부산대학교 도서관

This study aimed to conduct a comparative analysis of two distinct variants of a cloze test—namely, the C-test (where the second half of every second word is mutilated) and the X-test (where the first half of every second word is mutilated)— within the context of Korean English as a Foreign Language (EFL) learning. The research sought to explore their potential effectiveness as assessment tools for evaluating English language proficiency. The study intended to investigate several key aspects: firstly, the differences in item facility and item discrimination between the two tests; secondly, the influence of different test types (C-test and X-test) and varying levels of self-assessed learner proficiency (high and low) on both the frequency and success rate of strategies used to complete the blank-filling task; and lastly, how these variables affect the frequency of error patterns observed in failed restorations. To address these questions, the study involved 24 university students, divided into groups based on self-reported proficiency levels, with equal numbers assigned to C-test and X-test papers in each group. The data was collected through think-aloud protocols during the blank-filling test, supplemented by retrospective interviews. The data, based on think-aloud and interview transcription, were categorized into 11 inference strategies (including unidentifiable) and eight error patterns in failed restorations (including unidentifiable). These strategies were divided into two overarching strategies: top-down and bottom-up processes. The main findings were as follows: Firstly, regarding test difficulty, both the C-test and X-test demonstrated similar difficulty levels. However, in terms of discriminative ability, the X-test exhibited stronger discriminatory power compared to the C-test. Secondly, the frequency analysis of blank-filling strategies revealed that contextual inference and syntactic analysis were the most frequently employed strategies among the various approaches. This finding remained consistent across test types and learners’ proficiency levels. Thirdly, the analysis of success rates in blank-filling strategies highlighted that the strategy ‘using phrasal cues’ exhibited the highest overall success rate, while ‘using morphological cues’ showed the lowest. This trend was consistent across test types and learners’ proficiency levels. When ‘using orthographic cues,’ both tests displayed a success rate of approximately fifty percent. Notably, the highly proficient learner group demonstrated slightly higher success rates employing this strategy, whereas the lower proficiency group exhibited the opposite result. Finally, the analysis of failure cases indicated that ‘breakdown in text-processing’ was the most common cause of failure, prevalent in the X-test and among low-proficiency learners. In the C-test, grammar errors were the most frequent, and while highly proficient learners exhibited fewer overall failures, errors resulting from scoring rigidity for synonymy were notably prevalent among them. Lower proficiency learners showed a high frequency of errors overall, with the occurrence of ‘breakdown in text-processing’ alone surpassing the total error frequency seen in higher proficiency learners. The detailed comparative analysis of the frequency and success rates of the strategies, as well as the error patterns observed in failed restorations employed by adult English learners during blank-filling tasks, can potentially serve as fundamental data for developing effective English proficiency tests.