"Why We Sleep" Conclusion

The conclusion of an earlier version of the essay read

If you take one thing away from this entire essay, remember this: as long as you feel good, sleeping anywhere between 5 and 8 hours a night seems basically fine for your health [19] (see Section 1), regardless of whatever Big Sleep wants you to believe.

All of the evidence we have about sleep and long-term health is in the form of those essentially meaningless correlational studies, but if you’re going to use bad science to guide your sleep habits, at least use accurate bad science.

[19] Assumption here being that “feeling good” is a valid indicator of how much sleep your organism requires.

The conclusion now reads

We have literally no idea about the optimal for long-term health sleep duration. [19]

All of the evidence we have about this is in the form of those essentially meaningless correlational studies, but if you’re going to use bad science to guide your sleep habits, at least use accurate bad science.

[19] Correction: originally, this sentence read, “as long as you feel good, sleeping anywhere between 5 and 8 hours a night seems basically fine for your health”, but as several people pointed out, ironically, the only support for this statement comes from the correlational data, which I claimed cannot be used to establish causality.

The alteration appears to be a retreat, but the new conclusion is a stronger and less defensible claim.

  1. The original conclusion did not make a claim about sleep or causality. Saying something “seems basically fine” is a claim about one’s mental state. The charitable reader understands “seems” as “seems to me”, making this a self-documenting claim about the author’s point of view after 150+ hours of research. The conclusion states what the author concluded. The only way to discredit this statement is to question Guzey’s self-awareness, employing the Insanity Offense.

  2. The “as long as you feel good” [i.e., not sleepy] disclaimer further reduces the scope of the claim. “Drink more water if you feel thirsty” is not a perfect rule, but it strikes a good balance between the risk of promoting anxiety among people who fail to drink exactly eight glasses a day and the risk of endorsing rare but genuine disorders.

  3. Correlations are not meaningless. When people are eating more ice cream, the risk of murder is in fact higher. It is not completely unreasonable to worry about murder more when ice cream sales rise.

  4. We surely have some idea about sleep duration and health. Drowsy driving is a realistic sleep-related health concern. Most people in good health probably sleep 5-8 hours and shouldn’t worry about sleep duration, despite Dr. Walker’s appeals. People sleeping less than four hours or more than nine hours should consider investigating the cause of their outlier sleep habits, which could help them identify health risks.

  5. “Big Sleep” is funnier than “at least use accurate bad science.” The new phrase doesn’t make sense with the previous statement that bad correlational science gives us “no idea” about how sleep duration and health. What is the purpose of the red dashed line in the charts, if not to suggest that if you are getting 5-8 hours of sleep (and feel good), you probably don’t need to worry about your sleep duration?

  1. You’re kind of right but I don’t actually know if sleeping anywhere between 5-8 hours is actually good for your health. I do not have the data that would support this.

  2. This does not reduce the scope of the claim. I still write “seems basically fine for your health”

  3. I believe that in this case they are

  4. Drowsy driving is a bad example. It has no relation to the kind of health I discuss

  5. It is funnier. I do think that phrase makes sense. The purpose of the red line is to highlight the mortality at 8 hours of sleep suggested by Walker and at some other level where the mortality is the same

  1. So don’t claim to know that 5-8 hours is good for health. Just state that, after 150 hours of research where you found no good evidence of danger in that range, it seems to you that 5-8 hours is “basically fine.” Someone will object that they used to feel great sleeping 8 hours, but after changing jobs they only sleep 6 and they are always sleepy and miserable, so

  2. reduce the scope of your statement to people who sleep 5-8 hours and also “feel good.” The “seems basically fine for your health” language only appears in the footnote 19 correction, where it seems to be redacted.

  3. Do you have evidence showing that all the sleep-health correlations are meaningless? Perhaps “apparently meaningless” would be more defensible than “essentially meaningless.”

  4. Figure 1 depicts “risk of all-cause mortality.” The source is a meta-analysis, the first reference of which mentions accidents as the first reason for concern about sleep:

Several reasons underlie the increasing interest in sleep. First, sleep problems are associated with accidents and human errors. By 2020, the number of people killed in motor-vehicle crashes is expected to double to 2.3 million worldwide, of which approximately 230 000–345 000 will be due to sleepiness or fatigue. 11 Similarly, disturbed sleep has been shown to double the risk of a fatal accident at work over a 20-year period. 12

  1. Nothing like dissecting a joke: in the original version, the reader could use bad science to feel at ease with the 5-8 hours of sleep previously mentioned. The new version removes the 5-8 hours and says “We have literally no idea” so of what possible use, even jocular, is the bad science?
  1. The thing is, that statement is baseless. Why white 5-8 and not 5-9 or 4-9 or 3-10?

  2. this doesn’t change anything…

  3. https://www.gwern.net/Causality does this work?

  4. It seems to me that the primary source for this does not have these 230 000–345 000 figures and does not say anything at all about the number of deaths caused by sleepiness or fatigue. It only mentions “sleep-related” accidents, which it seems, mean accidents involving a driver who was sleepy or fatigued.

  5. the bad science can still be used if one takes the correlational data seriously, which is mentioned in that sentence and which is shown in the picture right after