(Original post on the polyphasic sleep experiment here.)

Welp, this got a little messy. The main culprit was Burning Man, though there were some other complications with data collection as well. Here are the basics of what went down.

Fourteen people participated in the main experiment. Most of them were from Leverage. There were a few stragglers from a distance, but communication with them was poor. 

We did some cognitive batteries beforehand, mostly through Quantified Mind. A few people had extensive baseline data, partially because many had been using Zeos for months, and partly because a few stuck to the two-week daily survey. Leverage members (not me) are processing the data, and they'll probably have more detailed info for us in three months(ish).

With respect to the adaptation itself, we basically followed the plan outlined in my last post. Day one no sleep, then Uberman-12, then cut back to Uberman-6, then Everyman-3.

Most people ended up switching very quickly to Uberman-6 (within the first two or three days), and most switched to Everyman-3 after about five to seven days on Uberman-6. Three people tried to hold the Uberman schedule indefinitely: One person continued Uberman-6 for two full weeks, and two held out for twenty-one days. Afterwards, all three transitioned to Everyman-3. 

During the originally planned one-month period, five people dropped out. Nine were on some form of polyphasic for the whole month. One returned to monophasic at the end of the official experiment with only partial adaptation achieved. 

Then Burning Man disrupted everybody's sleep schedule. Afterward, one person continued experimenting with less common variations of the Everyman schedule. Three went back to Everyman-3. One switched to Everyman-2. Two people have flexible schedules that include two hours less sleep per day. One person's schedule was disrupted by travel for a while after Burning Man, and they're now re-adapting.

Now that all is said and done, eight of the original fourteen are polyphasic.

I'll hold off on concluding very much from this until I see the results of the cognitive battery and such, plus the number who are still polyphasic after three months. In the mean time, I'll just stick with this: Some people are capable of going polyphasic and staying that way (probably?). Sleep is complicated and confusing. I don't know how it works. I don't think anyone else really does either. More research is desperately needed.

I know three months is a long way away. I'm feeling impatient too. But details will arrive! In the mean time, here's a video of what zombie-Brienne is like during the really difficult stretches, and here is how she entertained herself when she could manage to do things besides pace. (I was one of the few who bailed out early :-p)

New Comment
69 comments, sorted by Click to highlight new comments since:

Note that since the entire human species (7 billion people) spends around 1/3 of its days asleep or trying to sleep, and sleep is desperately important to human productivity, given the total value of this information, we should expect that it will be explored primarily in an ad-hoc, unfunded way by people in the quantitative self community, rationalist community, Leverage, etcetera. #civilizational-inadequacy

So, I discussed my polyphasic sleep plans extensively with my parents before, during, and after them. My father, trained as a fighter pilot, knows quite a bit about sleep from the military's training on it, and was intensely skeptical of uberman, and correctly predicted that it would not work for me. (My family was atypically well-rested when I was growing up, which is probably due to some combination of that training and our personalities.)

It seems to me that the military in particular has done a lot of well-funded research into sleep, and is willing to try fairly exotic things, and is willing to sleep-deprive people during training. It's possible that they haven't tried polyphasic sleep or that they're generalizing incorrectly from studies they've already done in dismissing polyphasic sleep, so that the military isn't doing this already is not conclusive that polyphasic isn't widely useful enough to adopt. But it does push some of the probability to the hypothesis that polyphasic does not work long-term for many people without serious drawbacks. In that case, only ad-hoc, unfunded research on it matches what you would expect for a project to, say, invent an engine that runs on water.

Matt Fallshaw's 2012 presentation to minicamp on polyphasic said that it trades flexibility for time - you sleep less, but you must follow a rigid schedule. It's possible that the military need too much flexibility to make it practical.

There are lots of situations where "it's been 4 hours, I need to take a nap" would be a critical handicap, but there are others where it wouldn't. Having only two shifts on a submarine (2 on 2 off) rather than three (6 on 12 off) seems like it would be useful.

I would have thought a submarine could sometimes hit a crisis that required all hands. But a stronger example might be drone pilots.

I would have thought a submarine could sometimes hit a crisis that required all hands. But a stronger example might be drone pilots.

Possible, but that's when you pop your caffeine gum (or your speed) and skip a nap. I know that submarines are space-limited, and so dropping the crew requirements might be helpful- I'm not sure what the limitations are for drone pilots. (The flexibility that means they don't have crises might also mean there's no real gain from having pilots that can be awake for more of the day.)

I'd have thought that the military would at least shy away from Uberman polyphasic because getting enough sleep on that schedule is very fragile-- if there's an emergency, there's much less reserves than for people who sleep in larger chunks.

I have indeed heard many rumors of the military running fascinating sleep experiments. But I've not had much luck getting access to their results, let alone the details of methodology and such. I admit, however, that I've spent less time and effort looking for them than I should have. If you happen to have illuminating links, please share them!

I'd also like to note that anyone on a successful polyphasic schedule has a pretty huge comparative advantage, so if I were the leader of a polyphasic army, I might not want my opponents knowing anything more than "I hear there is evil there that does not sleep," if even that.

this advantage is directly proportional to how many people you have on this sleep schedule. The fewer people who do it the lower the payoff, but the more people who do it the less chance you have of keeping it secret. If the army invents an advanced airplane, it can be kept secret and used rarely, but if the army invents a new kind of carbine, it's much more valuable if ALL the soldiers have it. Polyphasic seems a lot more like the carbine than like the fighter jet.


The US military did a lot of research with modafinil and amphetamines to keep pilots awake for long-haul flights.

Is that research public? If not, maybe someone could make an effective Freedom of Information request for the information?

I do not think FoI requests apply for classified military studies

Polyphasic is useless for the military IMHO.

Its not flexible. Pilots are less useful if they must be sleeping on multiple rigid times.

Its risky. Because if you lose a nap you are fucked. And this is exactly the type of screw-up the military likes least. You train people to stay functional under duress. Not to be germless lab pigs that fall like a feather missing a nap.

Thanks for the update. I eagerly await the results of the cognitive tests.


Small correction: I actually found Everyman 3 to be a very doable schedule at Burning Man. It's desirable to stay up really late since lots of neat stuff happens at night, and it's desirable to not need to sleep past 10am since it gets very hot. So a 3 hour core from 6-9am plus a few opportunistic naps in the shade is an excellent solution. Both Cathleen (who runs operations at Leverage) and I were on duty supporting Paradigm, the effective altruist camp, most of the week, and I think it's fair to say the quality of experience the camp achieved was due in no small part to the long hours we were able to put in.

The hard part has actually been sticking to the diet after the event, due to being quite sick for a while.

Looking forward to publishing data as soon as we have the time. As a preview, since it doesn't take any time, here's a plot of my sleep since starting the experiment. The no data part is Burning Man, which was a similar distribution to the period right before it.

Thanks for doing this; I, too, am really looking forward to the cognitive-test data.

I'm curious/concerned about the illness you mentioned. A lot of the argument against polyphasic sleep is that there may be poorly-documented negative health effects, and immunology is an area where downsides are likely to be found. Some infectious illnesses happen randomly, but it could be that polyphasic sleep is a risk factor and it would be bad to miss that, so investigating (or at least thoroughly reporting) things like that seems important.

What's "the diet"?

I just meant the schedule. I've taken to calling it a diet since I'm avoiding it sortof the way people do with food.

What are you using to measure your REM/light/deep sleep?

The Zeo. It isn't designed to handle short naps, so you have to manually copy down the data right after a nap into a spreadsheet or notebook or something.

I wondered why Burning Man should disrupt polyphasic sleep except possibly by intoxication. But what I get from your earlier post and www.polyphasicsociety.com is that polyphasic has the following disadvantages:

  • requires strict sleeping patterns
  • requires discipline to keep the pattern
  • limited task and schedule flexibility

Obviously you need a job and a family and friends where this is no problem. I have thought about trying DC1 which is the only option that might work with four children.

I wonder what the consequences of falling out of polyphasic may be. From my own sleep deprivation 'experiments' I'd guess:

  • Involuntarily sleeping much longer than planned out of order, thus completely wrecking you schedule.
  • Being tired and drowsy for a time thus being unusually unproductive and potentially unreliable.

From your graph I'd guess that falling out of polyphasic (accidentally or not) takes a week to recover. If such accidents happen too often (and once a month may be enough) all the disturbances this causes (missed deadlines, bad quality) may quickly eat up all the nominal efficiency gains from more awake hours.

This all assumes that the saved sleep has no other positive benefits. There might be:

  • physical regeneration
  • subconscious learning (there are some posts on LW about the habit to reflect the lessons of the day before sleep)

With these considerations in mind I decided to stay with my sleep rhythm (0:00 to 6:30). I don't want to risk long time health effects ('the candle that burns brighter burns half as long').

The main thing I took out of this are the recommendations about night lighting: http://www.polyphasicsociety.com/polyphasic-sleep/adaptation/night-lighting/

The graph is a little misleading on how long it takes to recover, I'd say, since the falling out part was due to being quite sick. There is also an important additional note, which is that on E3 so far, I'm getting way less REM than I used to get with normal sleep (see plot of my baseline data below). As I eventually concluded with Uberman, it really seems like the standard E3 schedule has no possible way of giving me the amount of REM that I used to get (over two hours per night on nights when I would sleep 8-8.5 hours and feel well rested), so I'm going to test a diet that will take 5.5 hours rather than 4, but will have a legitimate chance of matching my baseline numbers for REM and SWS (slow wave sleep). This would beat my normal sleep time goal of 8 hours by 2.5 hours a day, but would include lots of naps, so it's yet to be seen if it is worth it. It will come down to whether I turn out to be able to take advantage of several 30-60 minute waking periods that are part of the rotation I'm going to try.


subconscious learning (there are some posts on LW about the habit to reflect the lessons of the day before sleep) I just stumbled over this study:

Memory reactivation and consolidation during sleep, 2004 by Ken A. Paller1 and Joel L. Voss


which argues:

We propose that declarative memories change both during waking and during sleep, and that such change contributes to enhancing binding of the distinct representational components of some memories, and thus to a gradual process of cross-cortical consolidation. As a result of this special form of consolidation, declarative memories can become more cohesive and also more thoroughly integrated with other stored information. Further benefits of this memory reprocessing can include developing complex networks of interrelated memories, aligning memories with long-term strategies and goals, and generating insights based on novel combinations of memory fragments.

No analysis/release of the data collected?

Not yet. Watch http://www.leverageresearch.org/ . I'll post about it as soon as I know more.

Maybe you could just release the raw Zeo data?

Still nothing?

Is anyone doing some sort of polyphasic schedule while going to 9-to-5 day job where napping on the premises isn't practical?

This basically means that you have one nine hours or so block which you need to be awake for, but outside that anything goes. At least the dual-core schedule looks like it could accommodate this, but the article doesn't seem to really describe how they come up with the schedules or how they can be customized.


Not polyphasic but

I looked into this at some point, and the reports I found of people who had actually tried this seemed to mostly say that it was horrible if you actually tried to keep up with it. It's probably a bad idea to go mess with the 24-hour cycle unless if it's not broken to begin with.

One possibility I'm thinking of is to have the two core sleeps around the workday, for example 4 to 8:30 sleep, then 9 to 17 work, then 17:30 to 19:30 sleep and a 20 minute nap sometimes around 23.

I'd just like to mention a shout-out for the sleep schedule Dual Core 1, which I think is even more easily sustained than Everyman.

I can add a +0.5 to this; one of my friends adapted to DC with very little fanfare or sleep deprivation.

This is good to hear, because 6hrs is definitely better than 8hrs. However, evidence of sustained cognitive performance quality is even more important to me on this sort of schedule, because we already know that many people on swing shifts (and in, say, grad school) seldom sleep more than six hours a day, and often in separate chunks, but we also know that both the short-term and long-term effects of this sort of chronic partial sleep deprivation are detrimental. If someone's only sleeping two-hours a day and is still able to talk, it seems much more likely that something strange and completely not understood is going on, and chronic partial sleep deprivation is not what's happening.

DC1 is different from DC (aka Segmented) is only a little shy of 5 hours (3hr core, 1.5hr second core, 20 min nap), not 6.

If someone's only sleeping two-hours a day and is still able to talk, it seems much more likely that something strange and completely not understood is going on, and chronic partial sleep deprivation is not what's happening.

I think there's a fallacy here. It's possible Uberman might succeed enough to allow some cognition (and more than we might otherwise expect), but not optimal cognition.

Ah, my mistake about sleep length in DC1.

Sorry, I meant for "still able to talk" to stand in for the large group of things constituting "appear basically normal", which the standard literature suggests they should not.

I agree with you and believe in polyphasic sleep, but my point still stands -- one can "appear basically normal" without actually having optimal cognition.

Puredoxyk reports being able to do work while laying down, etc, which is basically impossible if you're seriously sleep deprived.

I'd appreciate a sentence or two about the outcome of this.

Ah, so would I! I think I never actually got to see more info on the outcome. I don't know whether or not anything was actually compiled. Once it was in Leverage's hands, I guess I lost track of it somehow, but I don't remember why.

I didn't expect a response! It's probably too late now, which is a shame—there's not that much information on polyphasic sleep. But thanks for answering anyway!

Do we have any results for this yet?


So glad to see this. I'm trying to get into an Everyperson 2-Nap schedule as mentioned here...I'm 5-6 days in, but unfortunately I slept 3 hours through my alarm (I think I turned it off in my sleep) during core yesterday morning. Any reasons I should not use that schedule in particular?

It's a really funny coincidence to me that you just posted this given that I started the schedule last week and did a search on Less Wrong and found your post. My influences to try polyphasic sleep were outside of Less Wrong.

Before this experiment, I would have strongly advised trying Everyman 3 rather than 2 first, because the information out there suggests successful adaptation to E3, especially via preliminary adaptation to Uberman, is more reliable. But my confidence in the quality of existing information about these things has gone way down, and I'm now very interested in finding out what will happen if you go straight for E2. I want lots of people trying lots of things. PM updates whenever you feel like it on what you're doing and how it's going would be awesome. That goes for everyone else reading this too.


Will do.


Will do.

[This comment is no longer endorsed by its author]Reply

What happened to the cognitive test results from this? It seems like they vanished into the æther.


It's been more than three months. Any results yet?

Has anyone tried orexin-RA-1 for inducing REM?

Not to my knowledge. Have you? Do you know anyone who has? What do you know about it?

All I know is from a press release from 7 years ago. Googling it produces rumors that it is behind schedule but still going forward.

Added: it seems to have renamed almorexant and failed a phase 3 trial. But there are other orexin antagonists still being pursued.

Where would one even get an orexin?

dunno. Orexin-RA-1 isn't a type of orexin, but a drug to block it.

Yes, that's what I meant: where does one even get useful quantities of these drugs which, AFAIK, have not gotten further than the early clinical trials? They're not really something you can just hop onto Silk Road and buy.

More than one orexin-ra has done phase 3 trials. I think that the key to creating a black market is communicating demand.


Given that results from this clearly aren't forthcoming, can we safely assume that the experiment was a failure?

Can you at least post the raw data for the cognitive tests so other people can analyse it, or was that data not properly kept?

I would propose that before you post the results of the experiments it might be worth to decide on the format in which you publish the results and create go to prediction book to create sample predictions.