Cortisol's effects on memory follow an inverted U-shaped function such that memory retrieval is impaired with very low concentrations, presumably due to insufficient activation of high-affine mineralocorticoid receptors (MR), or with very high concentrations, due to predominant low-affine glucocorticoid receptor (GR) activation. Through corresponding changes in re-encoding, the retrieval effect of cortisol might translate into a persistent change of the retrieved memory. We tested whether partial suppression of morning cortisol synthesis by metyrapone, leading to intermediate, circadian nadir-like levels with presumed predominant MR activation, improves retrieval, particularly of emotional memory, and persistently changes the memory. In a randomized, placebo-controlled, double-blind, within-subject cross-over design, 18 men were orally administered metyrapone (1g) vs. placebo at 4:00 AM to suppress the morning cortisol rise. Retrieval of emotional and neutral texts and pictures (learned 3 days earlier) was assessed 4 hours after substance administration, and a second time one week later. Metyrapone suppressed endogenous cortisol release to circadian nadir-equivalent levels at the time of retrieval testing. Contrary to our expectations, metyrapone significantly impaired free recall of emotional texts (p < .05), whereas retrieval of neutral texts or pictures remained unaffected. One week later, participants still showed lower memory for emotional texts in the metyrapone than placebo condition (p < .05). Our finding, that suppressing morning cortisol to nadir-like concentrations not only impairs acute retrieval, but also persistently weakens emotional memories corroborate the concept that retrieval effects of cortisol produce persistent memory changes, possibly by affecting re-encoding.