The presence of visible flexural dermatitis is a key feature in diagnosing atopic dermatitis. We describe a protocol for recording this sign in population-based studies, which could be used by suitably trained non-dermatologists. The protocol was developed using a standard set of photographs and accompanying instructions, which define the terms 'dermatitis' and 'flexural'. When assessed during a survey of skin disease in primary schoolchildren, there was excellent agreement between a trained nurse and a dermatologist with regard to the presence or absence of this sign in 73 consecutive children, with perfect agreement in 71 children (97%), and a chance corrected agreement index (kappa statistic) of 0.90 (95% confidence interval [CI] 0.77-0.99). Agreement between two nurses on 114 consecutive schoolchildren was not quite as good, with perfect agreement in 102 children (90%), and a kappa value of 0.51 (95% CI 0.26-0.77). The protocol is easy to perform, does not require children to undress, takes < 1 min to carry out, and is highly acceptable to children and staff. This protocol may be useful in standardizing the assessment of atopic dermatitis in population studies of children.