Adrenal glucocorticoids (GC) secreted during stress modulate memory. Animal and human studies investigating the effects of acute GC treatment on memory have reported conflicting (enhancing as well as impairing) results. Several theories have been proposed to integrate these contradictory findings. Among the variables discussed are the timing of the GC treatment (before learning or before retrieval) and the time of day (morning versus afternoon). Here we review meta-analytically the results of 16 studies, which experimentally investigated the acute impact of cortisol treatment on human memory. The results revealed that the timing of GC application in the course of a study is a relevant variable which explains a substantial amount of the significant heterogeneity within the effect sizes. The studies which administered cortisol before retrieval (n = 4) reported a significant decrease (average effect size of d = -.49) in memory performance. Studies which administered cortisol before learning (n =12) found on average no effect (d = .08), but there is heterogeneity within these effect sizes. Further analysis on these experiments indicated that studies, which administered cortisol in the morning found a significant memory impairment (d = -.40), while studies conducted in the afternoon observed a small but significant memory enhancement (d = .22). This meta-analysis supports the idea that the timing of GC administration (before learning or before retrieval) is a major determinant of the effects of GCs on human memory. We discuss methodological limitations of the current analysis and suggest several areas for future research.