User:Kinkreet/sandbox2

Overview
Memory consolidation is the idea that memories and experiences formed during waking periods are stabilized and enhanced during sleep.

This hypothesis relies on the notion that memories are not formed as a single event, but rather are processed through different stages. First, there is the acquisition phase, where the experience is encoded somehow into a form the brain can process and store. Then the memory is processed further, primarily through the process of consolidation. The process of consolidation is further subdivided into stabilization followed by enhancement. Stabilization is where the memory is prevented from interference (from new experiences) or decay (or become forgotten), whereas enhancement takes the stabilized memories and improve its recollection beyond the time of acquisition.

Stabilization has been shown to occur both during wake and sleep, whereas enhancement is known to occur only during sleep. This section will thus focus on the enhancement part of memory consolidation, since it appears to be sleep-dependent. At the end, I will outline reasons why memory enhancement might be an essential process, and not just a bonus that came out of the evolution fruits machine.

Different types of memories
Like sleep, memory is not a homogeneous state, and is commonly categorized into declarative and procedural/non-declarative memory. Declarative memory describes memory which are can be consciously accessed (fact-based), such as personal experiences (episodic memory) and facts (semantic memory); procedural memory describes things which are unconscious, such as learning a skill, or understanding the grammar behind a language.

Consolidation of declarative memory have often reached mixed conclusions. While both types of memory have been shown to be consolidated after sleep,  the evidence on the procedural memory is stronger.

Procedural memory
Brawn et al. proposed that players of first-person shooter games (FPS) such as Unreal Tournament and Quake requires primarily procedural memory, as players must develop the skills to incorporate and process a mixture of visual, auditory and spatial inputs, and produce outputs in the form of motor movement (pressing buttons on controller).

Subjects were first familiarized with the controls in the game Unreal Tournament 2003 for 2 minutes, after which they were tested for 7 minutes in a different environment and the score was recorded. Subjects are then trained on a different game with similar principles (Quake 3), in four different environments, each lasting 7 minutes each. A post-test is carried out after a certain amount of time after the last training session and the score of this post-test is recorded and compared to the pre-test score to get an indication of how well the procedural memory is retained.

Subjects were trained either in the morning (at 9am), or in the evening (at 9pm). When the subjects were tested straight after training (control group), their score improved by 8.63±1.62 and 7.71±1.23. When subjects trained in the morning are re-tested after 12 hours of wake (at 9pm), their score only improved by 4.28±1.08, showing a decay in memory. However, different subjects who was awake after training but had a night's sleep sees their score improve by 9.81±1.29, non-significantly higher than the control groups. This shows that sleep is able to enhance memory, even those that have decayed during the day.

This studied also showed that memory that was consolidated during sleep is more stable. Subjects that were tested in the evening and allowed to sleep before being tested 12 hours later, showed enhancement of 10.34±1.55. A different group that were trained in the evening but was tested 24 hours later, showed a 9.95±1.86 improvement on the score, meaning the consolidated memory is not subject to the same level of decay as seen prior to consolidation.

This study can be deemed reliable due to a large sample size (results from 178 subjects were used) and the use of established protocols (based on a previous study by the same laboratory). However, its conclusions cannot be definitive because the sample are all from University of Chicago students with no more than 10 prior experiences of playing FPS games. This is a problem because 1) the age group of the subjects were confined to ~19.9 years old, and younger/older subjects might show different results, and 2) "no more than 10 prior experiences playing FPS games" means it is not a truly novel experience, and there were no way in ensuring the students were being truthful.

Although this study is by no means definitive, it supports other studies which have shown sleep-dependent enhancement in perceptual learning of synthetic speech, .........

Purpose of sleep
Previous studies have shown that new memories have a time period of approximately four to five hours where it is vulnerable to interference from more recent memories. During this period, the memories are being stabilized. Stabilization have been proposed to work by strengthening relevant associations and weakening irrelevant ones. During sleep, the sensory inputs are dimmed and this prevents memories from interference, to prevent 'false' memories from being incorporated, ensuring the integrity of the memory.

Enhancement appears to occur exclusively during sleep, and is proposed to work by taking partially retained memories and cognitive maps, and rebuild said memories and maps. Indeed, there are evidence to support that memories developed during the day are replayed during sleep. Subjects trained on a serial reaction time task had their brain activity measured (using positron emission tomography (PET) imaging) during the task as well during sleep; results showed that regions of the brain that was active during the tasks were significantly more active during REM sleep. The same group later found out that the extent of the replay correlated with the quality and strength of the memory developed, suggesting the similarities in brain activity during REM sleep is due to the processing of these memories.

Replay is also seen in songbirds at the single-cell level. The firing patterns of neurons in the forebrain regions associated with learning vocalizations during singing, closely resembled that during sleep.

These memory consolidation effects appear to be independent on the circadian rhythm, as a 60 - 90 minutes nap has shown to be as effective as an 8-hour nocturnal sleep for learning a perceptual skill, and sleep during the day proved as effective as a night's sleep for a finger-tapping task.

Further evidence that sleep is essential for memory consolidation is the phenomena that learning new information might actually increases the amount of sleep required. For example, learning a foreign language increased the amount of REM sleep of the learner, and the subsequent success of the learning correlated with the amount of REM sleep. This increase in sleep might be a homeostatic response to the demand for a particular stage of sleep. This has also been shown in animals, where animals who learned quickly had the biggest change in sleep structure. In flies, it has been shown that flies which had a more intense or complex social experience requires more sleep after.

Sleep inspires insight
Researchers took 66 subjects and asked them to carry out a Number Reduction Task. In this task, subjects are given an eight-digit string of numbers, which consists of only three types of numbers - '1', '4' and '9'. They must process this string sequentially using paired digits stemming from left to right, using two simple rules. Rule 1 states if the pair of digits are the same, the output is the same as the digit, and rule 2 states if the pair of digits are different, the output is the other number not in the pair. This output is then used to pair with the next digit on the string, until we reach the last output, which is the final solution.

The important part of this task is that there is a hidden rule - that subjects do not need to complete the 8-digit string to get to the final solution, but the strings are designed so that the second output is always equal to the final solution. Subjects who realizes this hidden rule will be able to complete the task much faster, as they only need to do two iterations of the rules rather than seven.

Three task blocks (each consisted of 30 strings) were carried out. The subjects are then required to have 8 hours of nocturnal sleep, nocturnal wakefulness, or daytime wakefulness. After this 8 hour period, the subjects carried out the task again, this time lasting ten task blocks. The number of people that gained insight into the hidden rule (detected by their quickened elucidation of the final solution, as well as a questionnaire asking them about the hidden rule) is substantially higher in subjects who received 8 hours nocturnal sleep (59.1% as opposed to 22.7% in the wake groups).

To ensure it is not the effect of sleep on attention that is responsible to the gain of insight, supplementary experiments were carried out where the tasks were carried out as 13 continuous task blocks, in the morning after sleep, or in the evening after daytime wakefulness. Both groups had 25% of its subjects gaining insight. This shows that the gain of insight requires processing of previous experiences in the brain during sleep, and not as a result of sleep on improving attention.

Mechanism
Brain plasticity is the permanent structural and functional changes in neural pathways and synapses as a response to stimuli. Therefore, if sleep-dependent memory consolidation is true, then we should observe sleep-dependent neuroplasticity.

Using PET scans to image brain activity, it shows that after learning a motor skill, subsequent use of that skill employs different parts of the brain dependent on the time after the completion of practice. Brain activity shifts from the prefrontal regions of the cortex to the premotor, posterior parietal, and cerebellar cortex structures. It has been proposed that this shift is part of the consolidation process whereby the representation of that skill is changed into a more stable form. The primary motor cortex (M1) is critical in consolidation of newly acquired motor skills.

It has been shown that procedural memory enhancement last for a relatively long period. In a human study (n=25), subjects were trained on a pursuit task where only the horizontal trajectory is predictable. One half of the subjects were deprived of sleep the first night after training, whereas the other half were allowed to sleep. The subjects were then retested on the same task three days later. The three-day gap is to account for undesirable effects in attention-retention due to sleep deprivation, which is an undesirable factor in the experiment. Subjects which were allowed to sleep showed improvement whereas the sleep-deprived group showed no improvements. Using functional MRI (fMRI), they've also shown that brain activity in the superior temporal sulcus (STS) were more active in the sleeping group. This study shows that the sleep in the night immediately after memory acquisition is essential in memory consolidation, and this consolidation involves changes in the neural pathways that is retained after three days.

Different types of sleep
Sleep is not homogeneous - there are different types of sleep. In mammals, sleep is categorized into two broad categories - rapid eye movement (REM) and non-rapid eye movement (NREM) sleep. In humans, NREM sleep is further subdivided into NREM-1 to NREM-4, with the former representing the lightest sleep, and the latter the deepest sleep. Human goes through a few alternating cycles of REM and NREM sleep per night, each lasting around 90 minutes. Normally, a night's sleep will consists of four REM stages, each one longer than the subsequent one. Although the length of each cycle remains roughly the same, NREM-3 and NREM-4 dominates in the early stages whereas NREM-2 and REM sleep dominate the latter periods of the night.

Declarative memories appear to be unaffected by REM sleep, whereas procedural memories are. While sleep immediately or soon after learning a skill is essential for its consolidation, different types of memory may be improved during different stages of sleep. For example, formation of procedural memory for visual texture discrimination are enhanced after early (slow wave) sleep, rather than REM sleep. However, a previous study have shown that REM sleep is essential for visual discrimination learning, and indeed both types of sleep might both be required for memory consolidation for visual discrimination skills. For non-declarative motor skills, both REM and NREM-2 sleep late in the night is important for consolidation. In animal models, REM sleep, and more specifically the P waves in REM sleep, is responsible for learning of an avoidance task. An injection of carbachol (a cholinergic agonist which binds and activates the acetylcholine receptor) into the P-wave generator of rats prevented the lack of retention of the procedural memory from REM-sleep-deprivation.

Objections
Although sleep-dependent consolidation of memory is widely established, there remains some objections to this hypothesis. One of the most popular is that sleep-deprivation not only deprives the subject of sleep, but also induces additional stress. Recently, researchers have developed flies which can be induced to sleep using a temperature-gated nonspecific cation channel Transient receptor potential cation channel(UAS-TrpA1), and shown that this induced sleep (which the lack of do not provide additional stress) boosted memory of learning the flies undergone prior to sleep.

Synaptic homeostasis model
During wake, as we experience or learn novel things, the synapses between the neurons involved are strengthened, mostly through synaptic potentiation. This increase in synaptic number and strength can mean synapses are more frequently stimulated, more dendritic receptors, more neurotransmitters etc. This requires more energy, more space and prevents further potentiation, leading to saturation and an inability to learn further. This hypothesis suggests that sleep functions to clear unused or irrelevant synapses, or their potentiation, to make way for new ones.

This hypothesis is supported by behavioral studies in fruit flies, rodents and humans, which suggests that the more complex or intense an experience/training/social situation is, the more sleep that is required. EPSCs, which measures the depolarization of postsynaptic membrane potential after receiving a signal from the presynaptic neuron, showed that the frequencies and amplitude of these depolarizations increases after wake, and decreases during sleep. This supports the fact that synaptic strength is built up during the wake state, and is decreased during sleep - consistent with the hypothesis. Furthermore, the overall level of synaptic proteins in all major parts of the brain decreases gradually during sleep.

The hypothesis also states that slow-wave sleep (NREM-3/4) are responsible for this 'synaptic downscaling'. SWS is dominant during the first half of a normal sleep cycle, and REM sleep dominates the latter half. Since REM sleep have been implied in memory consolidation, the process of clearing out irrelevant connections prior to them being consolidated makes sense.

The synaptic homeostasis hypothesis proposes the following: During wake, we interact with the environment and acquire information which we process as memories. During wake, high levels of noradrenaline (NA), phosphorylation of cAMP-responsive element-binding protein (CREB), likely leads to the up-regulation of long-term potential (LTP)-related genes such as Arc, Brain-derived neurotrophic factor (BDNF), nerve growth factor-induced gene A (NGFI-A), Homer, Neuronal activity-regulated pentraxin (Narp).

These changes favors the storage of newly-acquired information by increasing the strength of the synapse. This increase in the number of synapses and synaptic strength requires increased demands of energy and space, a continuous increase leads to saturation of neural connections and prevents further learning.

Sleep, and more specifically slow-wave sleep (NREM-3 and NREM-4) is proposed here as a homeostatic response to proportionally down-regulated the strength of all synapses of a particular neuron, desaturating neural connections, and ensure efficient use of space and energy, while preserving memory.

The link of SWS to synaptic downscaling is proposed because of evidence which shows a positive correlation between synaptic potentiation during wake, with the amount of sleep during sleep; and that SWS sleep coincides with downscaling.

Slow-wave sleep
The occurrence of SW are due to the increased synaptic strength developed during wake. Because the neurons are highly connected and potentiated, the firing of one neuron will lead to the firing of the downstream neurons in its circuit, leading to synchronizing and a large amplitude. Stronger connections have also been linked to a prolonged intermediate phase between the hyperpolarized and depolarized states, lowering hte frequency of oscillation.

The synaptic homeostasis hypothesis is support by anatomical studies which shows global increase in synaptic density in animals which are experiencing an rich environment. This is likely to be a local effect, as stimulating the whiskers of mice led to the selective local increase in the formation of dendritic spines with GABAergic synapses in the corresponding barrel cortex.

Slow wave sleep may downscale synaptic density by down-regulating LTP-related genes, maybe as a consequence of lowered noradrenaline levels.

Animals which have lesioned noradrenergic system, and thus unable toactivate LTP-related genes, had a dampened SWA. And so SWS is likely a homeostatic response to increased synaptic potentiation. The noradrenergic system is unlikely to affect SWS, as during normal sleep the level of noradrealine is markedly decreased.

The environment of the brain during sleep is also different, there are less acetylcholine, noradrenaline, seratonin, histamin and BDNF. Furthermore, this reduction is likely to involve dephosphorylation and subsequeny internalization of α-N-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptors. In contrast, the level of molecules implicit in synaptic depression, such as calcineurin, protein phosphatase 1, meatbotropic glutamate receptor subunit 2, molecules involved in vesicle recycling (e.g. N-ethylmaleimide-sensitive factor (NSF)), and insulin are upregulated.

As synaptic strength decreases, the synchronisity and amplitude of the oscillations decreases, represented as a deviation from SW, and a more desynchronized waves characteriztic of REM sleep.

Synaptic homeostasis appears to be a very important process, as SWS deprivation during the first 3 hours of sleep led to an increase in SWA in the sleep that follows.

Conclusion
Humans are unique amongst the mammals largely because we have a large cortex. This cortex is involved in learning and memory, which requires permanent changes in the structure and function of the brain - a process known as brain plasticity. It is likely that any experience we have during wake, which are perceived and processed in the brain, will lead to plastic hcnages in the brain. Not all of these experinces are useful, or indeed make any sense. Thus sleep is a homeostatic mechanism to prune away irrelevant connections, as well as to make sense of existing experiences.

The need for sleep for synaptic downscaling cannot be dismissed as redundant to mechanisms in circadian rhythm. Neither can effects o sleep on memory consolidation, as 30-90 minute nap has been shown to be as effective as a whole cycle of noctural sleep.

Instead, circadian cycle and sleep should be viewed as two distinct, but related, process, which occur parallel to each other, with the circadian component regulating the time to fall asleep, and the homeostatic component regulating the amount of sleep.