| Summary: | The Vocabulary Levels Test (VLT) was constructed by Paul Nation in 1983, and was expanded and revised by Schmitt (2000) and Schmitt, Schmitt, & Clapham (2001). The test has been used very widely. In spite of that, the central issues of the VLT have not been satisfactorilya ddressedt;h ey all derive from its matching-formatc onsisting of six words and three definitions as a cluster. Those issues are concerned with local item dependence within a cluster, the relationship between target words and distractors, and guessing. The purpose of the thesisw was to reappraise the VLT by addressing them. A series of experiments were conducted for that purpose. The quantitative and qualitative empirical evidence shows that items within a cluster cannot be regarded as independent. The findings also show that distractors cannot play the role of target words. Protocol analyses indicate that learners, regardless of the proficiency level, gain some scores through a variety of guessing strategies, implying that the VLT has the tendency to overestimate their vocabulary knowledge. An attempt was carried out to control guessing by confidence rating and Shizuka's (2000) Clustered Objective Probability Scoring (COPS) method. The results claim that the proposed correction procedure looks extremely promising for the measurement of test-takers' true vocabulary knowledge. The thesis suggests that if the VLT adopts this correction procedure, the test will continue its role as a vocabulary measurement tool in the foreseeable future.
|
|---|
