does sleep disarm dangerous emotions?
Imagine you witnessed a terrible accident earlier today. Two cars slammed into each other just ahead of you on the motorway—and nobody got out. You rushed to help and managed to extract a little girl from the back seat of one car, but she was unconscious and bleeding. You did everything you could, but when the paramedics finally got there, they said it was no use—she had died in your arms. This was a scarring experience, which left horrendous images in your mind that just won’t go away. You’ve talked to a few friends, and they calmed you down a bit, but all the conversations finished with useless suggestions like, “Get some rest and you’ll feel better.” Are you really going to feel better about this horrific event after a night of sleep? One of the new theories in the study of sleep and memory suggests that you will. The overnight therapy hypothesis suggests that active processing during REM sleep causes negative memories to lose their bite.1 We are all familiar with the idea of feeling better in the morning, but is it really true, and could it work for these horrible images you have of the crash?
Let’s back away for a minute and take a broader look at how emotions work in the brain. Once this is clear we will evaluate evidence for and against the concept of overnight therapy.
The almond-like amygdala located deep in the temporal lobe of your brain can, very loosely, be thought of as a fear detector (although it also responds to other emotions such as anger and even happiness). It reacts very quickly to emotional situations, drastically increasing neural firing rates as soon as you experience something scary or upsetting. Although this little structure has a massive influence on what happens next, in terms of physiological responses, it isn’t the only part of the body which reacts early. When you’re really frightened by an outside stimulus, many bodily systems engage quickly in preparation for fast action. This is often called the “fight or flight” response because it gets you ready for action, whether by fleeing the scene or standing up to the danger. Blood is pumped from the skin and organs to the muscles, priming them for movement; pupils dilate, presumably to allow more light in; the heart pounds and hairs can stand on end; and the interesting thing is that all of this happens before you are even aware of the emotion.
There is evidence that you don’t actually feel afraid (happy, sad, etc.) until your brain has picked up on these various bodily responses and interpreted them. If this is correct, then emotion actually results directly from a sort of read-out of the body’s state.2 A really fascinating perspective on this is the idea that nothing can feel scary if the body doesn’t respond to it. That might sound strange, but this idea is precisely what forms the basis of the overnight therapy hypothesis.
So how could overnight therapy work? The increased heart rate, clenched teeth, tightened buttocks—whatever you consider as part of the unconscious bodily response to fear—is largely controlled by a neurotransmitter called norepinephrine. Without norepinephrine your body just wouldn’t have those terror-inspiring responses even in the scariest situations. And here’s the trick: Brain levels of norepinephrine reach an all-time low during REM sleep. So, if memories are replayed during REM, then no matter how scary they are, they won’t be associated with the normal bodily response because there simply isn’t enough norepinephrine around to elicit these. In other words, when you’re in REM sleep and dreaming of a frightening event, your emotional system cannot respond normally. The overnight therapy hypothesis proposes that, even though the content of the memory may be strengthened through this type of emotion-free replay, its emotionality could be completely lost (if it is replayed often enough, that is).
Overnight therapy is attractive as an idea not just because of its conceptual tidiness but also because of vast bodies of anecdotal evidence. We have all experienced the therapeutic effects of a night of sleep when we are upset, and we all know it is often better to “sleep on” a disagreement before coming back to it. Sleep undoubtedly helps to soothe frayed nerves, calm tempers, provide perspective, and generally alleviate difficult emotional situations. This might seem like the ideal way to process traumatic events.
Nevertheless, there is plenty of evidence that emotional memories are often preferentially strengthened during sleep (see chapter 9). Proponents of overnight therapy neatly sidestep this issue by pointing out that upsetting events may indeed be better remembered after sleep—but they simply won’t be associated with the same kind of emotional response postsleep as they would have elicited had they not been processed through sleep. The idea is that emotional content is essentially decoupled from the memory as a result of replay in REM sleep. This theory is supported by a recent study in which volunteers were shown an array of pictures, some of which were deeply disturbing scenes of accidents, surgery, disfigurement, and the like.3 Some were neutral scenes, such as landscapes or pictures of rooms or people at leisure, and some were happy scenes of children playing, kittens or puppies, mouth-watering foods (chocolate cake, ice-cream sundae), or couples kissing. People were asked to rate each image for emotional intensity both before and after a night of sleep. Fascinatingly, the results revealed that they felt the exact same images were less emotionally intense when seen again after sleep. Responses of neurons in the amygdala to the pictures were also markedly reduced after sleep, and (perhaps most interestingly) these changes were predicted by the neural firing which occurred at frequencies between 25 and 100 Hz during REM sleep. Because neural firing at this frequency (called gamma-band) is a marker of the concentration of the neurotransmitter noradrenaline, this correlation shows that the less noradrenaline there was in the system during REM sleep, the more the pictures lost their emotional impact. The same ratings of emotion were taken before and after a day of wake and revealed no specific change in how people perceived the pictures.
One really fascinating angle of the overnight therapy idea relates to pathological conditions like posttraumatic stress disorder (PTSD). Soldiers coming back from battle and people who have witnessed horrific accidents are just two of the groups that often suffer from this disorder. The flashbacks they experience can come at any time, they may have trouble sleeping, and the rest they do obtain is often haunted by horrific dreams about their experiences. Simply put, PTSD is the repeated, intrusive remembering of highly arousing (and upsetting) negative memories, and the consequences can ruin marriages, destroy lives, and lead to long-term depression and even suicide. If REM really does serve to dissociate memories of horrific situations from the emotional responses that originally accompanied them then it clearly isn’t doing its job properly in people with PTSD. Something has gone wrong with the system.
As mentioned earlier, sometimes the best way to verify a neurological process is to see how people who lack one of the parts of the equation fare. There are people who lack the ability to experience normal REM sleep. This is referred to as disregulated REM sleep, and those individuals are at greater risk of PTSD.4
We mentioned that one of the reasons it is good to relive the scary events through dreams during REM is the lowered level of the neurotransmitter norepinephrine. It turns out that higher than normal levels of norepinephrine during REM are linked to high risk for PTSD.5 This fits into the overnight therapy idea like a perfect puzzle piece since having more norepinephrine in the system means that unconscious bodily responses to emotion (such as a faster pulse or dilated pupils) are not reduced. Abnormally high norepinephrine in REM could therefore prevent the decoupling of emotional content from memories when they are reactivated. But hold on a minute. Are we really debating the question of whether replaying a memory with or without autonomic responses could make a difference to what you remember later? How can merely replaying a memory change it for good, irrespective of whether or not that means removing emotionality? After all, aren’t memories reasonably fixed and solid?
Reconsolidation
To answer this question we need to back up and talk about a concept called reconsolidation. Memories evolve across time and sleep. The way they are represented in the brain changes, the way they integrate with other memories and with general knowledge changes, and of course they may also be forgotten. Whether or not we can influence or control this evolution of memory poses a really tantalizing question. Imagine how fantastic it would be if you could shape your memories just as you like them (this might not lead to accurate memory, but at least you could have a good time—after the fact that is—and maybe even boost your self-esteem to boot).
Reconsolidation is the idea that memories become flexible and fragile every time we use them, and as such, it offers a potential mechanism by which we can modify them in a semicontrolled manner. To understand reconsolidation, you almost need to think of memories like library books which get stored away somewhere in the deep, dark depths of your brain for years at a time and don’t change much once they’re stored (except for a gradual rotting and moldering and also an ever-increasing possibility that you won’t be able to find them when they’re needed). Other than these minor dangers they are pretty safe while in the book-stacks. Once they are called back and brought out for use, however, these books are vulnerable. Sometimes they are slightly rewritten or scribbled on, sometimes they are grouped with related books before they are put back in storage, and sometimes they are damaged or lost. Reshelving is an active process, and messing this up can be so disastrous that these memories are lost completely—for instance, if you don’t have the resources to reshelve them or if you somehow put them in the wrong place. Two components of this analogy—the idea that memories are flexible once they have been retrieved, and the idea that storing them again is active and can be disrupted—capture the essence of reconsolidation.
This phenomenon of memory lability has been studied extensively in rats. If these furry creatures learn an association—for instance, between a beep and an imminent electric shock—they normally remember it for months (so long as they don’t hear the beep without the shock too many times, that is). One clever experiment used this type of memory to study reconsolidation.6 Two groups of rats learned to associate a specific sound, called the CS or “conditioned stimulus,” with a shock, called the US or “unconditioned stimulus” (Fig. 25.) The fact that they had learned this properly was obvious because every time they heard the beep they froze in fear of the imminent pain. Anisomycin, a substance that prevents cells from creating the proteins needed for consolidation, was injected into the rats’ amygdalas 14 days after initial learning. One group of rats heard the beep again once (but without a shock) about four hours before the injection (top). The other group heard nothing (bottom). Twenty-four hours after the injection, all the rats were tested to see if they remembered that the beep was scary. The rats who hadn’t heard it since training were just as scared as ever. Amazingly, however, the rats who had heard the sound before the injection ceased to associate it with the electric shock. They showed no sign of fear when they heard the beep on day 15. This acquired amnesia didn’t happen if rats were not injected with anisomycin, so it wasn’t simply a matter of hearing the beep without the shock and thus learning that it wasn’t scary.
Fig. 25 Reconsolidation of conditioned freezing in rats
So why the difference between these two groups of rats? Could hearing the beep just prior to the anisomycin injection really have had such a big impact on what was remembered? Karim Nader and colleagues from McGill University, who conducted this research, think it could. They suggest that hearing the beep caused the memory to be retrieved (just like getting that book out of the archives), but when the rat tried to put it away again, he couldn’t because consolidation requires the construction of new proteins and the anisomycin injection prevented this from happening. Basically, the injection prevented the book from being reshelved, and therefore the memory got lost and was forgotten. The group that didn’t hear the sound before the anisomycin injection didn’t have this problem because their memory was never retrieved: It was still safely stored away in the archive, so the anisomycin had no impact on it at all. It was this surprising observation—that, once reactivated, memories have to be actively processed if they are to be remembered later—that led to the concept we call reconsolidation.
In rats at least, memories appear to be somewhat fragile after they’ve been retrieved. But why is this important, and what does it mean for humans? It may be that this memory lability is important because it gives us a chance to change memories, and sometimes that is essential. We often want to update our knowledge (imagine a social situation in which two friends were a couple for three years, but now they have separated and she is seeing someone else), link previously unrelated concepts together (she is actually seeing a colleague you knew from a completely different social circle, so she has become part of that group as well), and sometimes even remove unwanted components (for instance, the strongly negative emotion associated with really horrific memories like the one you formed of the little girl dying in your arms). It is for the removal of unwanted information that the reconsolidation concept has proven really useful. This is because it looks as though reconsolidation can be used to selectively wipe out the most negative aspects of really disturbing memories.
As a matter of fact, clinicians have even started using reconsolidation as a treatment for PTSD. Treatments of this type usually rely upon a combination of REM-like eye movements and talking therapy in which the patient imagines the traumatic scene they are trying to get rid of. It isn’t entirely clear what the eye movements do in this therapy, but some argue that they help to minimize physiological responses associated with the emotions in the memory. In this way, participants are able to call back a traumatic memory without evoking the associated autonomic responses (just as they would have done by replaying it in REM sleep, where norepinephrine levels are low), this means newer consolidation can replace the old memory with a less emotionally charged version. Although the connection between eye movements and reduced physiological response remains somewhat murky (and in fact many people argue that the eye movements are unnecessary), this treatment is surprisingly effective, with just a single session completely curing profound PTSD in some cases. Such results provide convincing evidence that reconsolidation can alter human memories and specifically the traumatic memories which cause problems in PTSD.
What does reconsolidation have to do with sleep? There is actually a strong link here. A study by Matt Walker and his colleagues at Berkeley showed that retrieving memories before sleep can influence the way they are consolidated during subsequent snoozing.7 Instead of injecting a protein synthesis inhibitor, this study used interference, or learning another memory which is very similar but not quite the same as the original one, as a way of disrupting the initial memory. The paradigm was as follows: On the first day, people learned to tap their fingers in a particular sequence (let’s call it sequence A, 4–1–3–2–4, for example, if the fingers on one hand minus the thumb are numbered 1 to 4). People had to tap out this sequence as fast as possible. They were given time to practice this before being tested to see how fast they could do it. You might remember from chapter 1 that if people are allowed to consolidate this type of sequence overnight they get faster at it—up to 20 percent faster, in fact (Fig. 26a). People in this study didn’t just learn sequence A; they learned a second sequence as well (let’s call it sequence B)—imagine this was 3–1–4–2–1, for instance. The problem here is that, if the sequences were learned one after the other, then the second sequence interfered with the first, such that memory for the first sequence didn’t improve overnight. However, if sequence A is learned on day 1 and sequence B is learned on day 2, then on day 3 people show improvement on both sequences (Fig. 26b). Here comes the trick (and the link to reconsolidation). If sequence A is learned on day 1 and practiced just once on day 2 right before sequence B is learned, then on day 3 sequence A shows no improvement (Fig. 26c). This might be pretty confusing when you read it, but take a look at the figure to get a better picture.
If you think about it, this is just like the experiments in which anisomycin was injected into the amygdala right after rats were reminded of fearful associations with a sound: The memory for sequence A was (however briefly) called back from the library stacks of the mind, and then, before it could be reshelved, sequence B came along and scrambled it. However, if people were able to sleep between learning sequence A and learning sequence B, there was no interference, suggesting that sleep allowed (or even facilitated) a thorough tidying away of sequence A before sequence B was learned.
Fig. 26 Reconsolidation in humans, demonstrated
using interference between two tapping sequences
The idea that sleep consolidates things such that they aren’t so easy to disrupt doesn’t just hold up for finger tapping. Another study showed similar findings using the memory task, which we talked about in chapter 6. In this task, eight pairs of identical pictures that resemble playing cards are set out in a
4 x 4 array (so there are 16 cards but only 8 different images). At the start of the game, the cards are all face down so you can only see their backs, which are all identical. The task is to collect pairs by flipping over one card and then trying to remember where its match is and choosing that card next. People who play the game gradually form a representation of where all the pictures are, so they can easily make pairs every time, and they tend to remember this better if they are allowed to sleep in between their initial attempt at the game and a next try in which the cards are laid out in the same pattern as before. This consolidation-related memory advantage can be boosted by triggering replay of memory of the card game in sleep. This can be done by presenting a specific smell (in this case a rose scent) while people play the task initially and then re-presenting that same smell to them while they sleep afterward (see chapter 12 for more on this).8
How does this relate to the reactivation of memories? A more recent study used exactly this paradigm but added cognitive interference.9 Everyone first played the card game with cards set up in array A and with rose scent in the background. Half the participants then slept for 40 minutes, while the other half stayed awake. During these 40 minutes, everybody smelled the rose odor again, which should have triggered reactivation of the memory. Next, everybody did what we call an interference task, something designed to disrupt the memories that had already been formed. They played the game again, but this time the second card in every pair was in a different location; they had to learn a whole new spatial setup, which presumably disrupted their memory for the original layout. After learning the new setup, everybody was tested on the original layout. How did performance differ between people who had slept before the interference task and people who had stayed awake? Both should have reactivated the memory representation of the first spatial layout just before they learned the new layout and thus presumably experienced interference. Fascinatingly, however, people who slept before the interference task did markedly better on the final test than people who remained awake. Just like the finger-tapping study described above, this finding suggests that sleep acts to stabilize the original memory, making it less susceptible to subsequent interference. Reactivation of the memory during that sleep doesn’t appear to make it labile in the way that reactivation during wake presumably would. Instead, sleepy reactivation appears to boost the stabilization process.
All in all, the evidence in favor of memory reconsolidation is overpowering. Memories really do become labile, and thus fragile, every single time we use them. Once in this state they can easily be disrupted, either by newer learning which interferes with them or by chemicals that prevent them from being stored (or reshelved). Reconsolidation provides the perfect mechanism for updating memories. Sleep, on the other hand, appears to be critical for “battening down the hatches,” or strengthening a memory such that it is more resistant to interference (so long as it doesn’t get reactivated in subsequent wake, that is). Critically, reconsolidation also provides the missing mechanism for the overnight therapy concept: Reactivation of memories in sleep without the associated bodily responses essentially disarms the memory, stripping it of emotional content.
Although overnight therapy is compelling as an idea and fits beautifully with the literature on reconsolidation, there is a fly in the ointment. Quite a few studies have failed to show the expected effects of sleep on emotional intensity ratings and responses in the amygdala. For instance, one study found that people rated images as less emotive after wake and observed no change in ratings of emotion across sleep.10 This finding opposes data showing that emotional images are less jarring after sleep. This negative finding is especially convincing because it supports an older study in which picture ratings taken before and after REM-poor early night sleep revealed that, rather than decreasing, the emotional responses evoked by the pictures increased over this period.11 Unfortunately, the balance of evidence seems to lean heavily against the idea of overnight therapy. Memories simply do not lose their emotionality after a night of sleep in normal healthy people. In fact, recent research in rats has shown that depriving animals of sleep for a few hours after a traumatic experience significantly reduces the probability that the trauma will be remembered later on, suggesting that sleep may actually strengthen pernicious memories in some cases.
But what about those tantalizing data described at the start of this chapter, which did show a reduction in emotionality, and in amygdala response, after sleep? These findings are real and certainly shouldn’t be overlooked. The fact is, this type of conflict in the scientific literature may be confusing, but it is also exciting—how can we explain such apparently different results?
One answer could relate to memory. In the study that showed reduced emotional reactions after sleep, participants weren’t asked to remember anything, and they were not tested on memory. On the other hand, all of the studies that showed increased emotional ratings and amygdala responses after sleep specifically examined memory. In these studies, people were presented with emotional images, or something that had been associated with these images, and asked whether they remembered them. This means people were actively trying to conjure up memories (and very likely mental images) of the pictures they had been shown. Could it be this act of conjuring which leads to the extra emotional response? After all, if people remember an image better after sleep they probably remember how they felt about it better too—but that doesn’t necessarily mean they still feel the same way as they did originally, it just means they can recollect those prior feelings more clearly. In fact, you could almost say people who are being tested for memory will be trying to re-create the original scenario, complete with a representation of the feeling that was present the first time around. This could explain why better memory after sleep is also associated with a stronger emotional response.
Another answer could relate to stress.12 A study by Hein van Marle and colleagues at the Donders Institute for Brain Cognition and Behaviour in the Netherlands showed that the extent to which emotional reactions are toned down across sleep relates directly to stress levels during sleep. This study used pictures in exactly the same way as the studies mentioned above, and participants were aware that they would have a memory test after waking up. However, in half of the people who participated, the stress hormone cortisol was artificially elevated during sleep. Although participants did not rate images for emotional intensity, the elevated cortisol changed the way negative memories were processed in sleep. Amygdala responses were increased during postsleep recognition of negative images in participants who had normal cortisol levels but not in participants with artificially high cortisol. This is a fascinating finding, because it suggests that the way sleep impacts upon emotional representations depends upon how stressed you are while you sleep. This idea fits perfectly with the literature on PTSD, since people who have abnormally low cortisol are much more likely to develop this disorder than people with normal levels of cortisol. None of the other studies discussed in this section measured cortisol levels, so it is difficult to know whether differences in this stress hormone could explain the disparate results. It is possible that participants in the study by Walker and colleagues who showed a decrease in emotional responses after sleep were simply much more stressed than participants in the other studies. After all, this work was conducted in a highly competitive university environment at Berkeley—maybe these participants were students undergoing abnormal amounts of chronic stress.
Summing up
This chapter has introduced “overnight therapy,” the idea that sleep disarms dangerous memories, helping us to cope with traumatic or unhappy situations. We looked at memory lability and how the reconsolidation of memories during sleep may allow them to be modified such that emotional content is dampened or removed. We also summarized some of the evidence which contradicts this theory by showing that sleep can in fact increase emotional responses to unpleasant pictures seen the day before. Two possible explanations for the conflicting data were discussed—one relates to whether participants were explicitly asked to remember the emotional stimuli they saw before sleep, and the other to stress levels during sleep.
Whatever the reason for the difference in opinions, it is critical that neuroscientists solve this conundrum since the overnight therapy theory suggests that people who have been traumatized should be allowed to sleep so as to dissociate emotion from the traumatic memory while the opposing view suggests that these same trauma victims should be kept awake in order to prevent negative impressions from being strengthened. Which method would you subscribe to after experiencing the traumatic traffic accident and the dying young girl?