Health service planning requires information on levels of health and illness in the population. Surveys, such as the British General Household Survey (GHS) rely on self-reports of health, illness and restriction, but interpretation of results is problematic. Multi-item measures such as the Nottingham Health Profile (NHP) tap different aspects of health and allow respondents less freedom to define health and illness. In a survey of 1862 adults, health questions from the GHS and the NHP were used, and the results compared. Responses to GHS questions were associated with NHP scores, but the strength of the associations between the four GHS questions and the six NHP items varied considerably. Reporting a recent restriction was only weakly associated with NHP scores. Associations between GHS questions and NHP scores were weakest for the NHP items measuring emotional reactions, sleep and feelings of social isolation. Reporting good health or no illness in response to GHS questions was no guarantee that respondents experienced no health problems. Those who use health data from the GHS, NHP or similar surveys should look closely at whether such data provide appropriate information for their purposes.