Using a testing-the-limits paradigm, the authors investigated the modulation (attenuation) of negative adult age differences in imagery-based memory performance as a function of professional expertise. Six older graphic designers, 6 normal older adults, 6 younger graphic design students, and 6 normal younger students participated in a 19-session program with a cued-recall variant of the Method of Loci. Older graphic designers attained higher levels of mnemonic performance than normal older adults but were not able to reach younger adults' level of performance; a perfect separation of age groups was achieved. Spatial visualization was a good predictor of mnemonic performance. Results suggest that negative adult age differences in imagery-based memory are attenuated but not eliminated by the advantages associated with criterion-relevant ability (talent) and experience. The purpose of this study was to examine the effect of imagery-related professional expertise on the magnitude of age differences in memory functioning with the Method of Loci, a mnemonic strategy for the serial encoding and recall of word lists (Bower, 1970; McDaniel & Pressley, 1987; Paivio, 1971). Recent research using a testing-the-limits paradigm (Baltes, 1987; Kliegl & Baltes, 1987) has reported the existence of large and robust age differences favoring the young in recall performance with the Method of Loci during adulthood (Baltes & Kliegl, 1992; Kliegl, Smith, & Baltes, 1989). The present study investigates whether age differences would also be observed in a group of older adults with large amounts of criterion-relevant preexperimental practice and high levels of criterion-relevant ability. Evidence from different lines of research suggests that visual imagery plays an important role in memory functioning with the Method of Loci. First, it has been shown that the reported The research described in this article was conducted in the context of the project "Expertise and Cognitive Aging" (principal investigators, Reinhold Kliegl and Paul B. Baltes). A part of the data presented here stems from dissertation work conducted by Ulman Lindenberger. The dissertation work was supported by a stipend of the Max Planck Society and by a predoctoral traineeship from the Volkswagen Foundation. We would like to express our gratitude to Ulrich Mayr, Jacqui Smith, Ursula Staudinger, and Gerhard Steiner for their valuable contributions to our thinking about the topic of this article and to three anonymous reviewers who provided helpful comments on an earlier version of this article. We would also like to thank Daniela Becker and Annette Rentz for assistance in data collection. Ulman Lindenberger is now a research scientist at the Academy of Sciences and Technology in Berlin.