Two studies help reveal the dynamics of memory. New memories that weaken during the day can be strengthened by a period of sleep. And when memories are reactivated, they must be re-stored in order to persist.
Memory research is undergoing a transformation. No longer is memory thought to be a hard-wiring of information in the brain. Instead, it seems to be a process of storage and re-storage1. Two papers in this issue take us closer to understanding the dynamic nature of human memory. On page 614, Fenn et al.2 show that sleep can rescue memories that were lost during the day. And on page 616, Walker et al.3 demonstrate that stabilized memories can be re-stored when they are reactivated — and that those memories are lost if this process is interfered with.
When we learn something new, the memory passes through two states. The first is called short-term memory. In this state, memories are 'labile' — interfering with their processing in the brain changes how the memory is stored. Over the following several hours the memory is stabilized, or consolidated4, by a process that requires neurons to synthesize new RNA and proteins (cellular consolidation)5. These long-term, consolidated memories are usually thought of as being fixed in the brain. A good example of memory transition is remembering a new telephone number: if you are distracted immediately after learning a new number, you are likely to forget it. But if the same distraction is delayed by a day you will probably remember the number. Immediately after learning the number, the memory is labile and sensitive to disruption, whereas 24 hours later the memory is consolidated and resistant to the same disruption.
Sleep has been linked to memory consolidation6 — it can often enhance the memory of information acquired the previous day. So sleep helps some memories 'mature' and also prunes out unimportant memories. But the benefits obtained from sleep are typically thought to be restricted to previously learned information. In other words, the sleep-enhanced memory of a particular sequence of movements (motor sequence) does not improve the performance of other motor sequences7.
Fenn et al.2 demonstrate a new effect of sleep on memory. The authors suggest that, as well as enhancing memories, sleep can rescue memories that are weakened during the preceding day. Remarkably, they also show that this benefit is not restricted to the specific information acquired on that day — the rescued memories can be applied to improve performance in new situations.
The authors looked at the effect of sleep on the performance of a perceptual learning task. Human subjects were asked to identify a series of phonetically similar words produced by a synthetic speech machine. The subjects never heard the same word twice, so their improvement in identifying new words depended on their ability to apply what they had learned during a previous training period. In other words, their ability to generalize was tested.
After a single training period during the morning, the subjects showed a 21% increase in the accuracy with which they could identify new words. After 12 waking hours, that accuracy was reduced to 10%, but accuracy was not decreased if the 12-hour interval included a period of sleep. And, crucially, the decrease in accuracy seen after 12 waking hours could be recovered with a subsequent period of sleep. Additional groups that were trained either in the morning or evening and then tested 24 hours later also showed an 18% increase in accuracy, indicating that the improvement was not due to general circadian effects on learning.
These results extend another finding — that power naps can reverse the detrimental effects of overtraining8 — in two ways. First, Fenn et al. show that sleep can rescue memories that have spontaneously deteriorated, and second, they demonstrate that memories involving generalization can be recovered. It will be interesting to see what other kinds of memories can be rescued by sleep and what phases of sleep cause this effect.
But how does sleep enhance memory? Cellular consolidation takes only a few hours. So according to theory, most memories should be fixed in the brain and insensitive to further manipulation before a period of sleep occurs. Walker et al.3 examined the consolidation profile of a motor memory by looking at interference between two related memories. Interference — when the acquisition of a second memory impairs the memory of the original task — is time dependent and is a classical approach to determining whether memories are labile or consolidated4.
The authors show that if a second finger-tapping exercise is learned immediately after a first exercise, the memory of the second sequence interferes with that of the first, and the performance of only the second sequence is enhanced by sleep. The memory of the second sequence is not somehow erasing the memory of the first, as the subjects performed the first exercise normally if they were tested immediately after the training for the second exercise. In contrast, if the second sequence is learned six hours after the first sequence, interference does not occur and the performance of both sequences is enhanced by sleep.
This finding demonstrates that the memory of the first sequence becomes consolidated within six hours. So how can a subsequent period of sleep enhance this already 'fixed' memory? The authors have previously suggested that there is a second time window for consolidation during sleep, but how can fixed memories become sensitive to further processing?
The answer probably includes the fact that reactivating a consolidated memory can return it to a labile state from which it needs to undergo cellular re-storage or reconsolidation in order to persist. For example, if auditory fear memories are reactivated in rats, protein synthesis is required for these memories to persist. If the fear memories are not reactivated, inhibiting protein synthesis has no effect on their persistence, demonstrating that reconsolidation is predicated on memory reactivation1. So memory storage seems to occur more than once.
There is excellent evidence that some memories are reactivated during sleep and that this might return them to a labile state9,10. The crucial question, however, is whether reconsolidation occurs in humans. Memory reconsolidation in people exposed to electroconvulsive-shock therapy was reported some 30 years ago11, but it was not clear whether the phenomenon could be detected under more natural conditions. Walker et al.3 have now demonstrated exactly this. They looked at whether consolidated motor memories, when reactivated, would return to a labile state that would again be sensitive to interference. Subjects trained on the first finger-tapping exercise on day 1 performed with greater speed and accuracy when they were retested on days 2 and 3. But if, on day 2, they were given the second exercise immediately after the retest of the first, all improvements in the accuracy of the first exercise were reversed by day 3. Strikingly, this interference was prevented if the first exercise was not retested on day 2 — in other words, if this memory was not reactivated.
These results show that reactivating a consolidated motor memory returns it to a labile state that is once again sensitive to interference. Walker and colleagues show that new memories and reactivated memories respond in qualitatively similar ways when challenged using interference, so the most parsimonious interpretation of the data is that the reactivation of the consolidated motor memory caused it to undergo reconsolidation. The interference-induced block to reconsolidation might represent the first convincing demonstration of the erasure of a consolidated memory in humans.
The finding that reconsolidation occurs in humans is a landmark discovery. But Walker et al. also demonstrate that not all memories are equal. Although the accuracy of the finger-tapping tests was affected by interference, the speed with which they were performed was not. So perhaps not all memories undergo reconsolidation. Alternatively, the reconsolidation of different memories might be affected by interference in different ways. Understanding the intricacies of this phenomenon is crucial if we are to exploit its clinical potential to address psychological disorders such as post-traumatic stress or addiction.
Nader, K. Trends Neurosci. 26, 65–72 (2003).
Fenn, K. M., Nusbaum, H. C. & Margoliash, D. Nature 425, 614–616 (2003).
Walker, M. P., Brakefield, T., Hobson, J. A. & Stickgold, R. Nature 425, 616–620 (2003).
Müller, G. E. & Pilzecker, A. Z. Psychol. Suppl. 1, 1–288 (1900).
Kandel, E. R. Science 294, 1030–1038 (2001).
Siegel, J. M. Science 294, 1058–1063 (2001).
Karni, A. et al. Proc. Natl Acad. Sci. USA 95, 861–868 (1998).
Mednick, S. C. et al. Nature Neurosci. 5, 677–681 (2002).
Wilson, M. A. & McNaughton, B. L. Science 265, 676–679 (1994).
Spear, N. E. & Gordon, W. C. in Sleep, Dreams and Memory (ed. Fishbein, W.) 183–203 (Spectrum, New York, 1981).
Rubin, R. D., Fried, R. & Franks, C. M. in Advances in Behavior Therapy, 1968 (eds Rubin, R. D. & Franks, C. M.) 37–44 (Academic, New York, 1969).
About this article
Molecular Psychiatry (2020)
Anxiety Disorder Research (2019)
Activation of the dopamine D1 receptor can extend long-term spatial memory persistence via PKA signaling in mice
Neurobiology of Learning and Memory (2018)
A Historical Review of Diachrony and Semantic Dimensions of Trace in Neurosciences and Lacanian Psychoanalysis
Frontiers in Psychology (2017)