Concept inventories in astronomy are typically organized around curriculum-defined content domains, on the implicit assumption that conceptual understanding partitions along those same lines. This assumption has rarely been examined empirically, in part because probing it requires an instrument both aligned with a specific curriculum and broad enough to capture cross-domain structure. We developed the Astronomy Curriculum-aligned Diagnostic Assessment (AcuDA), a 26-item instrument aligned with the 2022 revised Korean science curriculum and adapted from six established concept inventories, and administered it to 735 Korean university students (114 astronomy majors, 621 non-majors). AcuDA showed strong internal consistency (α = .844, ωtotal = .928). Factor analysis distinguished four content domains whose latent correlations ranged from .86 to .94. Given these high correlations, bifactor modeling, which separates a general factor from domain-specific factors, showed the general factor dominated (ωh= .895, ECV = .830). Cross-group invariance supported comparability between majors and non-majors, while distractor patterns revealed group-specific misconceptions on shared items. Because learner responses were organized primarily along a single latent dimension rather than four separable proficiencies, a single total score provides more interpretable and useful information than separate domain scores. Beyond providing the first astronomy concept inventory developed within and validated against the Korean national science curriculum, these results offer initial evidence that astronomy understanding may not partition cleanly along curriculum-defined content domains, and that its underlying cognitive architecture is better characterized as a shared reasoning dimension than as a set of separable domain abilities.

Introduction
Methods
    Validation Sample and Data Collection
    Instrument: AcuDA and Curriculum-Domain Mapping
    Content Validity: Expert Review and Translation Adaptation
    Scoring and Analytical Strategy
        Scoring and Data Preparation
        Analytical Strategy
        Software
Results
    Descriptive Statistics and Classical Item Functioning
    Multidimensional Latent Structure: 4-Factor vs. Bifactor Models
        Correlated Four-Factor MIRT Model
        Bifactor MIRT Model
        Sensitivity to Domain Assignment
    Stability of the General Factor (Bootstrap)
    Group-Specific Bifactor Structure
    Known-Groups Performance Profiles
    Measurement Invariance across Groups
    Diagnostic Distractor Analysis
Discussion
    Curriculum Domains and the Latent Structure of Astronomy Understanding
    Interpreting Total vs. Domain Scores
    Misconceptions as Structured Knowledge: Qualitative Differences between Groups
    Expertise and the Emergence of Domain Differentiation
    Measurement Invariance and Item-Level Equivalence
    Implications for Assessment, Curriculum, and Instruction
    Limitations
    Future Validation and Extension to Secondary Education
Conclusion
References
AcuDA Item Texts (English Reference Version)

• Gilgu Kang(Department of Astronomy and Space Science, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea)
• Ahra Cho(Department of Astronomy and Space Science, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea)
• Yonggi Kim(Department of Astronomy and Space Science, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea, Chungbuk Pro Maker Center, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea)
• Jiwon Park(Department of Astronomy and Space Science, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea, Chungbuk Pro Maker Center, Chungbuk National University, Cheongju, Chungbuk 28644, Republic of Korea) Corresponding author