It is widely believed that congenital cholesteatoma originates from epidermoid cell rests in the temporal bone. Congenital cholesteatomas of the anterior-superior middle ear may arise from such a rest, known as theepidermoid formation(EF), which has been described by Michaels and others. The EF is thought to disappear by 33 weeks' gestation in most cases; however, if it persists beyond fetal life, it may provide a nidus for cholesteatoma formation. The authors studied human temporal bones from individuals ranging from 20 weeks' gestation to 5 years of age to investigate pre- and postnatal occurrence of the EF. A total of 106 temporal bones were surveyed; 76 of these were fetal specimens and 30 were postnatal. EFs were present in 14 (18.4%) of the fetal specimens; they were identified in four(13.3%) of the postnatal temporal bones, with the oldest being 2 years, 7 months of age. Multiple EFs within a single temporal bone were also present in some cases. Although the EFs were composed of cells morphologically identical to those of the external ear canal epidermis, none of the specimens showed keratinization. These findings support the contention that EFs do occasionally persist into postnatal life and may be a source of cholesteatoma. However, a clear transition from epidermoid formation to unequivocal cholesteatoma has not yet been demonstrated.