However, I need to throw a flag on the field for isolated demands of rigor / motivated reasoning here—I think you are demanding a lot from sleep science to prove their hypotheses about needing >7hrs of sleep but then heavily relying on an unproven analogy to eating (why should we think sleeping and eating are similar?), the sleep patterns of a few hunter-gatherers (why should we think what hunter-gatherers did was the healthiest?), the sailing coach guy (this was the most compelling IMO but shouldn’t be taken as conclusive), and a random person with brain surgery (that wasn’t even an RCT). If someone had the same scattered evidence in favor of sleep, there’s no way you’d accept it.
For what it’s worth (and it should be worth roughly the same as this blog post), my personal anecdotes:
1.) Perhaps too convenient and my data quality is not great and non-random, but I found analyzing a year of my time-tracking data showed that sleeping exactly 8hrs (not more not less) maximized my total hours worked (an imperfect but still useful metric of output).
2.) Multiple semi-sustained attempts of mine to regularly sleep <8hrs (including several attempts at polyphasic sleep) did not improve productivity.
3.) Sleeping <6hrs definitely gives me a feeling of “ugh I can’t do this because I’m too tired”, noticeable “brain fog”, and I also have noticeably less willpower. (Though I’ve not tried to long term “adapt”.)
4.) I would agree that oversleeping (>8hrs of sleep) harms my productivity though.
(Note: some of what I comment here is repeating the opinions of other people I talk to, but these people remain uncredited.)
As a random anecdote, I followed something like the reasoning in the post for 3 years between 2013 and 2016, and slept 7.5 hours a day, followed by a half-hour nap in the midday. Without any alarms and with blackout curtains I sleep around 9 hours a day.
I mostly didn’t feel very sleep deprived from sleeping the 7.5 + 0.5 hours, but after I switched towards a 9-hour sleep schedule, I noticed a very large change in my emotional variability, and in-particular noticed that I was feeling substantially less anxious, and substantially less depressed, and also substantially less hypomanic at different times. I still notice that I feel much higher emotional variability if I don’t sleep full 9 hours.
I do think the change to my sleep schedule also came with a lot of other changes in my life, so this is far from even a single clear anecdote, but it did update me that at least for myself, I am quite hesitant to sleep less than 9 hours.
>I do think the change to my sleep schedule also came with a lot of other changes in my life, so this is far from even a single clear anecdote, but it did update me that at least for myself, I am quite hesitant to sleep less than 9 hours.
Do you mean that the change in sleep pattern came with the change in something like the job you had or the country where you lived?
However, I need to throw a flag on the field for isolated demands of rigor / motivated reasoning here—I think you are demanding a lot from sleep science to prove their hypotheses about needing >7hrs of sleep
Yes, I’m demanding a lot from sleep science but you are missing one of the key points: the evidence that sleep scientists have any idea about the amount of sleep we need is incredibly weak. They literally recommend at least 7 hours of sleep (1) without any long-term causal evidence about the effects of sleep on health and with (2) correlational evidence showing that people sleeping 6-7 hours have the lowest mortality and that people sleeping 4 hours have the same mortality as those sleeping 8 hours.
You continue:
then heavily relying on an unproven analogy to eating (why should we think sleeping and eating are similar?), the sleep patterns of a few hunter-gatherers (why should we think what hunter-gatherers did was the healthiest?), the sailing coach guy (this was the most compelling IMO but shouldn’t be taken as conclusive), and a random person with brain surgery (that wasn’t even an RCT). If someone had the same scattered evidence in favor of sleep, there’s no way you’d accept it.
The key point of the analogies I make is to show that the arguments people commonly make to argue for a particular way of sleeping (as much as you want) or for harms of acute sleep deprivation (you feel bad) prove too much, since they could also “prove” that eating as much as you want is good or exercising is bad.
When I use the analogies to go further and to argue that feeling sleep is normal, I’m being very transparent in the essay about the strength of my arguments and they can evaluate whether they are persuaded by a particular analogy or not. I very much agree with you that an analogy is not a strong argument in and of itself.
More generally, I’m always amused when people start listing the evidence I’m relying on and pick the pieces that sound weakest while excluding those that sound strongest. For example, in your summary of what I rely on, you decided not to include
the data I showed above about the association between sleep and mortality
the study about a a single-point mutation can decrease the amount of required sleep by 2 hours, with no negative side-effects.
the studies with thousands of people taking antidepressants for weeks and years who have dramatically lower amount of REM sleep and yet seem to have no issues with cognition.
I believe that your quote is out of context and misrepresents my point. Here’s what I wrote:
At this point, I’m pretty sure that the entire “not sleeping ‘enough’ makes you stupid” is a 100% psyop. It makes you somewhat more sleepy, yes. More stupid, no.
I also specifically wrote about decreasing sleep by 1-2 hours in the long-term not having negative health effects, not that no kind of sleep deprivation has any negative effects on health on cognition. This is patiently false.
1.) Perhaps too convenient and my data quality is not great and non-random, but I found analyzing a year of my time-tracking data showed that sleeping exactly 8hrs (not more not less) maximized my total hours worked (an imperfect but still useful metric of output).
I find it notable that, having criticized my arguments for lack of rigor, your first argument is a conflation of correlation and causation. It’s simply inappropriate to say that “sleeping exactly 8hrs (not more not less) maximized my total hours worked”, given that you only have access to an association between sleep and number of hours worked and there’s an infinite number of confounders affecting both your length of sleep and your productivity.
I don’t think that negative votes on this comment reflects its value, if people who downvoted this comment could explain why I think that would be helpful for discourse.
If I had to guess why this comment has been heavily downvoted I think it would be because the tone of the comment could seem a bit adversarial and possibly because the original commenter is well known within the EA community and some people don’t like attacks on their group. I hope that I am wrong and people are downvoting this on more reasonable grounds, but as I said before it would be great that could be made clear.
I want to apologize for both the tone and relative lack of informativeness in my responses—it seems to me that our thinking here is extremely difficult to reconcile in the low-bandwidth medium of writing.
This comment should be upvoted for its rigor and contribution to the discussion. But I am a bit disappointed that this comment is so highly upvoted and that Alexey’s response fails to communicate the higher level crux. [CoI: Alexey and I are best friends; I have benefited from his takes on sleep.]
The high-level objection I want to raise is something like “ugh, it’s not how you should think about mad science.” Ozzie Gooen described the difference between disagreeables and assessors quite well. This post is of highly-disagreeable-noticing-a-conspiracy-against-humanity-mad-science type and the comment is careful-measured-assessor type. Peter’s comment and even Alexey’s response were operating under “assessing,” which is fair but misses quite a bit as the essay is titled “Theses on Sleep” and not “A Systematic Review of Sleep.”
I can’t imagine the mindset behind this comment producing the core ideas of the post. Like, it’s very hard for me to imagine someone who is not overly dismissive of sleep science (e.g., thinking that it is 100% psyops) bearing through a harsh sleep-deprivation self-experiment and seriously considering that modern sleep is a superstimulus, contributes to depression, is unnecessary. It’s correct to point out that not every single piece of evidence about sleep has been fabricated and used in psyops but I think this ~passion is a cost of taking mad ideas seriously enough to engage with them.
While the above comment is epistemically virtuous on the level of evaluating the strength of evidence and noticing the priors, I think it misses the larger picture[1]: discoveries are often made in bizarre circumstances and look like epicycle upon epicycle [see: SMTM on scurvy (one, two) and The Copernican Revolution from the Inside].
We surely want to reward hypothesis generation. Ideally, great hypotheses would be generated by people who hedge and bow appropriately, but in practice, it takes bull-headedness.
Just as a prior, I would think it’s more likely for motivated reasoning to generate the belief “it is optimal from a health perspective to spend more time doing something that makes me feel better while awake, and that doesn’t require any productivity during that extra time,” than “it is not optimal to spend more time on that, and if anything it is probably optimal to spend less time on that so you can do more effortfully productive things.”
I agree with that prior as stated but I think it’s reasonable to expect that other sources of motivated reasoning might dominate here, in both directions.
I think there’s a lot that’s intriguing here. I also really enjoyed the author’s prior takedown of “Why We Sleep”.
However, I need to throw a flag on the field for isolated demands of rigor / motivated reasoning here—I think you are demanding a lot from sleep science to prove their hypotheses about needing >7hrs of sleep but then heavily relying on an unproven analogy to eating (why should we think sleeping and eating are similar?), the sleep patterns of a few hunter-gatherers (why should we think what hunter-gatherers did was the healthiest?), the sailing coach guy (this was the most compelling IMO but shouldn’t be taken as conclusive), and a random person with brain surgery (that wasn’t even an RCT). If someone had the same scattered evidence in favor of sleep, there’s no way you’d accept it.
Maybe not sleeping doesn’t affect writing essays, but in the medical field at least there seems to at least be an increased risk of medical error for physicians who are sleep deprived. “I’m pretty sure this is 100% psyop” goes too far.
For what it’s worth (and it should be worth roughly the same as this blog post), my personal anecdotes:
1.) Perhaps too convenient and my data quality is not great and non-random, but I found analyzing a year of my time-tracking data showed that sleeping exactly 8hrs (not more not less) maximized my total hours worked (an imperfect but still useful metric of output).
2.) Multiple semi-sustained attempts of mine to regularly sleep <8hrs (including several attempts at polyphasic sleep) did not improve productivity.
3.) Sleeping <6hrs definitely gives me a feeling of “ugh I can’t do this because I’m too tired”, noticeable “brain fog”, and I also have noticeably less willpower. (Though I’ve not tried to long term “adapt”.)
4.) I would agree that oversleeping (>8hrs of sleep) harms my productivity though.
(Note: some of what I comment here is repeating the opinions of other people I talk to, but these people remain uncredited.)
As a random anecdote, I followed something like the reasoning in the post for 3 years between 2013 and 2016, and slept 7.5 hours a day, followed by a half-hour nap in the midday. Without any alarms and with blackout curtains I sleep around 9 hours a day.
I mostly didn’t feel very sleep deprived from sleeping the 7.5 + 0.5 hours, but after I switched towards a 9-hour sleep schedule, I noticed a very large change in my emotional variability, and in-particular noticed that I was feeling substantially less anxious, and substantially less depressed, and also substantially less hypomanic at different times. I still notice that I feel much higher emotional variability if I don’t sleep full 9 hours.
I do think the change to my sleep schedule also came with a lot of other changes in my life, so this is far from even a single clear anecdote, but it did update me that at least for myself, I am quite hesitant to sleep less than 9 hours.
>I do think the change to my sleep schedule also came with a lot of other changes in my life, so this is far from even a single clear anecdote, but it did update me that at least for myself, I am quite hesitant to sleep less than 9 hours.
Do you mean that the change in sleep pattern came with the change in something like the job you had or the country where you lived?
Yeah, a bunch of job and context changes, and it happened something like a year after I had moved to a new country.
Yes, I’m demanding a lot from sleep science but you are missing one of the key points: the evidence that sleep scientists have any idea about the amount of sleep we need is incredibly weak. They literally recommend at least 7 hours of sleep (1) without any long-term causal evidence about the effects of sleep on health and with (2) correlational evidence showing that people sleeping 6-7 hours have the lowest mortality and that people sleeping 4 hours have the same mortality as those sleeping 8 hours.
You continue:
The key point of the analogies I make is to show that the arguments people commonly make to argue for a particular way of sleeping (as much as you want) or for harms of acute sleep deprivation (you feel bad) prove too much, since they could also “prove” that eating as much as you want is good or exercising is bad.
When I use the analogies to go further and to argue that feeling sleep is normal, I’m being very transparent in the essay about the strength of my arguments and they can evaluate whether they are persuaded by a particular analogy or not. I very much agree with you that an analogy is not a strong argument in and of itself.
More generally, I’m always amused when people start listing the evidence I’m relying on and pick the pieces that sound weakest while excluding those that sound strongest. For example, in your summary of what I rely on, you decided not to include
the data I showed above about the association between sleep and mortality
the study about a a single-point mutation can decrease the amount of required sleep by 2 hours, with no negative side-effects.
the studies with thousands of people taking antidepressants for weeks and years who have dramatically lower amount of REM sleep and yet seem to have no issues with cognition.
I believe that your quote is out of context and misrepresents my point. Here’s what I wrote:
I have no doubts that large, persistent sleep deprivation akin to that often faced by doctors has makes them more sleepy and makes them make more errors. In fact, this follows directly from the arguments I made in the section Occasional acute sleep deprivation is good for health and promotes more efficient sleep.
I also specifically wrote about decreasing sleep by 1-2 hours in the long-term not having negative health effects, not that no kind of sleep deprivation has any negative effects on health on cognition. This is patiently false.
I find it notable that, having criticized my arguments for lack of rigor, your first argument is a conflation of correlation and causation. It’s simply inappropriate to say that “sleeping exactly 8hrs (not more not less) maximized my total hours worked”, given that you only have access to an association between sleep and number of hours worked and there’s an infinite number of confounders affecting both your length of sleep and your productivity.
I don’t think that negative votes on this comment reflects its value, if people who downvoted this comment could explain why I think that would be helpful for discourse.
If I had to guess why this comment has been heavily downvoted I think it would be because the tone of the comment could seem a bit adversarial and possibly because the original commenter is well known within the EA community and some people don’t like attacks on their group. I hope that I am wrong and people are downvoting this on more reasonable grounds, but as I said before it would be great that could be made clear.
I’ve upvoted the response and am also confused why it is getting downvoted.
I want to apologize for both the tone and relative lack of informativeness in my responses—it seems to me that our thinking here is extremely difficult to reconcile in the low-bandwidth medium of writing.
This comment should be upvoted for its rigor and contribution to the discussion. But I am a bit disappointed that this comment is so highly upvoted and that Alexey’s response fails to communicate the higher level crux. [CoI: Alexey and I are best friends; I have benefited from his takes on sleep.]
The high-level objection I want to raise is something like “ugh, it’s not how you should think about mad science.” Ozzie Gooen described the difference between disagreeables and assessors quite well. This post is of highly-disagreeable-noticing-a-conspiracy-against-humanity-mad-science type and the comment is careful-measured-assessor type. Peter’s comment and even Alexey’s response were operating under “assessing,” which is fair but misses quite a bit as the essay is titled “Theses on Sleep” and not “A Systematic Review of Sleep.”
I can’t imagine the mindset behind this comment producing the core ideas of the post. Like, it’s very hard for me to imagine someone who is not overly dismissive of sleep science (e.g., thinking that it is 100% psyops) bearing through a harsh sleep-deprivation self-experiment and seriously considering that modern sleep is a superstimulus, contributes to depression, is unnecessary. It’s correct to point out that not every single piece of evidence about sleep has been fabricated and used in psyops but I think this ~passion is a cost of taking mad ideas seriously enough to engage with them.
While the above comment is epistemically virtuous on the level of evaluating the strength of evidence and noticing the priors, I think it misses the larger picture[1]: discoveries are often made in bizarre circumstances and look like epicycle upon epicycle [see: SMTM on scurvy (one, two) and The Copernican Revolution from the Inside].
We surely want to reward hypothesis generation. Ideally, great hypotheses would be generated by people who hedge and bow appropriately, but in practice, it takes bull-headedness.
Another big picture miss is that evidence has been strongly filtered by authority-prestige forces [see: post-structuralism and hard programme in sociology of science].
“Hypotheses on Sleep” would have been more accurate but would have undersold his Martin Luther energy.
(I am wondering if this is sort of what outsiders mean when they say “EA is too systematic and not speculative enough.”)
Just as a prior, I would think it’s more likely for motivated reasoning to generate the belief “it is optimal from a health perspective to spend more time doing something that makes me feel better while awake, and that doesn’t require any productivity during that extra time,” than “it is not optimal to spend more time on that, and if anything it is probably optimal to spend less time on that so you can do more effortfully productive things.”
I agree with that prior as stated but I think it’s reasonable to expect that other sources of motivated reasoning might dominate here, in both directions.