Two experiments were designed to examine Carey's (1978) theory that face recognition improves with age because young children (< 10 years) encode predominantly piecemeal details from unfamiliar faces, whereas older children (≥ 10 years) and adults rely mainly on configurational information. In Expt 1, children (7–16 years) were tested for ability to recognize unfamiliar faces presented upright and inverted. Performance in the inverted condition was significantly poorer for all age groups. In a second experiment, subjects (4–8 years) were given a forced-choice, face recognition task. The tendency of young children to select incorrectly paraphernalia cues as a basis for identity judgements was found to be dependent on the similarity of the faces paired in each trial. These results suggested that Carey's original data were contaminated by floor effects. It was argued that there is insufficient evidence to endorse Carey's explanation of an encoding switch at age 10 years as a satisfactory account of the development of face recognition.