The Test Constructor As A Factor Affecting The Validity Of The Grammar Gap-Filling Test

This study was conducted to explore the effect of test constructors on the validity of grammar gap-filling(G/F) test. Two test constructors were invited to write items for the test and 70 Senior Two students participated in the experiment. The two constructors have taught English for many years. One is a professor in a famous university and the other is a high school English teacher. The 70 students come from two parallel classes in a key high school in Guangzhou. Before the experiment, the two constructors composed items on the basis of the identical passage for Test 1 and Test 2 respectively. After that, an interview was conducted to elicit related information underlying the item construction. A proficiency test was first held to obtain the grammar proficiency of the two classes. The result demonstrates that no significant difference was revealed in the performance of the two classes, which makes sure that the two classes have the same level of grammar ability. In real experiment period Test1 and Test 2 were administered to Class 1 and Class 2 respectively. The result shows that there is a significant difference in test-takers' performance in Test 1 and Test 2.The major findings of the study are:(1) The proficiency test displays that the two classes have no significant difference in grammar ability, while their performance differs significantly in Test 1 and Test 2. According to the analysis of the items, it is surmised that the items constructed by different constructors lead to the score difference of the two classes, which suggests that test constructors may be a factor affecting test-takers' performance;(2) Constructors with different knowledge and experience have different understanding about the construct of grammar and grammar test and they hold different general principles of designing items. For this reason, they designed items with different item difficulty and distinct distribution of components of grammatical competence, which hence influences the interpretation of the score of the test. In this sense, it would affect the validity of the test. Besides, test constructors may introduce a test method effect because they would design items according to their personal idiosyncratic interpretation of the test, which would also undermine the validity of the test.The findings indicate that the test constructor effect should be paid much attention to in test designing. It is necessary to train test constructors and prepare them with a clear understanding about the construct to be measured before item construction, thus to ensure the validity of the test.