A Study On The Construct Validity Of The CET Spoken English Test

This thesis attempts to explore the construct validity of the CET-SET by examining the match between test-takers'intended performance and their actual performance from the perspective of discourse analysis. The analysis only focuses on the group discussion part, because it is the very part that is designed to engage candidates in interactive communication and measure their interactive speaking ability. The key research questions were put forward: (1) What is the overall distribution of interactional language functions (ILFs)? (2) Are there any differences in the characteristics of the use of each type of ILFs across test-takers'proficiency levels?Based on the previous studies on ILFs and the language functions listed in the CET Spoken English Test Syllabus and Sample Test (1999), an operational checklist of eight ILFs representing the construct of spoken language ability was developed to serve as the basis for this study. 20 group discussions of 60 candidates taken from November, 2007 administration of the test were recorded and transcribed, and finally analyzed for the ILFs employed by the candidates.The major findings of the present study are briefly summarized as follows: First, as to the overall distribution of the eight ILFs, Asking for opinions or information, (Dis)agreeing and Negotiating meaning are the top three most frequently elicited ones, each of the remaining five– Challenging, Supporting, Modifying, Persuading and Developing– accounts for a very low percentage. Consequently, the inadequate elicitation of ILFs from candidates threatens the construct validity of the CET-SET. Candidates'interpretation of the group discussion task as an assessment event is considered to be the possible reason for this low degree of interaction.Second, there do exist some differences in the use of ILFs across test-takers'proficiency levels. Higher proficiency test-takers seem to apply more ILFs than lower proficiency test-takers. Moreover, of the eight types of ILFs, higher proficiency test-takers use the functions of Challenging, Modifying and Negotiating Meaning more frequently than lower proficiency ones, while the latter uses (Dis)agreeing more often. These results point to the importance of considering functional competence when assessing speaking ability.