Readability Analysis On CEPT Reading Texts And The Development Of Lexical Checker

Reading test, playing a leading role in language testing, is of vital importance in item bank construction. Generally speaking, there are two variables influencing reading test: reading texts and readers. Reading comprehension study tends to focus on readers'factor in recent years, however, it is also meaningful to study the reading text itself.One important step in reading test is to select proper reading texts according to the exam outline. It is necessary to develop a scientific and effective method to simplify the process of reading material selection. According to practical application, the project team of College English Placement Test in Hunan University has developed Lexical Checker, which is a small computer program mainly used to check word frequency of a reading text.Readability test is a common used reading text test in English speaking countries. It applies scientific formula to quantize the linguistic features of a text and then provides its readability index. This paper focuses on the study of readability measurement and its application in language testing: trying to find out the most proper readability test for CEPT reading texts as well as the correlation between readability data of a text and the facility value, establish a linear regression equation and then apply it into Lexical Checker in order to effectively improve the program and widen its application in material selection for reading test.The research materials are mainly 150 reading texts from the item bank of CEPT, research data includes two parts: one is readability data of the reading texts, which is obtained through six different but common-used readability formulas; the other is data of facility values obtained through 2008 College English Placement Test in Hunan University, which involved 4794 test takers. The correlations among above data are analyzed by SPSS 13.0. Readability data is taken as independent variable while facility value as dependent variable to see if linear regression equations can be established.Through empirical study of data collection and analysis, the primary research results show that: (1) the six different readability formulas used in this study are highly correlated and can provide relatively accurate readability information of a text, among which Flesch-Kincaid Grade Level are considered as the most reliable and proper one for CEPT reading texts; (2) readability level or index of a text is also correlated with the facility value if the text is used for a reading test according to a certain exam outline; (3) linear regression equation modal can be set up to forecast facility value if readability information of a text is provided.All above research outcome from readability test can be taken as the foundation to improve Lexical Checker so as to further extend its application: reading texts can be selected or revised after the obtaining of readability data in order to control the difficulty of the texts. Thus, the facility values of the texts used for reading comprehension in one section are close to each other, which would make the item bank more scientific and fair. To apply readability test and linear regression equations can not only meet the requirement of CEPT to justly distribute test contents with different difficulty levels but also highly promote the practical application of readability test and Lexical Checker in language testing.