Hacker News new | past | comments | ask | show | jobs | submit login
Portions of the brain fall asleep and wake back up all the time (stanford.edu)
208 points by qwename on Dec 6, 2016 | hide | past | web | favorite | 96 comments

Taking some time to reflect on and consider my own thinking behavior has been interesting. Just like exploring any other system, subjecting myself to sleep stress (in order to meet aggressive deadlines) has allowed me to see what happens when certain things fail, revealing functional boundaries.

I see it as kind of analogous to when you starve a circuit/device of power - it behaves in unexpected ways, revealing implementation details.

Some examples that I remember: after a particularly long all nighter (I was a freshman in college and wanted to try it while I could still do it by choice and not necessity) I was editing an article I was writing for the school magazine. I was starting to fade and realized that I couldn't read and comprehend well-formed, meaningful sentences that I had just written.

I've noticed that when sleep deprived I notice different things and have a more diverse set of emergent thoughts/recall events. For example, today I noticed the plug for an electric oven at a restaurant I have not only been to at least 100 times, but have worked at for months. I randomly remembered the lyrics to China's five-year-plan song walking home from class. I will suddenly remember and think fragments in Spanish, despite not touching it for years.

Truly, it appears the nature of effective cognition is restricting all of the many responses to stimuli to those that are useful and relevant, and I think the parts of the brain that do that may have 'fallen asleep' in all those instances.

I have migraines and every once in a blue moon I get a very interesting symptom along with it. Fully awake and cognizant I simply lose the ability to read. Even Dick and Jane become nearly insurmountable. It happens even before the pain and aura arrive, I'm fully functional, just suddenly illiterate. Its the damndest thing. It's only for a few minutes but I know what it's like to be a fully grown adult with a relatively high IQ who can't read.

I find migraines like that to be quite fun (at least from the second time, after I spoke to a doctor and learned that there was no permenant damage). In my case I totally lose my short-term memory :D

The weird thing is that after the migraine is over I can actually remember most of what happened during it, so somehow things are still finding their way into my long term memory - though it's mostly just conversations like "hey, I'm having a migraine, I'm going to lie down for a few minutes, because I'm having a migraine, it's affecting my memory, so I'm going to lie down, I think this migraine is affecting my memory" "ok dude, take a break" "woah, how did you know I need a break? I was just thinking that I should take a break, because I'm having a migraine, and it's probably affecting my memory. Did you know I get migraines that affect my memory? I should probably go lie down"...

This reminds me of an article I read a while back [1]. He had years of memories come back that he was never able to remember before the surgery.

[1] http://qz.com/511920/a-tumor-stole-every-memory-i-had-this-i...

What you and the people replying to you are describing sounds a lot like what happens when you try to read during a dream: you just can't. If you look at text while dreaming you recognize the words but they don't make any sense, and if you look away and then back at them they'll all be different. Looking out for this is one of the techniques for recognizing the dream state and getting into a lucid dream.

Perhaps your migraines sometimes trigger a waking dream state, or something functionally equivalent, where you're still seeing the real world instead of one your mind has created, but parts of your brain that are used to interpret your sensory input have switched to dream-mode.

I have something vaguely similar when my migraine-like[1] headaches are at their worst - some words just look wrong, even when I'm 100% certain the world is spelt correctly. It's an odd feeling, made especially weird as when that is happening I find Japanese kana much easier to read, which is normally something I find rather difficult.

[1] Not formally diagnosed as migraines (migraines don't typically cause a single headache that never goes away for years, but just fluctuates in intensity), but my sister has been so I suspect there is a link.

That's interesting! I have a similar but not nearly as severe symptom caused by lack of sleep - the first thing to go for me is spelling.

Weirdly I can still code easily, even keeping in my head high level module interactions and abstractions, to low level implementation details, but I'll occasionally stop, stare at a word, convinced that it's spelt wrong, google it and see its correct, and go back and stare at it - because it just feels wrong.

When I've reached this point, I know it's time to break!

This is interesting. Does it happen before every migraine, or just sometimes? Have you got yourself checked out? Hope everything is fine but that shit would scare me like hell at least the first couple times.

Almost never. Once every two years or so. But it first happened when I was 15 and I thought I was dying! Turns out it runs in my family. When I told my mom she did the whole "there's something you should know..." It was like a superhero backstory, except it sucked instead.

I have the same thing when I have migraines. Apparently it can be a symptom? It's like I can see the words but I can't make sense of what they mean.

This is exactly what happens. The strange thing is I can still do math perfectly, just not read text. I have this little "proof of work" thing I do (usually to figure out how drunk I am!) where I derive Kirchhoff's laws from Faraday's law (a hold-over from my EE degree). I can do this just fine but I can't read beyond sounding out words like a 1st grader and even when I do that, I can't tell you what the sentence meant when I'm finished. Odder still, I can write text and then can't read what I just wrote!

The longest its ever lasted was about 1.5 hours. Usually about the time the freight-train of pain arrives, everything goes back to normal. Of course then I'm in no mood to read anything anyway.

Can you recognize letters? Try adding up the letters-as-numbers in each word.

>I've noticed that when sleep deprived I notice different things and have a more diverse set of emergent thoughts/recall events.

Focus and sleep are clearly related. When my sleep apnea goes untreated, I feel like I suffer from ADHD. In fact, I was certain I had it until I received treatment for SA. Its incredible what poor sleep can do to us. The list of things that go wrong with poor sleep is pretty long.

If you're getting this often, I would get checked for SA.

That's the "sound" of your braincells dying. At least that's what I heard relating to tinitus.

Not sure of that.

A relative of mine had tinnitus which was cured by hearing aids; the theory offered there was that tinnitus was effectively the brain turning the automatic gain control on hearing up too far.

What? Can you elaborate?

Tinnitus may be like phantom limb: when receiving no input from a group of sensors, the system fills in for the missing information. It's like a microphone with compression: when the speaker stops talking, white noise becomes louder. Your neural system does a lot of filling in for missing sensations, most obviously in tests of what happens when bits of spoken words are removed from a recording: people perceive that the bits were in the recording by triangulating from other information that the bits had been said. Similarly, when hearing loss prevents certain frequencies to be signaled by your sound sensors, the system fills in for the lack of signal in that range. There is still research and some doubt about this theory of what is going on: https://www.nidcd.nih.gov/health/tinnitus

You know how every generation conceives of the brain as a machine that's analogous to the latest technology of that generation? Like in the 19th century, the brain was a steam engine, and then in the 20th it was a computer? Well, apparently now the brain is a smartphone.

The difference is that this time around we have massively augmented our data processing capabilities and improved our ability to investigate brain function at multiple scales. We can now formulate falsifiable theories and test them. Science was still pretty immature in the 19th century in that it wasn't always possible to crisply state theories in ways that give rise to testable predictions.

Is the brain solely a computer? I don't know. But it certainly contains information processing aspects that we should study and understand, and as we understand them with these tools, we may find other frameworks to apply at different (perhaps higher) levels.

> We can now formulate falsifiable theories and test them. Science was still pretty immature in the 19th century in that it wasn't always possible to crisply state theories in ways that give rise to testable predictions.

Doesn't that still hold true today with string theory and social sciences?

String theory is more analogous to quantum field theory than it is to the standard model. The standard model is a model of quantum field theory. Likewise, a theory of everything could potentially be a model of string theory. Particular models of string theory are ruled out by the same experiments that rule out particular models of quantum field theory. The ultimate goal is to find a model of some theory that reduces to the standard model in the low energy limit and includes gravity.

Now, it's potentially the case that string theory makes predictions for a given model that can't be practically tested. That's definitely an issue. Another issue is that we haven't yet pared down the possibilities in string theory to a parameter space that can be reasonably searched experimentally.

Social sciences do generate testable predictions, but they often run poor studies that fail to falsify untrue predictions. 538 recently had a good explanation.


AFAIK string theory hasn't made any testable predictions beyond the standard model, but the standard model itself does have testable predictions that have so far all held up. So string theory is just as falsifiable as the standard model.

I don't know about the social sciences. That's painting with a pretty broad brush, IMHO.

True, and I was bit tongue-in-cheek as a social scientist myself. Replicability of many studies is a long-standing issue.

I'd like to extrapolate this further: We won't have true AI until mankind explains the human brain as a "form of AI"

or We won't have true AI until AI explains the human brain as a "form of AI"

We won't have true AI until AI stands on its own. Just as a bootstrapped compiler can compile itself, an AI should be capable to reinvent itself without outside help.

Until then, it's just a child-AI, dependent on human brains to be designed and optimized.

But humans aren't even capable of that, and we are intelligent. It still takes a lifetime of learning and experience to create a functional human being, and many never get it right.

Yes, but until AI is better than humans at everything it will not be perceived as intelligent, because true intelligence will be redefined as one of the things AI still is not capable.

One person can't create himself, but people can "compile" other people into existence and teach them, all without the help of technology.

That's a pretty arbitrary definition.

And how long time do you give it? The human race haven't solved that problem in some 20k+ years. If we build an AI today, would you wait that long to decide if it is "true"?

If humans create an AI, but that AI can't create an AI, then humans are still smarter than that AI.

But someone else being smarter doesn't mean the AI isn't intelligent.

We don't even have true consciousness so why should we have true AI. Thats not how things work IMO.

AI has alway struck me as an absurd nomenclature.

Inteligence is inteligence regardless of origin.

Might as well call us AI Accidental Inteligence.

I kinda like Engineered Inteligence.

You mean like a virus?

Self replication and self preservation are the essence of life and evolution. There is no primary purpose to a biological system other than self preservation and replication.

Dont forget that a computer program exists in the physical world. It is not unlikely for AI that understands self, to also want to preserve self.

I'm not sure how this relates to the article?

I believe the comment is referring to how the newer ARM CPUs in smartphones utilize small battery efficient cores for basic stuff and can ramp up to more power hungry cores for the heavy stuff: https://en.wikipedia.org/wiki/ARM_big.LITTLE

It's offering the smartphone as an analogy for the behavior described in the article.

It is a human illness wanting to explain something to just stratify our rational thinking. In reality whenever there is a paradigm shift, the previous explanation becomes obsolete. So stop trying to explain things and go out an do it and experience for yourself.

I don't understand. Are you suggesting that our drive to reason about things and to construct mental models is an illness?

with all the blood pumping ther is certainly a hydraulic aspect

The brain is a network of routers.

The brain is a series of tubes.

I just the other day got … a thought that was sent by my staff at 10 o'clock in the morning on Friday. I got it yesterday [Tuesday]. Why? Because it got tangled up with all these things going on the Brain commercially ... They want to deliver vast amounts of information over the Brain.

The brain is not just something you dump things on. It's not a big truck. It's a series of tubes. And if you don't understand, those tubes can be filled and if they are filled, when you put your thought in, it gets in line and it's going to be delayed by anyone that puts into that tube enormous amounts of material, enormous amounts of material.

I think we are getting closer to furthering the concept that dreaming isn't just a side-effect of thinking (or simply a defrag/clean-up routine upon the previous day's thoughts or the mind's expectations for tomorrow). It is itself the key to the thinking process, and sits above and commands the other "agents". It helps us deal with uncertainty and lack of rationality inherent in the universe and our incomplete understanding of it.

I don't understand this obsession with the "purpose" of dreaming. As far as I have experienced its just noise on the wire that we sometimes are conscious enough to recognize.

Having spent enough time in extremely sleep-deprived states for projects recently, I've experienced firsthand a lot of funky stuff going on (e.g. falling asleep and waking up with a full page of class notes in front of me, or realizing that I've started 'dreaming' while editing a document, the dream being seeded by the words I was processing.)

>As far as I have experienced its just noise on the wire that we sometimes are conscious enough to recognize.

If it was actual noise it would have absolute minimal statistical chance of having any coherent structure.

Humans appear to be hardcoded to find patterns even when (from a causal and mathematical perspective) they do not exist in the random data.

So "meaningful" output doesn't necessarily mean the input wasn't random.

It's noise in brainspace, which is already structured.

I have lucid dreams almost every night now. I'm talking about entire story arcs with plot twists.

Last nights had Donald Trump, my wife and Jennifer Love Hewitt. There were riots going on and people reading newspapers with moving video ink. There was an old book called "The Fall" that was filled with fascinating ink drawings of ritualistic acts (eg: a man sat cross legged on broken fragments of his own arms). Etc. That kind of stuff. Had a full story.

Noise on the line in Brain-space could still be the case, but if so we can learn something from that noise regarding the way brain structures work.

What I'm saying is if it is just noise, then the noise must be feeding into an entire projective area of creativity that amplifies that noise, not just locally but over time, and in ways that maintain consistency.

Really fascinates me

I'm seconding the story/plot elements.

I've had some dreams like short sci-fi stories. Some with time travel/causality loop tangles as plot. Some with twists. Complex character and places.

I don't think that explanation has much content.

Mangled data with structure are not noise -- at worse they contain some noise (like a dirty dataset).

Consider a c struct describing a circle, with fields for position, radius, and color. Imagine you have a buffer full of random noise, and you cast it to an array of said structure. You then render the array to a screen, you'll get a bunch of randomly placed colorful circles, rather than a bunch of visual noise.

In this model a series of consecutive bytes has a predefined meaning existing outside of them (namely in the source code that defines the struct), so when they are read they are cast as a struct-circle object.

This also only works well for scalar fields, not associations (pointers in the C program's case) and more finely grained items.

Also, where would that (the definition of the "struct") happen with the brain? Is there any literature showing anything similar, that we have memory fields that pre-correspond to specific objects through some mechanism?

Besides, dreams can have very clear structure and narrative, which takes them outside the realm of what can be achieved with noise, unless the "struct" has fields that correspond to things like emotions, faces, narrative changes, etc.

You can also take other analogies. Complex systems usually have several layers of planning or decision making and if you randomize on the top layers, it will still appear to be structured, because the top layers have different degrees of freedom.

To take another example, a self-driving car might make a decision to switch the lane. If you randomize on that level, it will randomly switch, but the movement will still be smooth, because this is done one or more layers below.

That is exactly why we study sleep.

We examine the circles not to learn about the data from which those circles are derived, but rather to learn about which structs and other properties are defined in the system. We might find circle struts, square structs, triangle structs, but never spheres, boxes, or pyramids. Or maybe some people (schizophrenics) have the metaphorical three dimensional structs and others do not.

Actual noise with no coherent structure is enough for the brain to start imagining patterns - can confirm that having tried it myself, with white noise in my eyes and ears for half an hour, I ended up having dream-like hallucinations while awake:


I want to say, pyramidal cell in visual cortex connectivity network decoherence - similar happens with mice after sleep deprivation, and is reminiscent of the phenomenon of HPPD, and with stochastic resonance.

Noise 'on the wire' in a neural network is going to look different from noise on an actual copper wire.

E.g. noise in the brain might be an increased incidence of errant impulses/signals. There's evidence that sleep involves pruning meaningless connections - noise in this case might mean signals propagating farther than they ought.

Noise was meant in an abstract way, not statistically random noise you see in physical circuits.

Rorschach tests seem to have coherent structure, despite being randomly generated blobs.

Only in the superficial way in which we see figures in clouds, tree barks, etc. that we can name. Not in the sense that somebody sees a specific figure in a Rorschach test. But in dreams we have the impression of very specific semblances and "movie-like" experiences.

> in the superficial way in which we see figures in clouds, tree barks, etc. that we can name.


I've seen very specific semblances in coffee spills, knot-holes, etc. The reason they're not movie-like is because they're still and static, and not connected directly to my brain.

Dreams have the advantage of being on the bare wires, whereas Rorschach tests and the like are air-gapped.

I have never experienced this. How sleep-deprived do you have to be before you start dreaming while editing a document?

For me at least, it seems to depend on why I'm sleep deprived and what I'm doing. I spent 100 hours straight through working on my dissertation (only breaks to go fetch food etc.), and felt fine for the entire period - no fatigue, no noticeable cognitive effects, I was alert and coherent - right up until I stopped writing. Then it all just hit me all at once - speaking coherently was a challenge, there was an afterimage if I looked around too quickly, my mind was full of scattered and fragmented half-thoughts (not dreaming, but a similar feeling), and I ended asking a friend to go with me to a building I'd been to hundreds (possibly thousands) of times because I wasn't sure I'd find my way there. To put it simply, I was mentally a mess, and just needed to crash for 18 hours to recover.

On the other hand, I'm currently sleep deprived due to insomnia (it's 1320 on Tuesday, and I've had less than an hour sleep since 0900 on Friday), and the effects have been much more gradual. It's a steady build-up of fatigue and general "brain slowness" for lack of a better word (e.g. I was playing Picross last night, and puzzles that normally would take me less than ten minutes took me more than half an hour), but I'm more or less fine - no visual effects, my thoughts are coherent, and it doesn't feel much different to how I'd be after a long day working (just significantly more severe).

That's amazing to me. I haven't been over 20 hours without sleep for several years, before that I think the max I did was about 35-40 hours which was enough to almost make me hallucinate. I guess I could go for longer, but at that point the only thing I want to do is sleep, any other life goals suddenly become distant second in importance.

It's certainly not something I'd recommend doing (I've had a couple of occasions where I've been up 40ish hours and started to notice the visual disturbances, so it can definitely come on sooner).

The dissertation was partly very poor planning on my part, but I also stumbled on a doctoral thesis that more or less disproved the central tech. behind my dissertation less than a week before the final deadline, which gave me two options:

A) finish up what I'd done, knowing it was fundamentally flawed and that I'd never be able to defend it if called for a viva voce (unlikely, but a definite possibility)

B) rewrite pretty much the entire thing (~30k words) to take a different approach, re-implement the hardware (audio filtering; didn't help that the spec. sheet lied to me and said it had an FPU but actually just emulating floating-point in software which was unusable slow), and come up with a convincing conclusion explaining why a year-long project produced nothing particularly useful.

I went with B, hence the 100 hours of solid work, and it went OK (mark was a marginal pass, but a pass nonetheless), but I'd hate to do it again. It took me 3 or 4 days to start feeling normal and alert again, despite sleeping 18 hours after crashing once it was finished.

The current insomnia is really weird though. I saw my GP this morning, took a fairly hefty dose of sleeping aids (15mg diazepam on top of the 20mg tamazepam and 75mg promethazine I was already prescribed[1]), yet somehow feel more awake than before taking those, despite the fact I'm only about three hours away from having had less than than an hour sleep total in the last six days. Honestly, I just feel bored more than anything at this point (hence this somewhat unnecessarily long post :p) - I don't have the physical energy to do a great deal, but mentally I feel more or less normal (though I suspect when I finally do get some sleep, it'll catch up with me and I'll have a rouge couple of days).

[1] FWIW, I've also got Zopiclone, but I saving that as a dug of last resort. The benzodiazepines cause me no side effects, but last time I used the Zopiclone it was rather unpleasant - it caused mild hallucinations (my room looked like it was bathed in a lemon-yellow light, despite it being the dead of night and no lights were on), and it had a strong amnesiac effect (my memory of the night pretty much goes "I'm in bed, why is everything yellow?", "I'm in the bathroom, how did I get here?", "I'm back in bed, but it's three hours later", and I have literally no idea what happened between each thought. It's just like a jump cut in a movie - there isn't even a hazy memory of I did something; it literally just goes from being in one room to being in other, despite there obviously being a chuck of time missing. It's pretty disconcerting).

It seems to happen while sitting down doing repetitive editing after ~40 hours wakefulness after a short night of sleep.

I'm also looking into evidence that a blood sugar deficit may have been involved. Reportedly some of the techniques used to maintain alertness can result in increased glucose metabolism coupled with the unfortunate side effect of having zero appetite.

This is a question where the "wait and see" approach works.

Perception operates like guided hallucination.

This reminds me of the Mental Modules popping in and out of existence from the Coursera course "Buddhism and Modern Psychology", in week 4: https://www.coursera.org/learn/science-of-meditation#syllabu....

I was recently reading my daughter a story after being woken up too early and 'half asleep'. I kept noticing that I had invented a line of text or two, and knew that I had done this because part of my brain wasn't awake enough to focus on or perhaps read the words on the page. So one part of my brain was watching another invent plot or dialogue, while aware that another was 'asleep'. Other times I became aware that the story-supplementing part was not functioning and was able to jolt myself back to a state where I could read and recite the current line.

Here's some amazing relevant research with so many consequences that I keep coming back to:

Partial sleep in the context of augmentation of brain function


Inability to solve complex problems or errors in decision making is often attributed to poor brain processing, and raises the issue of brain augmentation. Investigation of neuronal activity in the cerebral cortex in the sleep-wake cycle offers insights into the mechanisms underlying the reduction in mental abilities for complex problem solving. Some cortical areas may transit into a sleep state while an organism is still awake. Such local sleep would reduce behavioral ability in the tasks for which the sleeping areas are crucial. The studies of this phenomenon have indicated that local sleep develops in high order cortical areas. This is why complex problem solving is mostly affected by local sleep, and prevention of local sleep might be a potential way of augmentation of brain function. For this approach to brain augmentation not to entail negative consequences for the organism, it is necessary to understand the functional role of sleep. Our studies have given an unexpected answer to this question. It was shown that cortical areas that process signals from extero- and proprioreceptors during wakefulness, switch to the processing of interoceptive information during sleep. It became clear that during sleep all “computational power” of the brain is directed to the restoration of the vital functions of internal organs. These results explain the logic behind the initiation of total and local sleep. Indeed, a mismatch between the current parameters of any visceral system and the genetically determined normal range would provide the feeling of tiredness, or sleep pressure. If an environmental situation allows falling asleep, the organism would transit to a normal total sleep in all cortical areas. However, if it is impossible to go to sleep immediately, partial sleep may develop in some cortical areas in the still behaviorally awake organism. This local sleep may reduce both the “intellectual power” and the restorative function of sleep for visceral organs.


Fascinating stuff, imo. I've definitely felt the difference between partial sleep restedness and total sleep restedness, and the idea of sleep pressure seems to line up with my experience and with my idea of explorable, falsifiable models of phenomena. The idea that we dream while processing internal organ information is beautiful to me on multiple levels- maybe memory consolidation has a whole lot to do with pre-existing organ health! Maybe instead of the idea that we dream with only our heads, we dream with all of our organs- just imagine how differently Jung would interpret Pauli's dreams if he knew about this!

I wonder if this also explains some of the benefits of meditation. While focusing on one part of our higher functions, more areas of the brain cam slip into this low power state, without fully entering sleep. Which is why people feel refreshed and relaxed after a session.

Seems like the article misses an interesting analog in power gating:


I seem to recall similar research in rats or mice that found that "sleep" states occurred randomly when awake, and slowly synchronized into moving waves of "sleep" when asleep. If I am indeed remembering correctly, then it would seem more likely that this localized sleep mechanism is shared by all mammals, including humans. I'm sure many of us have experienced cycles of awareness while supposedly fully awake as well.

I wonder if it extends to reptiles, fish, etc.

Evidence that birds sleep in mid-flight


Thank you for this interesting paper!

So our brains have Speedstep and are able to shut down cores to save power? :)

Looks like my brain does full shutdown whenever there's an interview

Ha! Came here to say something similar.

For me it's in meetings or when someone's telling me a vital piece of information I must remember.

Feel you!

Seems to me that the brain is conservative on energy consumption here. The same I apply to my hobby electronic circuitry design, i.e. turn off sensory areas when not in use and back on as needed.

I would be interested to see how this relates to mental illnesses such as ADHD, schizophrnias catatonic states, and dissociative identity disorder. But maybe that will be another decade out.

This reminds me of dropout regularization - the key idea there is to randomly turn off units during training.

You're on to something!

If our brains would work fully all the time, our head would boil and explode. I don't see how this is so new.

I am glad that functioning of the brain is understood much better. Also, this could be utilized if we could find ways to increase focus, maybe even artificially, well artificially until brain gets trained for full flow state for example.

>If our brains would work fully all the time, our head would boil and explode. I don't see how this is so new.

Small price to pay if it gives you an edge in the valley!

>If our brains would work fully all the time, our head would boil and explode. I don't see how this is so new.

Citation needed.

I was referring the energy usage of the brain. It is never all used at the same time, just a small portion, but that doesn't mean other parts are not used, just they are all turned on and off.


See Carlos Casteneda and "assembling the self."

Dolphins put half their brains to sleep at a time.


Please don't introduce politics to a thread unnecessarily, especially this week:


I feel like this isnt political?

It's off-topic with respect to the submission. It's also a reference to the anti-vaxxer debate, which is definitely political. It's also a smear aimed at half of the population, so it's uncivil as well.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact