A Study Of The Reliability Of The Different Types Of Questions In Semi Direct Spoken Tests For Intermediate Level Students

Chinese spoken tests, as stylized standards to distinguish the learners’levels, are becoming more and more colorful and systematic. Except the HSK which is most famous as a kind of mandarin test, a lot of spoken Chinese tests have appeared recently, such as BCT oral test, spoken Chinese exams for primary and secondary schools, and so on. The diversity of test content, the variety of examination methods and the materialization of the test objects have become the development tendency. Developing the more reliable oral tests have become the hot issues. How to make oral tests more reliable is a very popular question and this is very meaningful to mandarin tests and mandarin education, how to make those tests more widely popular is also worth discussing in mandarin tests.The test questions, which are more reliable and more practical can promote the effective development of mandarin tests. Besides that, those test questions also can offer the students more typical test content, propelling their studies.Basing on the previous period data of text corpus in Nanjing university,this text classify all the questions according to the question types (narrative /descriptive/illustrative), we use the basic ideas and the basic methods of Generalizability Theory to measure the reliability of the text corpus and the reliability of the different types of the questions. The objects all belong to the immediate level. Through the demonstration and comparison of the number, we will give the suggestions to design the questions in questionnaires or in a text corpus.This content of this paper is divided into two main parts, the first part includes the reliability of the whole corpus and the reliability of the three types of questions, and the first part quantifies the scores which the students get and designs the experiments according to Generalizability Theory, including a two-facet crossed design and three single-facet crossed design. The result we get is that the reliability of the whole corpus is very high. And among the three types questions, the descriptive type is the most reliable, the second one is the illustrative type and the last one is the narrative type. We discover that though we have made the specific rules, the interaction between raters and students is still very obvious. When the number of the raters reached 4, we cannot only get the high reliability and also save the costs. So, we suggest that, when we test the oral Chinese ability of the intermediate level students, the number of the raters should be 4.The second part includes the questionnaires about the interviews between the learners and the teachers, who participated in the corpus, mainly talking about the questions in the test subjectively, we will test their tendency of the question types,and then analyzing and summarizing, giving some specific suggestions from their perspectives. The specific process is divided into four parts. Firstly we design the questionnaires according to previous studies, secondly we filter the objects according to current situations, and thirdly we start to interview the students one by one and record the content of the interviews. Last but not the least, we calculate the percentage of those data and try to analyze them.The result we get is most students tend to choose the narrative type and the direct oral tests.The two different parts have been analyzed, we discover that there are some different results we get from the two different parts and we have tried to detect the reason which caused the difference. From the perspective of learning and teaching, we analyzed the difference between the two parts and tried to give the feasible features of the different types of the questions, which are description and complex. Because the reliability of the descriptive type and the whole corpus is higher. In the last part, this paper give some suggestions in oral tests and oral teaching. For example, the learners should pay more attention in the different types of the questions and the teachers should improve the way of testing and try to increase the quantity of the questions. It is believed that this research might help the later investigations.