A Comparative Study On Banked Cloze And Multiple-choice Cloze Test Formats

The present study explores the test takers’ mental processes of taking banked cloze test format （BCfor short）and multiple-choice cloze test format （MC for short） in an attempt to examine what such testsactually examine and how does the different test format influence the test takers’ test-taking process. Thekey research questions are put forwards as follows:1).What is Facility Value (F.V. for short)&Discrimination Index (D.I. for short) of BC and MC respectively? What is the relationship between the testtakers’ performance and them?2).What information sources do the test takers employ when they areanswering the items in BC and in MC? Do those information sources influence their test-taking processes?If so, how do they influence? To answer the questions above, simultaneous introspection and immediateretrospection are adopted and forty-two junior middle school students take part in the procedure.Based on the previous studies, the researcher mainly adopts a classification framework proposed byBachman (1985) who classifies the redundancy information into four levels in the cloze test-takingprocesses. They are:1) within clause;2) across clause, within sentence;3) across sentence, within text;4)extra-textual. Considering the practical instances, the thinking aloud data in this study were categorized asphrase-level information, sentence-level information, text-level information and guessing-level information.According to this classification, the test takers’ thinking aloud data are future categorized and analyzed.The major findings of this study are briefly concluded as follows:First, the Mean F.V. of BC is higher than that of MC, which means that BC has more difficulty levelamong the items. And the Mean D.I. of BC is higher which implies that BC can distinguish between testtakers at different levels better. Thus the same test takers take the two test formats will get different scoresand they will perform better in MC. And the medium level will be most popular.Second, four classification of information sources are identified in the thinking aloud data of the testtakers, namely phrase-level information, sentence-level information, text-level information andguessing-level information. Due to the conjugation of simultaneous retrospection and immediateretrospection methods, though some students do not say anything during the simultaneous retrospection phase, they are instructed to give some additional information during the immediate retrospection phase. Ofthese four information sources, the sentence-level information accounts for the largest ratio, followed byphrase-level information, text-level information, and guessing-level information in BC. While in MC, thesentence-level information accounts for the largest ratio, followed by phrase-level information,guessing-level information, and text-level information.