Processes Of Taking Two Cloze Tests-Comparison Between Banked MC Cloze And Separate Item MC Cloze

The current study compares the two popular cloze tests in domestic language testing, namely, separate Item MC cloze (shorten as MC cloze) and banked MC Cloze (shorten as banked cloze). Through both quantitative and qualitative methods, the study is aimed at figuring out whether the two cloze tests with the same passage and same items tested can measure the same language ability. Answering such a problem can further provide evidence for the long running argument of what cloze really tests.73 sophomores took part in the two cloze paper tests and 2 joined the think-aloud pilot study. Then another 10 students were asked to do the think-aloud experiment. After the experiments, both quantitative and qualitative data were analyzed. The statistical results showed the significant difference between the two cloze test scores and test-takers were shown to perform better in MC cloze than in banked cloze. Item analysis of the two tests, however, further indicates MC cloze is easier and less discriminative than banked cloze. Thus, from the view of quantitative data result, banked cloze appears to be more effective in testing language ability than MC cloze.The analysis of think-aloud protocol, however, showed much similarity in test-taking processes of the two cloze tests: firstly, during the test-taking process, test-takers in both cloze procedures underwent three stages: meaning-decoding stage, choice-making stage and evaluation stage; secondly, test-takers were found to use much more within-sentence information in response to items. Among the information used by test-takers, clause-level information was found to be used most frequently; cross-sentence information was comparatively less in number and cross-text information, the least. Therefore, considering the information categories, the two cloze tests mainly measure lower-level language ability rather than higher-level language ability. Thirdly, test-taking strategies used by test-takers revealed that test-takers used some metacognitive strategies in answers to items. Besides, the use of test-wiseness strategies by test-takers, to some extent, posed threat to the validity of cloze tests.In addition, qualitative data also showed the differences between the two cloze test-taking processes. Test-takers were found to experience more complex cognitive process, use more information and take more strategies in figuring out answers. All these differences partially explain the reason why MC cloze is statistically shown to be easier and less discriminative than banked cloze.Based on the above results, the following suggestions are generalized as references for language test designers:First, it is advisable that test designers choose different cloze tests for different testing purposes; second, test designers might as well focus more attention on context-level information and design more items measuring higher-level rather than lower-level language ability so as to guarantee content validity of cloze tests; last but not least, besides quantitative analysis, the cognitive test-taking process from test-takers can also be regarded as an effective way to further validate cloze procedure, thereby helping design cloze tests with high reliability and validity.