Only in the last 10—15 years have social researchers acknowledged the importance of conducting survey research with children directly, rather than relying on qualitative findings plus prevalence data collected by proxy from adults. With reference to the question—answer model, I here present and consider a number of guiding principles for the design of appropriate and effective questions for younger respondents. Pretesting of questions is especially important when researching populations about which relatively little is known, or who are likely to be especially sensitive to failings of questionnaire design. Children fall into this category because relatively little is known about how the question—answer model applies to them, and because their cognitive, memory, communicative and social faculties are still developing. With this in mind, I describe two key methods of question-testing, namely review by expert panel and in-depth cognitive interviewing.