A major challenge for survey-based research is to reduce bias. In an interview, subjects may claim more favourable behaviour to please the interviewer or comply with accepted norms. This study aimed to: (i) determine if adolescents give different answers when using a self-completed computer-administered oral health questionnaire compared with a personal interview, (ii) ascertain if responses to a face-to-face interview are dependant on whether a dentist or nonclinical researcher administered the questionnaire and (iii) examine if responses were influenced by whether they undertook the computer questionnaire first or second. A randomized crossover design was used to investigate the responses to 15 closed questions on oral health-related practices. These were administered to 453, 12-year olds attending school dental inspections in South Wales. A total of 449 valid pairings of computer/interview responses were available for analysis. Responses to the questionnaire variables demonstrated good to very good levels of agreement (kappa 0.68-0.90) when comparing the face-to-face and computer-administered questionnaire. With the exception of questions on dental attendance, responses were not influenced by whether the questions were posed by the research officer or the dentist. A minimal order effect could be detected when undertaking the computer-administered questionnaire first. Whilst a study such as this cannot determine the true validity of the responses achieved, it is concluded that a computer-administered questionnaire, comprising closed questions, can be utilized to determine oral health-related behaviours in oral health surveys.