Cognitive interviewing: verbal data in the design and pretesting of questionnaires.

PURPOSE: The purpose of this paper is to discuss problems that occur in questionnaire responses and how cognitive interviewing can be used to identify problematic questions prior to using the questionnaire in the field.
BACKGROUND: Questionnaire design involves developing wording that is clear, unambiguous and permits respondents successfully to answer the question that is asked. However, a number of problems in relation to respondents' understanding and successfully completing questionnaires have been identified. Cognitive interviewing, an amalgamation of cognitive psychology and survey methodology, has been developed to identify problematic questions that may elicit response error. The overall aim is to use cognitive theory to understand how respondents perceive and interpret questions and to identify potential problems that may arise in prospective survey questionnaires.
METHODS: A literature review is used to examine the process of questionnaire design and how cognitive interviewing can be used to reduce sampling error and increase questionnaire response rates.
FINDINGS: Cognitive interviewing involves interviewers asking survey respondents to think out loud as they go through a survey questionnaire and tell them everything they are thinking. This allows understanding of the questionnaire from the respondents' perspective rather than that of the researchers. Cognitive interviews have been used in a number of areas in health care research to pretest and validate questionnaires and to ensure high response rates. Interviewing has been found to be highly effective in developing questionnaires for age specific groups (children and adolescents) and in ascertaining respondents' understanding in health surveys prior to distribution. However, cognitive interviews have been criticized for being overly subjective and artificial.
CONCLUSION: Cognitive interviews are a positive addition to current methods of pretesting questionnaires prior to distribution to the sample. They are most valuable in pretesting questions that are complex, where questions are sensitive and intrusive and for specific groups for whom questionnaire completion may pose particular difficulties.