A random sample of adults over 60 years of age was drawn from three different areas of England to take part in a survey of their dental status and needs. The paper describes sampling methods and operational problems in trying to measure oral health in such an age group. The local Family Health Services Authority lists, based on registrations with general medical practices, were used as sampling frames. These proved to be straightforward to manipulate and are believed to be almost complete for the population at this age. Final samples of around 800 individuals in each area were visited at home, representing 53-62 per cent of all subjects who could be contacted, except in one subgroup where only 39 per cent of contacts permitted a visit. There were problems related to difficulty making contact with a proportion of the names sampled. This may have resulted from list errors, illnesses and unrecorded changes of address. A letter from the subjects' family doctors appeared to be an important factor influencing response rates. Systematic bias among subjects refusing a visit was assessed using a follow up postal questionnaire which showed that the edentulous were under-represented in the final (visited) sample by 1-10 per cent, whilst the dentate non-attenders were under represented by 2-13 per cent.