A Comparative Study Of Three Different Types Of Cloze Test In English Reading Comprehension Tests

This paper describes a comparative research and study between the threecommon types of cloze tests in college English tests, that is, Multiple-Choice cloze(MC cloze for short), open cloze, and banked cloze. As it is known, MC cloze is acommon testing instrument used in English reading tests in China. However, is MCcloze really effective and valid when reflecting testees' comprehension of the testingpassage? Do these three types of cloze have any difference in the aspect of assessingtestees' reading ability? To answer these questions, the present researcher hasconducted this comparative study.The subjects of this study are seventy-two undergraduate students from threemajors in Tsinghua University. First they are asked to do a cloze test consisting ofthree testing passages in the formats of those three cloze types and a post-testquestionnaire. Their scores of this test are input into computer for statistical analysisthrough Excel and SPSS (version 10.0), a statistical package for social science. Theresults of this analysis are: first, among those three types of cloze, banked cloze hasthe optimal difficulty and differentiate coefficients, while MC cloze has the lowest,which shows MC cloze cannot clearly tell the differences between testees in readingability;second, MC cloze has a much lower correlation coefficient to the total scorethan open cloze and banked cloze, which indicates it cannot present the testee'sreading ability to such an exact extent as open cloze and banked cloze;third, theANOVA test between those three types of cloze and the testees' comprehensiveEnglish grades shows that it is more difficult to tell the students' comprehensiveEnglish level only judging from their MC cloze scores than from their open cloze andbanked scores.The analysis of those testees' answers to the questionnaire indicates that most ofthem do not think that MC cloze can precisely reflect their real comprehension of thetesting passages. This finding has got confirmation in the later interviews of ninereporters from the 72 subjects. These nine reporters' online verbal reports of theirdecoding of the passage and solving process of the questions after the passages showthat in MC cloze test there are more invalid items, whose answer cannot reflect thetestee's decoding of the context since it is gained through some test-taking strategiesrather than the testee's understanding of the passage. Thus the validity of MC clozemust have been impaired a lot. However, in open cloze and banked cloze, althoughthere are still some invalid questions, they do not occur so often as in MC Cloze,because they themselves have some methods to prevent the occurrence of the invalidanswers.Therefore, based on this study, the researcher suggests that test designers shouldchoose appropriate cloze item type in accordance with their different test purpose anduse;and MC cloze should not be used alone in a large-scale proficiency or selectingtest, but combined with other testing instruments.