It is generally agreed that sleep aids memory consolidation, but the reasons for this are a mystery. Part of the answer may lie in the patterns of synchronous brain activity unique to the state of slumber.
Only ten years ago, discussions about the purpose of sleep offered great hypotheses, but these were based on flimsy evidence. So scant were the data that some researchers argued that sleep might have no use at all. This led Alan Rechtschaffen, a pioneer in the area, to comment wryly that “if sleep doesn't serve an absolutely vital function, it is the biggest mistake evolution ever made.” Since then, researchers have produced a wealth of evidence for at least one function of slumber — the consolidation of memories. A wide range of converging data show that memories are replayed1, modified2, stabilized3 and even enhanced4 as we snooze, but an understanding of how this occurs remains elusive. The proposed mechanisms rely on two effects that are observed in the brain only during sleep: alterations in the levels of chemical neuromodulators5 and distinctive oscillations of electrical activity4. In this issue (page 610)6, Marshall et al. provide evidence for the second mechanism Footnote 1. They show that the direct induction of 'cortical slow oscillations' during sleep can improve the recall of word-pairs memorized the previous night.
Human sleep is divided into rapid-eye-movement sleep (REM) and non-REM sleep (NREM), with NREM sleep further divided into stages 1 to 4. These sleep phases are distinguished, in part, by well-defined patterns of oscillatory electrical activity in the brain, as measured by electroencephalography. Theta waves oscillate at 4–8 Hz, and are characteristic of REM sleep; NREM stage 1 is a transitional phase between full wakefulness and sleep, and is characterized by mixed-frequency waves; sleep spindles (12–14 Hz) typify NREM stage 2; and delta waves (1–4 Hz) distinguish NREM stages 3 and 4, which are known collectively as slow-wave sleep. Sleep spindles and delta waves rise and fall in concert with a yet slower (<1 Hz) oscillatory pattern known as cortical slow oscillations — it is these patterns that Marshall et al.6 have studied.
The authors asked 13 volunteers to learn 46 word-pairs in a training session before they went to sleep. Electrodes were then placed on the scalps of the sleeping subjects, and fluctuating electrical potentials were applied to induce cortical slow oscillations. Remarkably, the morning after this treatment, the subjects demonstrated enhanced recall of the word-pairs compared with their performance the morning after receiving a sham stimulation. This effect was not seen for all forms of memory. For example, no improvement was observed for a motor-skill procedural memory task — learning to trace around a shape by looking at its image in a mirror. Furthermore, the memory boost occurred only if the electrical stimulation matched the oscillation frequency of cortical slow waves (0.75 Hz) and not when the stimulation was at the theta frequency (5 Hz) of REM sleep. The time chosen for the stimulation was also crucial — the authors observed no effect on recall if stimulation was during the last 45 minutes of the night instead of the first 45 minutes.
The actual increase in words recalled was small; averaged over the whole group, the subjects remembered only 1.8 more words after stimulation than they did after the sham treatment. But the improvement in subjects' recall compared with their performance during the training session was statistically significant — on average, after stimulation, 4.8 more words were remembered by the group in the morning than had been recalled the previous night, compared with just 2.1 more words in the control experiments. These increases probably do not reflect a genuine overnight improvement, because they were based on comparisons with performance during the training session (when the words would have been imperfectly memorized), not after it (when peak performance would be expected). Nevertheless, sleep-dependent consolidation, enhanced by direct electrical stimulation, does seem to stabilize memories. But why does this happen?
The authors began electrical stimulation of the subjects' brains 4 minutes after the volunteers had entered stage 2 of NREM sleep — that is, 5 to 10 minutes before they would normally be expected to progress into slow-wave sleep; the stimulation was applied in five 5-minute periods, separated by stimulation-free pauses lasting 1 minute each. Marshall and colleagues' analysis of the stimulation-free interludes revealed greater electrical activity in the subjects' brains at slow-oscillation frequencies (up by 60%) and at sleep-spindle frequencies (up by 51%), as well as a 37% increase in the amount of time spent in slow-wave sleep, compared with sham stimulation in the same subjects. Arguably, any of these effects could underlie the observed enhancement of memory consolidation. Indeed, sleep spindles7 and the duration of slow-wave sleep8 have previously been implicated in memory consolidation. But Marshall et al.6 argue that the spindles are the key players. Spindles produce large influxes of calcium ions into cortical neurons, where they can trigger molecular cascades known to strengthen the connections between neurons — which presumably leads to memory consolidation9.
Might the effects6 of electrical stimulation be caused by alterations in the levels of neuromodulatory chemicals? The authors argue that they are not. They measured blood levels of some of these compounds — cortisol, growth hormone and noradrenaline — and reported no changes from normal as a result of the stimulation. Other neuromodulators were not studied, most notably acetylcholine, which is known to be involved in memory formation. This crucial point requires further exploration, as earlier reports from the same group indicate that low levels of acetylcholine5 and cortisol10are required for sleep-dependent consolidation of memories.
Such uncertainty indicates a need for continued research into sleep-dependent memory consolidation. But rapid resolution will not come soon, as several complicating issues remain unresolved. First, many different memory systems exist — those involved in memorizing word-pairs may be different from those required for learning to ride a bicycle, or those concerned with recalling the details of an emotional event. Memory consolidation in different systems clearly correlates with different stages of sleep, and possibly even with different components of those stages. Second, consolidation seems to be a series of events, but we do not know which of these are sleep-dependent (within any given memory system). Finally, our understanding of the cellular and molecular mechanisms underlying waking memory consolidation is still highly primitive. It is time for sleep researchers to tackle these basic issues.
This article and the paper concerned6 were published online on 5 November 2006.
Peigneux, P. et al. Neuron 44, 535–545 (2004).
Walker, M. P., Stickgold, R., Jolesz, F. A. & Yoo, S. S. Cereb. Cortex 15, 1666–1675 (2005).
Ellenbogen, J. M., Hulbert, J. C., Stickgold, R., Dinges, D. F. & Thompson-Schill, S. L. Curr. Biol. 16, 1290–1294 (2006).
Huber, R., Ghilardi, M. F., Massimini, M. & Tononi, G. Nature 430, 78–81 (2004).
Gais, S. & Born, J. Proc. Natl Acad. Sci. USA 101, 2140–2144 (2004).
Marshall, L., Helgadottir, H., Mölle, M. & Born, J. Nature 444, 610–613 (2006).
Fogel, S. M. & Smith, C. T. J. Sleep Res. 15, 250–255 (2006).
Stickgold, R., Whidbee, D., Schirmer, B., Patel, V. & Hobson, J. A. J. Cogn. Neurosci. 12, 246–254 (2000).
Sejnowski, T. J. & Destexhe, A. Brain Res. 886, 208–223 (2000).
Plihal, W. & Born, J. NeuroReport 10, 2741–2747 (1999).