A postal survey aimed at studying the prevalence of chronic pain and the need for care because of pain was carried out in a representative sample of a Swedish population. A random sample of the study population has been interviewed and assessed clinically to determine whether a questionnaire method gives the same results as clinical assessment. The agreement between estimates based on the questionnaire and those based on the clinical assessments varied depending on the questions used. By creating questions more independent of language habits and values, it was shown that it is possible to get better agreement between different descriptions. The best agreement was shown when answers to questions regarding different dimensions of pain were combined in a model for selecting individuals with more serious pain problems.