Forty-nine patients with atopic dermatitis entered a double blind placebo controlled cross-over study of mupirocin, a new topical antistaphylococcal antibiotic. Forty-five patients were evaluable. Quantitative bacteriological assessment before treatment showed that heavy colonization of the skin with Staphylococcus aureus was present in nearly all patients even in the absence of overt infection. However, the bacterial count was significantly reduced by 2 weeks' treatment with topical mupirocin, but not by the placebo. Moreover, a significant reduction of clinical severity was also observed after treatment with mupirocin, which was maintained over the following 4 weeks, although recolonization occurred during this period, with bacterial counts rising to pre-treatment levels. Despite recolonization, clinical deterioration was not observed during the trial period. No serious side-effects were observed. Phage typing showed that 50% of patients carried more than one bacterial phage type. Recolonization in eight patients (17%) was with a 'new' strain that had not previously been isolated.