This interview consists of six questions (see Appendix). The first question aimed to be a relatively easy question to talk about (Robson 2002; Newing in press). The second questio Race Class and Gender n was related to the first question, so that it was able to lead the interviewees to main topic of the interview. This question was one of main questions of this interview as well as the third, fourth, and fifth question. These questions were set to understand Chinese students’ attitudes towards the giant panda. The final question was not directly related to the subject and it was supposed to be a simple question as a “cool-off” question. However, it was found at the development stage of this interview that this sixth question invited the further discussion about the relationship between the giant panda and Chinese people. Therefore, the question was kept in each interview. Procedures This interview survey followed the procedure described by Robson (2002:277); “Int Race Class and Gender roduction, warm-up, main body of interview, cool-off, and closure”. In the introduction stage, interviewers and the students were introduced each other, and talked about their own courses at University of Kent as “warm-up”. During the interview, it is weighted to elicit information to maximize the advantage of a semi-structured interview. Therefore, the depth of answer was varied between the questions and the answerers. It is also important to note that the interview was often stopped to clarify what the interviewee meant or answered. In some cases, it wa Race Class and Gender s confirmed at cool-off stage or after the interview by exchanging e-mail. Analysis During the data collection, the interviewer tried to record annotations, memos, coding (Newing in press). At an initial stage of an analysis, the coding procedure was conducted followed the instruction described by Newing (in press: 218). As top codes, several values from Kellert’s nine values (1996) (see Table 1) were employed as predefined codes. For sub-codes, the detailed information related to the defin Race Class and Gender ed top codes was identified. At next stage, the procedure suggested by Sarantakos (2005) was taken to develop from open-coding to the concept. However, the coding procedure for this interview description was not sufficient for axial, selective coding since top codes us>