This post is a quick (gut) reaction to an article recently revised (republished?) on Psychology Today: Dreams Decoded, thanks are due to Edward Craig for tweeting it. A large part of the article is devoted to Revonsuo’s threat simulation theory (TST), you may read his recent review of the evidence here (Valli and Revonsuo 2009). The theory itself is interesting but I honestly fail to see why it was proposed as it is, because it contains the basic concepts that may underlie an obvious and much more convincing (at least, convincing to me!) theory extension.
I’m writing this post to briefly expose this theory extension and see if someone will be kind enough to show me why my present idea isn’t new at all, or why it doesn’t stand: please comment below if that’s the case, I would be most grateful. I must also add that I’m not a sleep or dreams specialist, not at all, but questions about sleep and dreams have haunted me for years, meaning that I’ve been looking for a comprehensive and convincing theory about sleep and dreams for a long time.
The question is clear: as far as I know, all animals sleep, at least starting with insects (clearly, a minimum level of complexity is necessary to display sleep-like behaviours), and most dream, I’ve not researched this in detail, but it seems that the jury is still out on the matter of reptiles : we don’t really know if they dream. Mammals and Birds however seem to be uncontroversial: they all have dreams. Evolutionarily, one can’t avoid asking “why did sleep evolve, and why dreams emerged along the way?”. Until today, I only had fragmented and vague ideas, and reading about TST finally sparked thought that I wish to share. At least, I have an idea about why dreams emerged, and I’m hoping it’s a promising start!
TST is explained in short at the start of Valli and Revonsuo (2009):
The threat simulation theory (TST) […] claims that dreaming originally evolved as an offline simulation of the real perceptual world and that its function in the ancestral human environment was to repeatedly produce simulations of the real threatening events that had been encountered during waking hours and had left a mark in emotional memory.
Critics to the theory point out that one of the major predictions of TST is that most dreams will revolve around threats, and that the evidence doesn’t clearly support this prediction. The issue is still debated, and I’ll have more to say on this, after explaining my own little extension.
My additional idea stems directly from the Psychology Today article:
In the ancestral environment, Revonsuo reasoned, our dreams served to protect us, teaching us how to respond when a wild animal was chasing us or when we got lost in the forest. That was why the dream world was so filled with peril: to simulate the potential threats and prepare us to react quickly (emphasis is mine)
The key word here is quickly: could it be that we dream to rehearse? By rehearsing we may obtain two major and evolutionarily convenient objectives:
- We experiment on how to best react to various situations, so to find out what is likely to work in real life
- Whenever possible, we commit successful strategies into procedural memory.
By procedural memory I mean that kind of memory that allows you to perform the sort actions that are needed to drive without having to think about it. We all know that establishing this kind of memory requires repetition, and that the results are stunningly useful, even in the relatively safe world that we’ve managed to build around us. But what about being able to rehearse situations that can’t be safely repeated over and over again? These situations tend to be dangerous, hence very important for survival. This consideration supports Revonsuo’s intuition, but I see no reason why dreams should be limited to simulating dangerous situations, they could be useful to strengthen all types of procedural memory, with the added bonus of achieving also objective 1 (above).
My hypothesis therefore is that “dreams evolved as a rehearsing system, a system used to ensure that what the subject has already learned is used to facilitate/reinforce ‘appropriate’ reactions to important situations, in particular those that require a fast response“. With “already learned”, I mean two main types of information: information like “when someone does A, then B happens, unless C is true, in which case D will follow” (T1), but also things about what situations are important, require immediate responses, and therefore are likely to need rehearsing (T2).
When dreaming, the brain would build dreams around T2 information, so that the dreamer’s “dreaming consciousness” will get a chance to try out different response strategies using the T1 type of info, both to design strategies and simulate the effect of these.
If you happen to be a Freud’s fan, there you go: the symbolic value of dreams stems out of the fact that T2 type of knowledge is symbolic, so different situations will be built from the same T2 class by representing the core symbol in different ways: one night you’ll be chased by zombies, the other by a pack of wolves, the generating symbol is a pack of deadly beings that want to eat you in both cases (to use a pretty silly example).
From this perspective, PTSD recurrent dreams may be explained. Let’s consider a fairly typical combat case: a soldier in the battlefield finds himself into a “new” and very uncomfortable situation, reacts as s/he was expected to (according to the learned drills) and, according to her/his own judgement, things go terribly wrong. I’ve seen examples of this on a recent BBC Panorama episode, they included heartbreaking reports were the traumatic situation resolved in horrible ways (apologies for arbitrarily filling in my memory blanks): for example, the PTSD victim’s best mate got killed, but they also reported about the shock of seeing the effect of their own shooting on the (killed) enemy.
These are experiences of type T2: they have high chances of getting rehearsed because what happened was fast, important, and the result highly disappointing. Because the soldier does not have alternative T1 type of info, something that will convincingly lead to better outcomes, the traumatic experience gets into dreams, but keeps repeating as the dreamer is unable to generate a strategy that does lead to better results.
In the case of soldiers, they actively drill how to react, so that they will automatically function in ways that their generals like (and that they can predict); their training is designed to force them into predetermined courses of action, and this makes the dreaming system malfunction, because the virtual drills (dreams) are constrained by the previous training and get forced into an always repeating and disappointing pattern. I may argue that PTSD recurrent dreams happen to other people in the same way: when they lack T1 knowledge that can provide a good answer to the traumatic experience, and/or when the traumatic experience does not allow for good coping strategies. For example, abused children may unfortunately have no way to avoid being abused.
This conjecture allows me to address the controversy about the TST prediction on the prevalence threatening dreams. With a quick Google search I could easily find papers that support or challenge this prediction, and the ones that I could read had an interesting methodological difference:
Zadra et al. (2006) found a prevalence of threatening dreams when analysing data about recurrent dreams. Malcolm-Smith et al (2012) could not, but they were not looking at recurrent dreams at all (I was looking for references that didn’t include the original TST author). This is easily explained by my conjecture on PTSD: recurrent dreams are the ones that cover a very important (traumatic) and irresolvable situation. In other words, the “rehearsing” hypothesis that I’m proposing here predicts that threatening episodes are likely to appear more frequently in recurrent dreams. This may be supported by the available evidence as my quick check seem to suggest, but of course my opinion here is not final. I may also add that I could not find any compelling evidence against my conjectures in the actual Valli and Revonsuo review (2009). However, I need to add that they also report that:
The main weakness of the theory is that there is no direct evidence of the effect of dream rehearsal (or the lack of it) on performance or on survival rates across generations of ancestral humans.
This means that my extension also suffers from the same limitation, probably even more than the original TST. However, other conjectures are possible.
First of all, the modified theory also accounts for the fact that sleep does not involve dreams all the time, it proceeds in different (physiologically identifiable) distinct phases. Hence one could build a working hypothesis: before the dreaming phase (REM or NREM), the brain will have time to collect T2 types of information, preparing a “list of topics”. After the dreaming phase the brain may need some time to evaluate and classify the drilling results: if the dreams produced bad results, some time may be required to strengthen the importance of the T2 topics that inspired the dream (making the same kind of dream more or equally likely to happen again), if the results were good, the T2 topics will lose importance while the successful reactions would be reinforced into procedural memory.
The Rehearsing concept also explains a couple of more observations:
- One frequently dreams about recent activities, especially if they imply high motor skills that are rarely practiced. When I go out for a long ride on my motorbike I will sometimes dream of riding during the following night, but if I go on a skiing week-end, I will always dream of skiing.
- People rarely dream of tasks that are fully committed to procedural memory: in the examples above, I sometimes dream of riding, because I’m relatively good at it, while I always dream of skiing, because I do it very rarely and therefore I don’t have well rehearsed procedural memories about it.
A few more words on Valli and Revonsuo’s review (2009): most of it is dedicated to verifying the main predictions of TST, and they conclude saying that the evidence so far does seem to support their working hypothesis. I agree, but think that this confirmation isn’t neither conclusive nor surprising. By extending the scope of the theory, from rehearsing about how to respond to threats, to rehearsing about how to respond quickly to situations that are classified as important, I am not negating the importance of simulating threatening situations, nor implying that they should be relatively rare: all threatening experiences are likely to be classified as important, and therefore their relative high frequency isn’t surprising at all. However, considering that, for some physiological reason, repetition is undoubtedly necessary to store useful routines into procedural memory, the proposed extension can only strengthen the (still to be empirically validated) hypothesis that dream rehearsals have a positive effect on real-life performance. This would explain why dreaming is such a highly conserved feature across the animal world: if building procedural memories requires many repetitions, the evolutionary advantage of being able to do so without having to expose to any additional task-related risk or energetic cost, is clear and obviously not negligible. So overall, the theory does look as promising preliminary solution to the evolutionary questions about dreams (if not sleep as well).