Children's performance on face identification tests improves dramatically between age 4 and adolescence, yet the source of this improvement is controversial. We used face identity aftereffects to examine whether changes in the organization of face-space during childhood could be a source of this improvement. Specifically we tested whether 7- to 9-year-old children, like adults, show patterns of aftereffects predicted by coding facial identity relative to a norm or the patterns predicted by exemplar-based coding. Consistent with use of norm-based coding children's aftereffects were larger (a) for opposite than non-opposite adapt-test pairs equated for perceptual similarity, and (b) for adaptors far from the average than for adaptors closer to the average. In addition, face identity aftereffects were present by age 5, suggesting adult-like face-space properties by 5, though we did not conduct specific tests to distinguish norm-based from exemplar-based coding in this age group. We conclude that children's poor face identification skills cannot be attributed to a failure to use norm-based coding.