills lists that ignore prerequisite relations among reading processes. This study seeks to (i) identify the core cognitive attributes underlying lower-secondary second language (L2) reading comprehension and (ii) construct a hierarchical model of those attributes. A two-stage expert-elicitation pipeline was implemented with 12 experts. An initial set of 15 candidate attributes was screened by the multidimensional Fuzzy Delphi Method (FDM) for Importance and Relevance, while Measurability and Specificity were applied qualitatively to refine wording and merge overlaps. Findings from interpretive structural modeling yield a four-level hierarchy. Level IV consists of C1 Recognizing word meaning and C2 Retrieving explicitly stated information, Level III (C3 Identifying main ideas and C4 Distinguishing facts from opinions); Level II (C5 Making inferences), and Level I (C6 Drawing generalizations). MICMAC classed C1/C2 as low-dependence, C3/C4 as linkage and C5/C6 as dependent. Implications include hierarchically constrained Q-matrix design for CDMs, load-aware instruction sequencing (Level IV → I), and item blueprinting that tags task features to prevent attribute leakage. The study contributes to the existing body of knowledge by proposing a theory-aligned, expert-derived hierarchical model of ESL reading attributes and providing actionable structures that can inform Q-matrix design, item generation, and instructional decisions. Limitations include a small, context-bound expert panel, inherent subjectivity in expert judgments, and the absence of learner-data validation. We recommend broadening the expert pool, validating the model with learner data, and processing complementary data.