The value of IgG and IgA gliadin antibodies (AGA) was compared with that of IgA endomysium antibodies (EMA) for the diagnosis of coeliac disease. Three hundred and six of 340 (90%) children with untreated coeliac disease (flat mucosa) had EMA and 338/340 (99.4%) had IgG AGA and/or IgA AGA. Only 1/340 (a 7 year old boy with selective IgA deficiency) had neither AGA nor EMA. Absence of EMA is more frequent in coeliac patients younger than 2 years than in older patients (32/277 compared with 1/62). EMA were present in 4/211 (2%) of comparison subjects (normal mucosa), IgA AGA in 12/211 (6%), and IgG AGA in 74/211 (35%). The specificity of AGA cannot be calculated from these figures as they are biased. The combined determination of AGA and EMA, taking advantage of the high sensitivity of AGA and the high specificity of EMA, gives an excellent prediction of the condition of the mucosa: 247/248 patients (99.6%) with positive EMA and positive IgG AGA and IgA AGA had a flat mucosa, whereas 136/137 patients (99.3%) with neither AGA nor EMA had a normal mucosa. During a gluten free diet EMA and AGA disappear. Their presence or absence is therefore an indicator of dietary compliance. After reintroduction of gluten into the diet 110/134 (82%) of the patients who had a flat mucosa at diagnosis relapsed, but 24/134 still had a normal mucosa after 2-15 years of challenge. All these patients without a morphological relapse were less than 2 years old at diagnosis so we conclude that patients who are young at diagnosis should be challenged. AGA often reappear earlier than EMA. After one month of challenge 93% of patients are AGA and 69% EMA positive. After more than three years of gluten intake the percentage of AGA positive patients decreased to about 50% whereas the percentage of EMA positive sera was then highest (93%). Therefore EMA are more sensitive for the detection of 'silent' relapse after prolonged periods of gluten intake.