An Empirical Study Of The Approach To The Validation Of Cloze Tests Based On Redundant Information Theory

Cloze test is a widely used testing means, with its economic characteristics which gets the test developers’ approval, but it is also the subject of huge debate, so the research on cloze tests is not in the minority. This thesis reviews its origin, development, types, information levels and the research status at home and abroad, emphasizes its theoretical base. It focuses on the Information Theory founded by Shannon and others, and its core concept that is the redundant information on the basis of the Information Theory, which elaborates from the language redundancy, its functions, the factors that influence the language redundancy, the forms of redundancy in English and the redundancy in English text. By using the questionnaires and analyzing the data, the thesis is mainly on the empirical study of the approach to the validation of cloze tests based on Redundant Information Theory.This thesis aims to explore the validation of cloze tests of NEMT, CET-4,CET-6, TEM-4 and so on, in order to find whether the information levels of cloze test papers focus on a certain one or not. This research focuses on solving three questions:(1)How is the distribution of multi information levels in cloze tests? (2) How is the distribution of multi information levels in cloze tests in students’ answering process? (3)How do the multi information levels affect the validation of cloze tests? To answer these three questions, this thesis gives an empirical research on the clozes of representative English tests.According to the research questions, this study mainly uses two questionnaires and the method of data statistics. The object of this study is 18 freshmen from a Career Technical College in Hunan,18 graduate students from a Foreign Language College of one university in Hunan, and 3 experts, one is a full-time English teacher from the Career Technical College in Hunan, one is a full-time English teacher in the Foreign Language College of one university in Hunan, and one is a senior high school English teacher. The 18 Career Technical College freshmen are required to complete a questionnaire which contains four cloze tests and the questionnaire questions, chosen from the 2010 NEMT Hunan volume, the 2011 NEMT Shanghai volume, the 2012 NEMT national volume I. The 18 graduate students are required to complete a questionnaire which contains three cloze tests and the questionnaire questions, chosen from the CET-4 volume of December in 2006, the CET-6 volume of June in 2007, and the TEM-4 volume of 2008. The three teachers are required to complete a questionnaire which contains the above 7 cloze tests and the questionnaire questions.To solve these three questions, under the guidance of the Redundant Information Theory, by filling in 7 questionnaires two times and then using SPSS21.0, the author analyzes the information data of each test questions. Among the 115 selected cloze test question, there are 95 questions which can be determined based on a certain information level, and the left 20 questions can be determined based on two or more information levels, which are about 17% of all the selected cloze test questions. In the answering process, the students’choices on the multi information levels are widely distributed, which is more than 50% in each cloze test. The author takes the Section C of 2010 NEMT Hunan volume, which asks the students to fill in the blanks with a suitable word or phrase, as an example, by using the paired samples T test to verify that taking all the information levels at the same time, which are the key of a question answer, into account has a noticeable effect on the validation of cloze tests.We can learn from the thesis that when we use the information levels to test and verify the student’s behaviors in answering process, considering all the information levels which determine the answer’s key information is a more appropriate method. In the inspection process, the students select one of the related levels should be a correct answering behavior and should be accepted. In this way the accuracy of the validation of cloze tests can be improved.Although the number of sampling and selected tests of this study are inadequate, and the reasons about the distribution of the students’choices of multi information levels are not mentioned, but the main achievements combined with the actual situation has a certain depth on the validation of cloze tests. Because of the limited number of the subjects and the limited sample sizes of this study, the conclusion has some limitations, the research results are for reference only.