Memory functions involve three stages: encoding, consolidation, and retrieval. Modulating effects of glucocorticoids (GCs) have been consistently observed for declarative memory with GCs enhancing encoding and impairing retrieval, but surprisingly, little is known on how GCs affect memory consolidation. Studies in rats suggest a beneficial effect of GCs that were administered during postlearning wake periods, whereas in humans, cortisol impaired memory consolidation when administered during postlearning sleep. These inconsistent results raise the question whether effects of GCs critically depend on the brain state during consolidation (sleep vs. wake). Here, we compare for the first time directly the effects of cortisol on memory consolidation during postlearning sleep and wakefulness in different measures of declarative memory. Cortisol (13 mg vs. placebo) was intravenously infused during a postlearning nap or a time-matched period of wakefulness after participants had encoded neutral and emotional text material. Memory for the texts was tested (a) by asking for the contents of the texts (“item” memory) and (b) for the temporal order of the contents within the texts (“relational” memory). Neither postlearning infusion of cortisol during sleep nor during wakefulness affected retention of content words of emotional or neutral texts. Critically, however, the retention of temporal order within the texts, known to rely most specifically on the hippocampus proper within the medial-temporal lobe memory system, was distinctly improved by cortisol infusion during the wake phase but impaired by cortisol during sleep. These results point toward fundamentally different mechanisms of hippocampal memory consolidation, depending on the brain state.

Retrieving a memory is a reconstructive process in which encoded representations can be changed and distorted. This process sometimes leads to the generation of “false memories,” that is, when people remember events that, in fact, never happened. Such false memories typically represent a kind of “gist” being extracted from single encountered events. The stress hormone cortisol is known to substantially impair memory retrieval. Here, in a double-blind, placebo-controlled crossover design, we tested the effect of an intravenous cortisol infusion before retrieval testing on the occurrence of false memories and on recall of correct memories using a modified Deese–Roediger–McDermott paradigm. Subjects studied sets of abstract shapes, with each set being derived from one prototype that was not presented during learning. At retrieval taking place 9 hr after learning, subjects were presented with studied shapes, nonstudied shapes, and the prototypes, and had to indicate whether or not each shape had been presented at learning. Cortisol administration distinctly reduced susceptibility to false memories (i.e., false recognition of prototypes) and, in parallel, impaired retrieval of correct memories (i.e., correct recognition of studied shapes). Response bias as well as confidence ratings and remember/know/guess judgments were not affected. Our results support gist-based theories of false memory generation, assuming a simultaneous storage of the gist and specific details of an event. Cortisol, by a general impairing influence on retrieval operations, decreases, in parallel, retrieval of false (i.e., gist) and correct (i.e., specific) memories for the event.

Sleep crucially contributes to the off-line consolidation of memories. Although this view was confirmed in numerous studies in adults, it is not known whether it can be generalized to sleep during development. Here, we examined effects of sleep on implicit memory formation considered of particular relevance in children, because brain structures underlying implicit learning develop earlier in ontogeny than structures supporting explicit learning. Subjects were 7- to 11-year-old children ( n = 14) and 20- to 30-year-old adults ( n = 12) tested on a serial reaction time task before (learning) and after (retest) equal length retention periods of overnight sleep and daytime wakefulness. At learning, after eight training blocks, all subjects had acquired implicit knowledge of the probabilistic rules underlying the sequential stimulus materials, as indicated by a substantial difference in response time to grammatical versus nongrammatical trials in two test blocks that followed the training blocks. At learning, this response time difference was greater in children (48.49 ± 6.08 msec) than adults (28.02 ± 3.65 msec, p < .01), but did not differ between sleep and wake retention conditions in either age group. Consistent with previous studies, retesting in the adults revealed that the reaction time differences between grammatical and nongrammatical trials increased by 9.78 ± 4.82 msec after sleep, but decreased by −12.76 ± 5.49 msec after the wake retention period ( p < .01). Contrary to this finding in adults, sleep in children did not lead to an increase, but to a decrease in the reaction time difference averaging −26.68 ± 12.25 msec ( p < .05), whereas across the wake retention interval the reaction time difference remained nearly unchanged. The sleep-dependent deterioration in measures of implicit sequence knowledge in children was in striking contrast to the gain of such knowledge in the adults during sleep ( p < .01). Our findings indicate that the functional role of sleep in implicit memory consolidation depends on age. We speculate that the overnight decrease of implicit knowledge in children reflects a preferential effect of sleep toward the enhancement of explicit aspects of task performance that interferes with implicit performance gains.