Several researchers have proposed that developmental improvements in children's face recognition abilities might reflect an increasing reliance on configural information (i.e. spatial relations between features) in faces (Carey & Diamond, 1994; Mondloch, Le Grand & Maurer, 2002). We investigated 4- and 5-year-olds’ use of configural information for upright and inverted faces using Tanaka and Sengco's (1997) configural change paradigm. Participants saw a photograph of a child's face (e.g. Luke). Memory for features (e.g. Luke's mouth) was then tested in a face with the same configuration as the study face (e.g. Luke's face), in a face with a novel spatial configuration (e.g. Luke's face with the eyes shifted further apart), and presented on their own (e.g. Luke's mouth alone). We found that preschoolers and adults recognized target features from upright faces better when tested in the context of a face with the same configuration as the study face, than when they were embedded in a face with a new spatial configuration or when they were presented in isolation. This effect was lost when faces were inverted. The results suggest that adult-like styles of face processing are present from 4 years of age.