The traditional listening comprehension part of College English Test is mainly composed of short conversation and monologue. In this paper, the author questioned the construct validity of listening comprehension part of CET and suggested that the Longer Conversation could make the listening test more valid. The author put forward her hypotheses and designed tests to verify them. Two text types, the Longer Conversation and the Talk of the same topic were compared. The Longer Conversation contained more oral features compared with the Talk. The formal experiment was administered to 110 EFL learners, who were then divided into two matched groups. In each group, test takers listened to the different version of each topic and answered the same questions. The results of these two groups were compared and analyzed. Results indicated that the text type did affect the performance of the test takers. It was found that the Longer Conversation was comparatively more difficult than the talk and had comparatively higher item discrimination index. In order to make the listening comprehension of College English Test more valid, the author recommend that listening comprehension test of College English Test should include the Longer Conversation.