Objectives: To validate a method for salivary diagnosis of measles and to assess the diagnostic accuracy of notified cases of measles. Design - Blood and saliva samples were collected within 90 days of onset of symptoms from patients clinically diagnosed as having measles and tested for specific IgM by antibody capture radiommunoassay. Setting: 17 districts in England and one in southern Ireland during August 1991 to February 1993. Subjects: 236 children and adults with measles notified by a general practitioner. Results: Specific IgM was detected in serum in only 85 (36%) of the 236 cases.In cases associated with outbreaks and tested within six weeks of onset, 53/57 (93%) of samples were IgM positive, thereby confirming the sensitivity of serum IgM detection as a marker of recent infection. The serological confirmation rate was lower in cases with a documented history of vaccination (13/87; 15%) than in those without (70/149; 47%) and varied with age, being lowest in patients under a year, of whom only 4/36 (11%) were confirmed. Measles specific IgM was detected in 71/77 (92%) of adequate saliva samples collected from patients with serum positive for IgM. In cases where measles was not confirmed, 6/101 had rubella specific IgM and 5/132 had human parvovirus B19 specific IgM detected in serum. Conclusions: The existing national surveillance system for measles, which relies on clinically diagnosed cases, lacks the precision required for effective disease control. Saliva is a valid alternative to serum for IgM detection, and salivary diagnosis could play a major role in achieving measles elimination. Rubella and parvovirus B19 seem to be responsible for a minority of incorrectly diagnosed cases of measles in the United Kingdom and other infectious causes of measles-like illness need to be sought.