If you are not a new comer, you probably know that sleep is one of my favourite subjects. I like sleeping, but my fascination is not the result of this fact alone. It’s more than that: neuroscience is actually getting closer to a comprehensive explanation, and what we already know, is already enough to make me declare that one of the big mysteries of human nature is not much of a mystery any more.
Incidentally, the other mysteries, as far as I’m concerned are:
- Consciousness. What is consciousness? How does it work? Why seeing red feels in its own particular way? To be fair, also consciousness is not as mysterious as some neuro-philosophers claim. In fact, I have not blogged for a couple of weeks because I was busy preparing (for peer review) a paper on the subject. You will not read much on consciousness in here until I’ll manage to get it published, it’s an ambitious work, so I wouldn’t hold my breath.
- Music: as far as I know, music is ubiquitous in every human culture. There isn’t a place on earth where humans never indulge in some form of musical activity, why is it so? Clearly, whatever the reason is, it must be an important one, but it is a puzzling phenomenon: requires time, energy, considerable skill, and doesn’t fulfil any recognised primary need. I have my ideas on this one, but they are nothing more than sketchy, so I’d better remain silent for a little longer.
- Humour: laughter itself may not be a big mystery;uttering a friendly, playful noise is a way of communicating your intentions and in fact laughter-like vocalisations are not uncommon in other species. However, humour, the cognitive mechanism that makes us laugh when we hear a joke, well, that’s a different matter. How does it work? But more, importantly, why does it happen? The more I think about this matter the more I believe that there is an important cognitive function hidden somewhere between laughter and humour. I am investigating this, and have one or two exciting ideas on how to develop the subject in here.
Anyway, let’s go back to sleep: we spend one third of our time on earth sleeping. We know that sleep plays a role in memory formation and memory management, but it is still puzzling that we need to spend so much time sleeping. In short: we know that sleeping helps memory formation and consolidation, and that this is true for implicit (procedural) memories (what you need to play an instrument) and explicit memories (what you learn at school). What we don’t know is exactly how this is done, and why sleep is divided in cycles, each characterised by a rigid succession of distinct phases. What happens in each phase? Why? And above all, why does it require so much time? What follows is my own little contribution to the last question, but in order to get there I need to summarise what we do know.
The perfect place to start is the website of the Dana Foundation, a “private philanthropic organization that supports brain research through grants, publications, and educational programs”. In fact, the Dana Foundation is such a good resource, that I will be able to build this whole post using only articles that appear there. In particular, this article (by Jim Schnabel) gives us a very good overview, the take home messages are:
SWS [Slow Wave Sleep, a non-REM phase of sleep] strengthens not only procedural memories but also place memories—learned routes within a town—as well as declarative verbal memories, such as homework-type memories of historical facts. Overall, the degree of memory consolidation seems to depend on factors such as the duration and type of sleep (REM sleep having little or no effect), the type of memory (declarative memories being helped most consistently) and the importance of the memory in the life of the person.
For a declarative, who-what-where memory—also called an “episodic” memory—the initial connections are thought to be made in the hippocampus […]
The hippocampus works a bit like an old-fashioned telephone switchboard, making quick but temporary connections […]
During SWS, these hippocampal links come alive again, apparently to consolidate the new memory. […]
One likely effect of this hippocampus-initiated memory reactivation is to strengthen the new hippocampal connections. But its more important goal seems to be the stimulation of slower-growing connections among memory-associated neurons outside the hippocampus, mainly in the neocortex. These slower-growing connections are more direct and permanent, and correspond to what we usually think of as long-term memory. [emphasis is mine]
The second bit of the puzzle can be found in an article by Tononi and Cirelli, again published by Dana.
But what are the cellular mechanisms by which sleep benefits learning and memory? Many sleep laboratories are currently trying to address this question. There are two general hypotheses, not mutually exclusive. One is that sleep further reinforces the synaptic connections that had been strengthened in wake during learning, leading to synaptic consolidation.
According to the second idea, the so-called synaptic homeostasis hypothesis (SHY) that we have proposed, sleep may be a time not so much for rehearsal, but for down-selection. […] For example, synapses that are activated strongly and consistently during sleep would be protected and survive mostly unscathed. They may actually consolidate, in the classic sense of becoming more resistant to decay and interference (though they would not become stronger in absolute terms). By contrast, synapses that are comparatively less activated would be depressed (therefore, consolidated synapses would end up stronger in relative terms). […] Clearly, learning by strengthening synapses cannot go on indefinitely – day after day – and something must be done. That something, says SHY, is the down-selection of synapses down to a baseline level that is sustainable both in terms of energy consumption and cellular stress.
They also observe that in fruit flies, it is possible to observe that during their own night-sleep the number of synaptic spines (the little protrusions where neurons form synapses) decreases to up 70% in the whole brain. I find this hypothesis very convincing, and would summarise it as follows: during sleep, we certainly consolidate memories, but we also eliminate what is not needed.
Finally, to address the question about why we need to sleep so much, I’ll use another brilliant article from Stanislas Dehaene (another scientist that, like Tononi, never disappoints): this one is not about sleep, but about learning, and in particular about learning to read. You can read it here (if you are only going to follow one link, use this one, it’s well worth your time). The reason why I’m citing this article is that Dehaene and colleagues were able to demonstrate that learning to read involves the creation of new connections between distant brain regions. This is the final block in the puzzle, which confirms what Schnabel suggest: SWS is where slow-growing, new-connections are made. Dehaene doesn’t discuss this, but it is very plausible: when we learn some new ability, conscious effort is needed, and while learning, it would be impossible to perform the desired task without giving it full attention. Then something happens, and we find ourselves able to perform the same task in auto-pilot, quite literally, at least when the task is “driving a car”. In order to do so, the brain reorganises itself, areas devoted to task A may be re-assigned to task B, new connections between areas are formed, and so on. All this is well-known, and beautifully explained by Dehaene’s work. So here is the key: such reorganisations require time, axons don’t extend centimetres in minutes, their growth speed is normally measured in millimetres per day. This means that there is no chance that the new connections may be made while actively trying to learn the new task, they may start growing at that point, but they will need more time. Crucially, this slow reorganisation may benefit from an off-line simulation that, once the new connections are established, is able to verify that they actually can do the required job. This isn’t a simple feature, and certainly requires time.
So here is my hunch: during sleep, we transfer the relevant memories from the “daily storage” (hippocampus and the like) into long-term warehouses; probably different kind of information is routed into different and specialised brain areas. At the same time, new procedural memories are (when required) re-wired, in order to acquire the ability to perform them automatically. Both processes will need two very important mechanisms:
- They will need a system to identify what should be retained and transferred to a permanent location or into a permanent new connection pattern. What remains will be flushed away and forgotten.
- While the original “to be remembered” information (procedural or declarative) is still in the daily memory, it would be useful to compare it with the new “permanent” version; otherwise, we would need to look for an error free “transfer system” that, especially when the transfer implies making new, previously non-existing connections between distant brain regions, is unlikely to be possible at all.
All this requires time, and presumably separate transfer, verify and correct phases repeated more times, so to refine the results. Voilà: you get a good night sleep, during which you may cycle through long and active “preparation” phases, where new memories are consolidated, and new connections are made, and then verify them in your dreams. At the same time, you can also use dreams to double-check the emotional relevance of what you are learning, and flush away what doesn’t seem to be interesting after all.
Yes it’s all cheap speculation, and will probably take several hundred working years of real scientists to provide a final proof or refutation of what I’m suggesting. But what can I do? I get this sort of hunches every day, and I guess that writing them down in the public domain doesn’t have any drawback (apart from the possibility of making me look silly), so here they are, my thoughts on Sleep and Dreams are explained above. Enjoy.