A Comparative Study On Three Cloze Variants Of Rational Cloze, Multiple-Choice Cloze And C-Test

Cloze, adopted wisely in large scale language assessment and classroom exercise in recent years, usually harbors five variants:fixed-ratio cloze, rational cloze (open cloze or gap-filling), multiple-choice cloze, c-test, and banked cloze. A lot of researches have been conducted in this field at home and abroad, whereas they rarely touch upon comparative study on multiple variants of cloze. Meanwhile, most of them are product-oriented rather than process-oriented. Involving rational cloze, multiple-choice cloze, and c-test, and combining the product-oriented with the process-oriented study, this thesis obtains its own necessity and practical application to some extent.In this study, the experiment is conducted by requiring the same sophomores majored in English from the Faculty of International Studies of Henan Normal University to take different variants of cloze test recomposed from multiple-choice cloze component in TEM-4 held in April,2009. Due to precluding the diversity on test-takers' language ability, the differences in the test results demonstrate the existence of distinction among the three test papers. This thesis mainly probes into what significant differences exist in three test results, what significant distinctions are at difficulty and discrimination level among the three variants of cloze test, and what sorts of information and how many information sources are adopted by test-takers in the process of taking each test respectively.The result of statistical analysis suggests that test-takers perform best in multiple-choice cloze but worst in rational cloze, for the former is the easiest yet the latter the most difficult test format, and c-test is able to distinguish between test-takers to the largest extent. Test-point level analysis and the analysis on the status of adopting information source indicate that the test-point level is highly related to test validity. Therefore, the item with the test-point in higher-level requires test-takers to adopt more information sources, which complicates the inferential process to enhance the test validity.This study can be of help for test-makers to choose the appropriate test format according to the corresponding test objective and environment, which increases the diversity of cloze test task to refrain from negative effect owing to test-driven language teaching and thus enhance the test validity.