A working party of 13 dermatologists, two family practitioners and a paediatrician was assembled, with the aim of developing a minimum list of reliable discriminators for atopic dermatitis. Each physician was asked to select 10 consecutive new cases of unequivocal mild to moderate atopic dermatitis and 10 controls with other inflammatory dermatoses. Each subject was examined by two independent observers, who were blind to the clinical diagnosis and study aim, with regard to 31 clinically useful diagnostic features for atopic dermatitis. Two hundred and twenty-four patients were studied (120 cases and 102 controls). Using the key physician's clinical diagnosis as a gold standard, the sensitivity and specificity of each of the 31 diagnostic criteria were tested. Using multiple logistic regression techniques, a minimum set of diagnostic criteria for atopic dermatitis was derived. These were: history of flexural involvement, history of a dry skin, onset under the age of 2, personal history of asthma, history of a pruritic skin condition, and visible flexural dermatitis. Adjustment for age, sex, region, social class and ethnic group did not alter the choice of final criteria. The discriminatory value of these criteria was also satisfactory when tested against a further sample of 150 patients drawn from the community, who did not have skin disease.