The thesis1 is concerned with a study of the oral component of PETS Level 3 using a Conversation Analysis (CA) approach. To be specific, the study focuses on the Pair Discussion of Spoken English Test for PETS Level 3 (the PETS(e)-SET Pair Discussion) and it has two broad purposes. First, it adopts a CA approach to analyze the candidates oral performance in the Pair Discussion to see how well they are able to apply the three types of interactional organizations-turn taking organization,sequence organization and repair organization in spoken interaction under the test conditions. Second, the study further investigates whether the test facilitates conditions for eliciting legitimate samples of candidates oral performance in terms of the tests construct validity. Although this study is preliminary in nature, it is data driven and significant in that it provides empirical evidences which help to reveal problems in candidates oral performance and underline the need for further improvement of the test. The research findings show that the candidates do not know how to appropriately take part in spoken interactions and their knowledge of turn taking, adjacency pairs, repair devices and turn taking strategies is far from satisfactory. The study also shows that poor administration and some test-method errors impair the construct validity of the test. The findings of the study have important implications for the improvement of the test design and the teaching and learning of oral English as well.