A commonly used test of non-verbal memory, which measures recognition for unfamiliar face pictures, was developed by Warrington (1984), the Recognition Memory for Faces (RMF) test. The task has been widely used in adults in relation to neurological impairment of face recognition. We examined the relationship of RMF scores to age in 500 young people aged between 6 and 16 years. A.linear relationship obtained between 6 and 10 years, followed by a 'plateau' between the ages of 10 and 13, followed by further improvement. Abilities on the test correlated with both verbal and non-verbal intelligence, but the nonlinear function relating age and RMF survived partialling for intelligence in the younger age groups. The improvement of the adolescents compared with older children also survived partialling for IQ. We found no significant influence of gender or reported pubertal status (which was not obtained for all participants) on RMF once age was taken into account. Performance on other face-processing tasks (emotion classification and accuracy in line of sight detection) correlated significantly, if moderately, with RMF scores. Despite its age and imperfections, RMF test may nevertheless be a useful indicator of 'face expertise' in a developmental context.