Comparative evaluation of C-test reliability using classical and modern psychometric methods

Abstract

This study offers a comparative analysis of classical reliability measures, including Cronbach’s alpha, test-retest, and parallel forms reliability, alongside modern psychometric methods such as the Rasch model and Mokken scaling, to evaluate the reliability of C-tests in language proficiency assessment. Utilizing data from 150 participants across two testing sessions, the results indicate high internal consistency (Cronbach’s α = .91, .88), strong parallel forms reliability (r = .82), and moderate test-retest reliability (r = .76). Rasch analysis provided nuanced insights into item difficulty and person reliability, while Mokken scaling confirmed a reliable ordinal hierarchy among test items. This combined methodological approach offers a more comprehensive view of C-test performance and its psychometric properties. Future studies should investigate scalability across diverse linguistic contexts to enhance the test’s generalizability and practical applicability.