Theories of childhood amnesia and autobiographical memory development have been based on the assumption that the age estimates of earliest childhood memories are generally accurate, with an average age of 3.5 years among adults. It is also commonly believed that early memories will by default become inaccessible later on and this eventually results in childhood amnesia. These assumptions were examined in 2 prospective studies, in which children recalled and dated their earliest memories at an initial interview and did it again 1 year (Study 1) and 2 years later (Study 2). Systematic telescoping errors emerged: Children substantially postdated their memories for the same events at the follow-up interview, particularly for memories initially dated from earlier ages. These findings have critical methodological and theoretical implications for research on childhood amnesia and autobiographical memory development.