SLEEP AND MEMORY CONSOLIDATION

More than two hundred years have passed since David Hartley (English psychologist and philosopher, 1705-1757) first proposed that dreaming might alter the strength of associative links between memories, and more than one hundred years since Sigmund Freud (Austrian neurologist and founder of psychoanalysis, 1856-1939) suggested that dreaming served to process traumatic memories. But it has only been since 1953, with the discovery of rapid eye movement (REM) sleep by Eugene Aserinsky and Nathaniel Kleitman, that studies of sleep's role in memory processing began in earnest. Since then, a wide range of studies have provided converging evidence on the important role sleep plays in the off-line reprocessing of waking memories. Whether dreaming plays a similar role remains unclear.

Sleep's complex role in memory consolidation makes our understanding of that role more difficult. Multiple memory systems exist within the brain store and process different types of information in separate anatomical regions. For example, episodic memory recall in humans is dependent on the hippocampus, whereas access to procedural memories is not. Numerous mechanisms can contribute to memory consolidation. Consolidation can refer to the simple strengthening of a memory, its movement from one memory system to another, or its functional linking to other, associated memories. As a result, sleep's contribution to memory consolidation depends on the precise memory system involved and the form of consolidation being considered. Sleep represents a similarly complex phenomenon because it consists of a series of brain states with different neurophysiological and neurochemical properties. This deceptively simple question of sleep's role in memory reprocessing must be expanded to address the contribution of specific stages of sleep to various mechanisms involved in the consolidation of several distinct forms of memory.

The REM Sleep Cycle

Sleep in mammals follows a rhythmic pattern of alternating REM and nonREM (NREM) sleep. In humans, this cycle has a period of approximately ninety minutes, and continues throughout the night (see Figure 1, top panel). NREM sleep is divided into four stages, ranging from very light (stage I) sleep onset sleep to deep slow wave sleep (SWS; stages III and IV), so named because of its characteristic electroencephalographic (EEG) pattern (see Figure 1, second panel). Throughout the night this ninety-minute period remains relatively constant, but a slow shift from a preponderance of SWS early in the night toward a preponderance of REM sleep at the end of the night occurs.

Several physiological parameters vary across the REM cycle. As sleep progresses from stage I to stage IV, EEG patterns show progressively slower and higher amplitude waves, whereas REM sleep shows a high frequency low amplitude EEG pattern. Thus, REM sleep has been given the alternate name of paradoxical sleep, based on the similarity between its EEG pattern and the pattern seen in waking. These distinct EEG waveforms seen in different sleep stages may contribute differentially to various aspects of memory consolidation. Tonic muscle activity, measured by the electromyogram (EMG), decreases with descent through light NREM sleep into SWS, but is at its lowest level in REM sleep (see Figure 1, third panel). During REM, a descending spinal path from the brainstem bulbar reticular formation actively inhibits the voluntary muscles. The resulting atonia produces a functional paralysis during REM sleep, which is necessary to prevent the physical acting out of dreams. Spontaneous eye movements, recorded in the electrooculogram (EOG) also show distinct stage-dependent patterns (see Figure 1, bottom panel). While NREM sleep stages II through IV show little or no eye movements, both sleep onset (stage I) and REM sleep show stereotypical patterns of movements. During sleep onset, slow rolling eye movements, lasting one to three seconds are observed, whereas during REM sleep phasic bursts of rapid eye movements are seen. These bursts correlate with times of peak dream recall.

The shift from NREM to REM sleep is accompanied by an increase in release of acetylcholine in the brain and a simultaneous near-cessation of release of norepinephrine and serotonin. Brain imaging studies show that most brain regions become less active during NREM sleep; several distinct regions, including the anterior cingulate and medial orbitofrontal cortices and the amygdala become more active in REM sleep. Together, these changes are thought to control the variations in memory reprocessing and dreaming seen across the sleep cycle.

Human Procedural Skill Consolidation

The clearest evidence of the important role sleep plays in human memory reprocessing comes from studies of the consolidation of procedural learning. Posttraining sleep can improve both perceptual and motor skill learning.

In visual texture discrimination tasks, training only leads to improved performance after a night's sleep (see Figure 2). Subjects trained and then retested the same day show no improvement, whereas subjects retested after a night's sleep show significant improvement. Similarly, subjects that are sleep deprived the night after training and then retested two days later, also show no improvement. Thus, sleep the night after training appears critical for the consolidation of this learning. Those stages of sleep that are most important remain unclear. Studies have found that selectively decreasing either REM or SWS can block improvement, and overnight improvement correlates both with the amount of SWS early in the night and the amount of REM sleep late in the night. These findings in humans match findings in rats, suggesting that a two-step process of sleep-dependent memory

[Image not available for copyright reasons]

consolidation requiring SWS followed by REM sleep occurs.

In contrast to these findings with perceptual skill tasks, motor skill tasks can show dependence on stage II NREM sleep. Individuals show a twenty percent increase in speed on a finger-tapping task after a night's sleep, correlating with the amount of stage II sleep

[Image not available for copyright reasons]

obtained, especially late in the night. These differences between perceptual and motor skill learning may reflect the fact that different brain regions are involved in each of these forms of learning.

Human Declarative Memory Consolidation

Less convincing evidence exists for sleep's role in the consolidation of declarative memories. Sleep deprivation, particularly REM sleep deprivation, has little or no effect on the retention of simple declarative memories such as paired word associates, but may interfere with the retention or consolidation of more complex declarative memories, such as recall of lists of words grouped into categories, or the acquisition of such complex skills at BASIC computer programming, foreign languages, or Morse code. It is, however, unclear whether it is the specifically declarative portion of such learning or more subtle, nondeclarative components that are being affected. Nevertheless, these results suggest that some aspects of complex declarative memories are supported by sleep.

Animal Memory Consolidation

Because the classification of a memory as declarative requires a verbal statement of recall, it is impossible to know whether animals possess such memories. However, many forms of animal learning are clearly impaired by subsequent REM sleep deprivation, indicating that the role of sleep in memory consolidation is not uniquely human. In rats, posttraining REM deprivation impairs both aversive and appetitive tasks, although simpler tasks may not share this property. Thus, simple shock avoidance training is unaffected by subsequent REM sleep deprivation, whereas a more complex, shuttle box avoidance task is REM sleep dependent. Memory consolidation appears to be sensitive to REM deprivation only during specific REM windows (Carlyle Smith, 1985), often occurring hours to days after the initial training. These REM windows are further characterized by increased REM sleep following training. During periods when REM-dependent memory consolidation is occurring, the brain appears to produce more REM sleep.

In humans, declarative memories are dependent on the hippocampus for their encoding and initial recall, whereas procedural skill learning is largely independent of this structure. In rats, spatial learning tasks are hippocampally mediated, and it is possible to look at the role of sleep in consolidating hippocampally mediated memories in rats by comparing spatial and non-spatial tasks. Surprisingly, both types of tasks are found to be sleep dependent; posttraining REM deprivation impairs performance on both the Morris water maze and the eight-arm radial arm maze. Thus, these tasks may correspond to the complex declarative memory tasks in humans, which show a similar REM dependency.

Memory Reactivation in Sleep

Studies showing that patterns of brain activation seen during learning are reactivated during sleep provide additional support for sleep dependent memory processing. Evidence for this comes from both animal and human experiments. Recordings from the rat hippocampus show the most direct evidence, revealing that the rat hippocampus, during both REM and NREM sleep, activates patterns of neuronal activity that mimic patterns seen earlier when the rat was navigating a maze. The patterns are sufficiently complex that one can visualize the virtual maze running activity of the sleeping rat. Interestingly, this repetition is seen during NREM sleep only in the first thirty minutes after maze running, while the REM reactivation is seen twenty-four hours later. Thus, as has been suggested from behavioral studies, memory processing might occur first during NREM sleep and only subsequently during REM sleep. Similar patterns of reactivation have been seen in the neo-cortex. Unfortunately, no behavioral studies exist to show that this replay of patterns of either hippocampal or neocortical neuronal activity during sleep are actually associated with memory consolidation.

Less direct evidence of the reactivation of neuronal ensembles coding memory traces is found in dream reports collected during the sleep onset period following intense video game play. Sixty to ninety percent of subjects that played either the video game Tetris, or the arcade game Alpine Racer II, reported dreamlike images from the game when awakened during the first few minutes of sleep following several hours of intensive game play. In most cases, the images were accurate copies of game elements, suggesting that memory traces were being reactivated during the sleep onset period. Subjects playing Alpine Racer reported both visual and kinesthetic imagery, indicating that coordinated multimodal replay is occurring. Surprisingly, the reactivation is not hippocampally mediated, since amnesic patients with extensive damage to both hippocampi produce similar Tetris images despite being unable to identify the source of the images or to recall playing the game.

Some researchers continue to question sleep's role in memory consolidation, but the findings reviewed here point toward an important and complex role for sleep in the off-line reprocessing of learning and memory. The evidence is clearest for the role of REM sleep in the consolidation of procedural learning. The possible roles of deep sleep in consolidating procedural memories and of stage II NREM sleep in consolidating motor skill learning are less clear, as is sleep's role in consolidating and integrating declarative memories. There is mixed evidence of roles for both SWS and REM sleep in these processes. Taken together with evidence for patterns of neuronal replay during REM, NREM, and even sleep onset, a picture begins to emerge in which each stage of sleep makes a unique contribution to off-line memory reprocessing. Further work is needed to permit the unequivocal identification of these contributions.

See also:MEMORY CONSOLIDATION: MOLECULAR AND CELLULAR PROCESSES; MEMORY CONSOLIDATION: PROLONGED PROCESS OF REORGANIZATION; MOTOR SKILL LEARNING

Bibliography

Aserinsky, E., and Kleitman, N. (1953). Regularly occurring periods of ocular motility and concomitant phenomena during sleep. Science 118, 361-375.

Freud, S. (1900). The interpretation of dreams. New York: Basic Books.

Hartley, D. (1791). Observations on man, his frame, his duty and his expectations. London: Johnson.

Hennevin, E., Hars, B., Maho, C., and Bloch, V. (1995). Processing of learned information in paradoxical sleep: Relevance for memory. Behavioural Brain Research 69, 125-135.

Peigneux, P., Laureys, S., Delbeuck, X., and Maquet, P. (2001). Sleeping brain, learning brain. The role of sleep for memory systems. Neuroreport 12 (18), A111-124.

Siegel, J. M. (2001). The REM sleep-memory consolidation hypothesis. Science 294, 1,058-1,063.

Smith, C. (1985). Sleep states and learning: A review of the animal literature. Neuroscience and Biobehavioral Reviews 9, 157-168.

—— (1995). Sleep states and memory processes. Behavioural Brain Research 69, 137-145.

Stickgold, R. (1998). Sleep: Off-line memory reprocessing. Trends in Cognitive Sciences 2 (12), 484-492.

Stickgold, R., Hobson, J. A., Fosse, R., and Fosse, M. (2001). Sleep, learning and dreams: Off-line memory reprocessing. Science 294, 1,052-1,057.

Vertes, R. P., and Eastman, K. E. (2000). The case against memory consolidation in REM sleep. Behavioral and Brain Sciences 23, 867-876.

RobertStickgold