Objectives: To evaluate the hypotheses that subjective memory ratings are more accurate in the oldest old than in the young old and more accurate in a representative sample than in a convenience sample.
Methods: Outcomes of hierarchical regressions of subjective ratings and participant characteristics on recall were compared between a nationally representative sample of 6,446 adults ages 70 to 103 and a convenience sample of 326 adults ages 70 to 97.
Results: Education interacted with memory ratings in the prediction of performance in the representative sample, with better prediction for more highly educated participants than for participants with lower levels of education.
Discussion: The general pattern of findings was consistent across both samples and neither hypothesis was supported. Possible explanations for the similarity of results across samples are considered.