In a previous longitudinal study of basal cortisol levels and cognitive function in humans, we showed that elderly humans with 4- to 7-yr cumulative exposure to high levels of cortisol present memory impairments, compared with elderly humans with moderate cortisol levels over years. Here, we measured whether memory performance in two groups of elderly humans separated on the basis of their cortisol history over a 5-yr period could be modulated by a hormone-replacement protocol in which we inhibited cortisol secretion by the administration of metyrapone and then restored baseline cortisol levels by infusion of hydrocortisone. We showed that in elderly subjects with a 5-yr history of moderate cortisol levels (n = 8), metyrapone treatment significantly impaired memory performance, a deficit that was reversed following hydrocortisone replacement. In the elderly subjects with a 5-yr history of high cortisol levels and current memory deficits (n = 9), metyrapone treatment did not have any significant effect on memory performance, but hydrocortisone treatment significantly decreased delayed memory. These results suggest that memory function in elderly humans can be intensely modulated by pharmacological manipulation of glucocorticoids, although the direction of these effects depends on the cortisol history of each individual.