This study replicates and extends Lee-Ellis’s (2009) work by examining how different scoring methods affect the measurement characteristics of a Korean C-test. Thirty-six learners of Korean as a second language completed a five-passage C-test. Their responses were scored using Partial Credit Modeling, dichotomous item-by-item scoring, and dichotomous morpheme- by-morpheme scoring. Rasch analysis was conducted to compare reliability, person separation, item fit, and interpretability across these scoring approaches. The study also treated each text passage as a “super-item” to evaluate proficiency interpretation at the passage level. All three scoring methods demonstrated high reliability (above .97) and yielded highly correlated proficiency estimates. No single scoring method proved significantly superior in reliability, separation, or model fit. At the super-item level, the passages showed a narrow spread of difficulty indicating limited coverage of proficiency levels. These findings suggest that fine-grained scoring alone cannot compensate for limited test scope. To enhance the diagnostic value and level-referenced interpretation of C-tests, careful selection of text passages spanning a broader proficiency spectrum is essential.

I. Introduction
II. Literature Review
    1. C-Test as a Language Proficiency Measure
    2. C-Tests for Korean as a Second Language
III. Method
    1. Participants
    2. Procedure and Task
    3. Analysis
IV. Results and Discussion
    1. Person Ability
    2. Item Difficulty
    3. Separation of Content/Function Morphemes
    4. Super-Item Analysis
    5. Discussion
V. Conclusion
References

• Hyunah Ahn(Kunsan National University)