The Memory Function of Sleep

How the Sleeping Brain Impacts Memory

Based upon the studies discussed in the previous entries (1 Opens in new window|2 Opens in new window|3 Opens in new window|4 Opens in new window), it is clear that sleep Opens in new window does have a strong influence on memory Opens in new window, but the question still remains of exactly how such a marked impact on various memory systems is afforded by sleep.

While the answer to this question is not yet known, a number of mechanisms likely work in tandem to create the effects seen in the aforementioned studies. These studies pointed toward the contribution of specific sleep stages to memory improvements.

However, it is important to note that this may be an oversimplification of how exactly sleep has its impact on memory. There are a number of unique physiological processes occurring during different sleep stages which may better account for these effects, including varying activation patterns, electrophysiological signatures, and neurochemical composition of the brain.

Brain activation during sleep

A variety of positron emission tomography (PET) studies have investigated rCBF during various stages of sleep and wake, and have found drastic differences in brain activation profiles between these states.

One such study had participants sleep in a PET scanner to compare rCBF during wake, SWS, and REM sleep (Braun et al., 1997). They found that during SWS there was an overall deactivation of the brain compared to wakefulness, including areas of the midbrain, limbic areas, and higher cortical areas such as the dorsolateral prefrontal cortex (dlPFC). In contrast, during REM sleep there were increases in general brain activation over that seen in SWS. Specifically, these increases were evident in the brainstem, basal ganglia, hippocampus, and nearby limbic areas, as well as in cortical areas.

Interestingly, these levels were comparable to, or even greater than, levels seen during wakefulness. Others have corroborated this evidence for increased activity of such structures during REM, especially for the amygdala Opens in new window, hippocampus Opens in new window, and cortex. Indeed, this activation pattern may underlie REM sleep’s role in emotional memory enhancements, as these regions are important for processing such memories (Maquet et al., 1996). Further, there is a common finding among these studies that the activity of the executive processes associated with dlPFC is reduced in REM sleep, which may, in conjunction with increased activity of the emotional regions of the brain, contribute to the peculiar, disjointed, and emotional content of dreams (Braun et al., 1997; Maquet et al., 1996; Stickgold et al., 2001).

The electrophysiology of sleep

Just as there are different neural activation profiles associated with different stages of sleep, so too are there different neural electrical signals which distinguish one stage of sleep from another, and these unique characteristics may hold an important key for each stage’s impact on memory. For instance, SWS is associated with an assortment of specific electrophysiological characteristics arising both from cortical structures and from the hippocampus Opens in new window. These features include:

• slow, low-frequency (0.8 Hz) EEG activity known as slow oscillations,
• higher-frequency (12-15 Hz) sleep spindles, and
• extremely high-frequency (200 Hz) bursts of activity within the hippocampus known as sharp wave ripples (De Gennaro and Ferrara, 2003; Siapas and Wilson, 1998; Steriade, Nunez, and Amzica, 1993).

All of these signatures have been implicated in learning. For example, sleep spindles have been implicated in assimilating new and old information (Tamminen et al., 2010) and have been proposed to play a role in long-term potentiation (LTP), a process of brain plasticity thought to be important for learning (Rosanova and Ulrich, 2005).

Additionally, slow oscillations have been suggested to power sleep-dependent hippocampal reactivation of memories that gradually transfer these traces to cortical long-term storage (Wagner and Born, 2008).

Further, both slow oscillation amplitude and sleep spindles have been seen to increase during sleep following a period of learning, for both declarative (Gais et al., 2002); Mölle et al. 2009) and procedural tasks (Fogel and Smith, 2006; Huber et al., 2004). Thus, these unique features of SWS, instead of SWS as a whole, could be responsible for a variety of memory enhancements, likely by working to transfer information into long-term storage (Born and Wilhelm, 2012).

Similar to SWS, key electrophysiological characteristics of REM can also be identified that may be sufficient for memory enhancements. For instance, evidence has begun to point toward an important role for REM theta EEG rhythms (4–7 Hz) in memory improvements for a variety of tasks (Fogel, Smith, and Cote, 2007; Jones and Wilson, 2005; Nishida et al., 2009).

Specifically, Fogel, Smith, and Cote (2007) had participants train on a word-associates task for a later memory test before undergoing a night of PSG-recorded sleep. A week later, participants again trained and were tested on the word pairs learned during the first session. The authors found participants to exhibit increased theta rhythms during REM after learning the word pairs as compared to a baseline night in which no learning took place. Further, the authors found that REM sigma rhythms (12-16 Hz) were also increased after learning.

Although REM is typically related to emotional and procedural memory Opens in new window, the contribution of such REM signatures to this declarative task highlights the importance of looking beyond simple sleep stages to the mechanisms that may be driving these effects. Theta in particular has been previously implicated in hippocampal LTP (Larson, Wong, and Lynch, 1986) and thus may provide a more specific sleep-dependent mechanism for memory formation at the neural level, as opposed to attributing memory effects to a broad sleep stage per se. This feature of REM has also been related to other hippocampus-dependent memories, such as those involving emotional content (Nishida et al., 2009) and spatial learning tasks (Jones and Wilson, 2005).

Changes in neurochemicals during sleep

It has further been suggested that the chemical milieu of the sleeping brain may have an extensive impact on how the brain processes information during sleep. NREM and REM sleep can be differentiated by their varying profiles of neurotransmitter and neurohormone composition.

For example, while NREM sleep is characterized by reductions in aminergic (norepinephrine, NE; serotonin, 5-HT) and cholinergic (acetylcholine, ACh) inputs from waking levels, REM sleep exhibits an even further reduction in aminergic tone, but an increase in ACh that is near or above levels seen at waking (Stickgold et al., 2001).

Critically, some of these neurochemicals have been explicitly implicated in declarative memory Opens in new window enhancements. For example, when levels of ACh are experimentally increased during “early” sleep dominated by SWS, the typical performance enhancements on a word-associates tasks observed after this portion of the night are diminished (Gais and Born, 2004).

Alternatively, elevated ACh has been shown to play an important facilitative role in emotional memory formation, especially because of its interaction with the amygdala Opens in new window (McGaugh, 2004). These features of Ach may explain why SWS appears to aid declarative memory Opens in new window when ACh levels are low, and REM appears to aid emotional memories Opens in new window when ACh levels are high.

Further, the glucocorticoid cortisol, a stress hormone arising from the activation of the hypothalamic-pituitary-adrenal (HPAP axis, has been shown to display significant circadian variation and has been implicated in memory processing. Specifically, although excessive elevations of cortisol hinder both LTP and declarative memory, moderate elevations enhance LTP and declarative memory (de Kloet, Oitzl, and Joёls, 1999; Payne et al., 2004). Further, receptors for this hormone are densely located within the hippocampus Opens in new window, which serves as an important element of the negative feedback loop for the HPA axis (de Kloet, Oitzl, and Joёls, 1999; Sapolsky, 2004). Cortisol is lowest in the beginning of the night when SWS is more prevalent, but increases occur throughout the night with peaks that track periods of REM until it is at its highest upon awakening (Dallman, Bhatnagar, and Viau, 2000; Fries, Dettenborn, and Kirschbaum, 2009; Wagner and Born, 2008).

Not surprisingly, cortisol has been found to have a significant interaction with sleep-dependent learning. For example, Plihal and Born (1999b) found that increasing cortisol levels during the early SWS-dominated portion of the night led to performance impairments on a subsequent word-associates memory test, but not a mirror-tracing test.

These results are crucial for understanding how sleep can optimize memory via cortisol. A low glucocorticoid environment, such as that observed during SWS, appears to be necessary for enhancements of declarative memories Opens in new window that rely on the hippocampus Opens in new window, like the word associates task.

Conversely, procedural memories Opens in new window that are independent of the hippocampus, like the mirror-tracing task, are unaffected by glucocorticoids and can thus be optimized without difficulty under the elevated cortisol concentrations of REM sleep.

This finding provides yet another example of the way physiological characteristics of the sleeping brain, as opposed to specific stages of sleep per se, can give rise to memory enhancements. However, it is important not to overly attribute such enhancements to any one of the physiological features discussed, as they likely all work concurrently to provide the optimal processing atmosphere for various memory systems at different phases of sleep.

Summary

Many studies show a relationship between memory enhancements and specific stages of sleep. However, as described in this entry, there are numerous neural events occurring in tandem with such stages that also affect memory performance.

For example, effects of REM sleep on memory could be caused by a multitude of the changes seen in brain physiology during this stage, including increases in activation of pertinent brain structures, theta rhythms, ACh, and cortisol.

Similarly, the brain structure deactivations, slow oscillations, spindles, sharp wave ripples, decreases in ACh, decreases in cortisol seen during SWS could individually or collectively have an impact on the memory enhancements found to be related to SWS.

Thus, while findings from the sleep and memory literature often point toward an entire stage of sleep as having a relationship to memory improvements, it is also important to consider the neural mechanisms associated with these stages that may be more directly involved in memory consolidation Opens in new window.

Conclusions

Sleep plays a critical role in memory processing. The dynamic environment of the sleeping brain provides the means necessary for enhancing a variety of different memory systems, including declarative, procedural, and emotional memories.

While SWS has been largely associated with declarative memory performance and REM sleep has been related to improvements in procedural and emotional memory domains, there are also many learning tasks that do not fit such a mold.

Further, it is also important to remember that many of the mentioned memory enhancements may be the result of the neural environment, including the physical, electrical, and chemical aspects of the sleeping brain that generate and underlie these dynamic stages of sleep. Although there is still much to learn about why we spend a third of our lives in such a vulnerable state, it seems very clear that sleep is critical for memory and learning.