A key challenge in the design of effective survey questionnaires is to write questions that respondents can understand consistently. Recommendations in the questionnaire design literature propose the use of respondents' own terminology to facilitate comprehension and the response process. In this paper, we propose an innovative questionnaire development method to construct items using respondents' own terminology. On the basis of statistical analyses of answers to an open-ended question asked in a survey measuring attitudes to biodiversity (N = 351), we first identified the words that respondents associate most frequently with the concept of 'biodiversity'. We then designed new attitude measures composed of the words identified as being central to representations about biodiversity. These items were tested in another survey using a different sample (N = 467). The results show that the attitude items designed on the basis of a social representation method satisfied validity and reliability quality criteria.