Fourteen nights at four, six and eight hours
80 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- State the design of the dose-response experiment, what it measured, and which of its numbers came out of a model
- Explain what cumulative, dose-dependent deficits means in terms you can restate
- Say what the experiment licenses and what it does not, given its participants and its outcome
Almost everything you have been told about sleep is an association. Somebody asked a lot of people how long they sleep, waited, and counted what happened to them. Lesson 3 is about that literature and it is enormous.
This lesson is the other kind. One experiment, forty-eight people, everything controlled, sleep taken away on purpose by an amount somebody chose. It's the strongest evidence in this course and it's also the narrowest.
The design
In 2003 a team published a study that did something the observational work cannot do: it assigned people an amount of sleep and kept them to it.1 This course read the abstract and didn't open the paper, so everything below is at that level, and the lesson will say so again where it matters.
Their design, in their own words: "The chronic sleep restriction experiment involved randomization to one of three sleep doses (4 h, 6 h, or 8 h time in bed per night), which were maintained for 14 consecutive days. The total sleep deprivation experiment involved 3 nights without sleep (0 h time in bed). Each study also involved 3 baseline (pre-deprivation) days and 3 recovery days."1
Their participants: "A total of n = 48 healthy adults (ages 21–38) participated in the experiments."1
And the setting: "Both experiments were conducted under standardized laboratory conditions with continuous behavioral, physiological and medical monitoring."1
Read those three sentences together and you have the shape of the thing. Three doses, fourteen nights, healthy people in their twenties and thirties, in a laboratory, watched continuously. The forty-eight is the total across both experiments, the fortnight of restriction and the three nights without sleep; how many were in each arm is not in the abstract. No sleep outside the assigned hours was allowed, of any kind: "All other sleep was prohibited."1
Notice the eight-hour arm. It is not a group sleeping normally; it is a group held at eight hours in bed, which for some of them will be more than they usually get and for others less. That is what a dose is, and it is why this is a dose-response experiment rather than a comparison of the sleepy with the rested.
And notice the unit, because the whole lesson has to hold it. The doses are hours in bed, not hours asleep. Nobody sleeps every minute they are in bed, so somebody who actually sleeps six hours is getting more sleep than the six-hour arm did, not the same. Every "six hours" below means six hours in bed.
What happened
The Results sentence is worth taking one clause at a time, because each clause is a different finding and they don't all point the same way.
"Chronic restriction of sleep periods to 4 h or 6 h per night over 14 consecutive days resulted in significant cumulative, dose-dependent deficits in cognitive performance on all tasks."1
Two words in there are doing the work. Cumulative means the deficit kept growing across the fortnight rather than arriving on night two and sitting still. Dose-dependent means four hours was worse than six, which was worse than eight. Put together: the cost went up with how much sleep was taken and it went up again with how long it went on.
And here is the first thing this course cannot tell you. The abstract says "all tasks" and never says what the tasks were. So this lesson cannot name them, cannot tell you how long they took, and cannot tell you whether any of them resembles anything you do. A lesson that filled that in would be making it up, and the gap is worth feeling, because it's the gap between what a study measured and what you want to know.
Before the next clause. The same study asked the participants how sleepy they felt. What do you expect it found, and in particular, do you expect the four-hour group to have reported feeling worse than the six-hour group?
Show the answer
Write your answer down before you read on, because this result is much easier to explain away once you know it.
Most people expect the ratings to track the performance: worse sleep, worse scores, and people saying they feel terrible in proportion.
What the abstract reports is that they came apart. Subjective sleepiness responded at once and then almost stopped moving, and it "did not significantly differentiate the 6 h and 4 h conditions".1
So the ratings did not separate the four-hour group from the six-hour group, while the performance did. The instrument inside your head, the one you actually use to decide whether you slept enough, was not tracking the thing that was changing.
The authors say what they think that means, in the Conclusions, and it is worth quoting rather than paraphrasing: "Sleepiness ratings suggest that subjects were largely unaware of these increasing cognitive deficits, which may explain why the impact of chronic sleep restriction on waking cognitive functions is often assumed to be benign."1
Read the hedge. "suggest", and "may explain". This is the authors offering an account of why a belief is common, not reporting a measurement of the belief.
The measures that didn't move
There is a third clause and it is the one nobody quotes, which is a reason to give it a section.
"Polysomnographic variables and δ power in the non-REM sleep EEG—a putative marker of sleep homeostasis—displayed an acute response to sleep restriction with negligible further changes across the 14 restricted nights."1
Delta power is slow electrical activity in the sleep trace, which is background this course is supplying rather than anything the abstract says. What the abstract gives is the authors' own phrase for it, "a putative marker of sleep homeostasis", and putative is their word for a thing widely taken to be a marker rather than a thing shown to be one.2
What matters here is not the physiology, it is the shape. The overnight measures jumped once and then sat still. The daytime performance kept falling for a fortnight. Two quantities, measured on the same people on the same nights, moving differently.
That is lesson 1's point arriving with teeth. A physiological measurement of sleep and a measurement of what sleep loss costs you are not the same number, and in this study they are visibly not the same number. If you had only the overnight trace, you would have concluded early in the fortnight that nothing much else was happening.
The arm with no sleep in it
The study ran a second arm: three nights with no sleep, and then compared the two kinds of loss.
"Comparison of chronic sleep restriction to total sleep deprivation showed that the latter resulted in disproportionately large waking neurobehavioral and sleep δ power responses relative to how much sleep was lost."1
So three nights with no sleep at all cost more, per hour of sleep lost, than a fortnight of short nights did.
The Conclusions then put an equivalence the other way round, and this is the sentence that gets quoted out of this paper: "chronic restriction of sleep to 6 h or less per night produced cognitive performance deficits equivalent to up to 2 nights of total sleep deprivation".1 That is a separate claim from the one above, not the same finding restated.
Hold "up to" firmly. It is a ceiling, over fourteen nights, on tasks the abstract does not name, in a laboratory. It is not a conversion rate you can apply to your own week.
The number that isn't an observation
One more sentence, because it is the one most often repeated as though somebody watched it happen: "A statistical model revealed that, regardless of the mode of sleep deprivation, lapses in behavioral alertness were nearlinearly related to the cumulative duration of wakefulness in excess of 15.84 h (s.e. 0.73 h)."1
"A statistical model revealed" is the authors telling you where the figure came from. It is a parameter they fitted, with a standard error attached, across four experimental conditions. Nobody observed a threshold at 15.84 hours and nobody measured one person's.
Why that distinction is worth a paragraph: a fitted parameter is a summary of a model that describes data, and it inherits every assumption in the model. A reader who takes 15.84 hours as "the number of hours you can be awake before you start making mistakes" has turned a description into a limit, and has also quietly applied a fortnight of laboratory data to a single Tuesday.
So is six hours a night enough, or isn't it?
Show the answer
The honest answer has two halves and the second is the more useful one.
On what this experiment measured, six hours in bed was not enough, and the deficits accumulated across the whole fortnight, with the abstract reporting no point at which they stopped. That's a real result, on healthy 21-to-38-year-olds, in a laboratory, on unnamed cognitive tasks, with everything else controlled.
And the question you actually asked isn't one this study can answer, which is worth being exact about rather than waving at. You want to know whether six hours is enough for you, doing your work, at your age, with everything in your life that the laboratory removed. This study had forty-eight people, none over thirty-eight, doing tasks the abstract doesn't name, with nothing else to do all day.
And the thing that makes the gap uncomfortable rather than academic is the sleepiness finding. The ratings did not separate the people held at six hours from the people held at four, while the performance did, which means the feeling you would use to answer the question for yourself is the one measurement here that did not track the cost. That's not a reason to assume you're impaired. It's a reason to notice that feeling fine is not evidence either way, which is this course's reading of the finding rather than something the paper says.2
Somebody tells you that 15.84 hours is the longest a person can safely stay awake. Name the two things wrong with that sentence.
Show the answer
Both are in the paragraph above, and the second is the one people miss.
It isn't an observed threshold. The abstract says a statistical model revealed it, which means it is a parameter fitted to data rather than a point anybody watched somebody cross. Fitted parameters inherit every assumption in the model they came from.
And it isn't about a person. It was fitted across four experimental conditions in a laboratory study of forty-eight people, and it describes the relationship between accumulated wakefulness and lapses across all of them. There is no version of it that belongs to one reader's Tuesday.
A third thing, if you want it. "Safely" is not in the sentence the paper wrote. Lapses in behavioural alertness is what was modelled, and safety is a claim about consequences that nobody here measured.
One person, worked through
This case is constructed, because no source read for this course follows an individual.2
Nadia has been in bed about six hours a night for the last two weeks, on purpose, to finish something. She feels tired in the mornings and fine by ten. Her work has gone out on time and nobody has complained.
Take the study to her, clause by clause, and see what survives.
What it licenses. She is in the age range the study used, and if she were in that laboratory doing those tasks, the expectation from this design is that her scores would be lower than they were a fortnight ago, and lower again than they were a week ago.
What it does not license. Anything about the work she actually did. The study measured performance on tasks the abstract doesn't name, not the output of a job. Nobody has shown her a task she was worse at.
And the thing that isn't evidence. That nobody complained, and that she feels fine by ten. The study's whole point is that those two reports came apart from the measurement.
Before the last line of this. Nadia asks you a fair question: given all that, what would actually settle it for her?
Show the answer
The answer is short, and how short it is turns out to be the point.
A measurement of her, on the thing she cares about, taken twice. Her own performance at something she can score, at the end of a normal fortnight and at the end of a short-sleep one, in some order that is not confounded with which project she happened to be on.
Which is a real experiment on one person, it is difficult, and almost nobody does it. That is the honest end of the road here, and the course project is the nearest practical version of it.
Notice what is not in the answer: no threshold, no number of hours, and nothing from the study above. A result measured on forty-eight strangers can tell her what to expect and cannot tell her what happened, which is the distinction this whole course is built on.
Three things people get wrong about this study
"Six hours is fine, you adapt." The design ran for fourteen consecutive nights specifically to look for that, and the abstract reports the deficits accumulating with no point at which they stopped.1
"Sleep debt is just a metaphor." The authors propose the opposite in their final sentence, and they propose it: sleep debt "is perhaps best understood as resulting in additional wakefulness that has a neurobiological 'cost' which accumulates over time".1 Perhaps best understood is a proposal about how to think, not a measurement.
"This proves short sleep is dangerous." It shows a cost to cognitive performance in a laboratory over two weeks. Danger is a health claim, and no health outcome was measured here at all.
Practice
Take twenty minutes.
This experiment shows a cost to performance. Suppose you wanted to show that six hours a night harms health rather than performance. Write down the study you would have to run.
Four things, and the fourth is the point.
- The outcome. Something a doctor would recognise, named specifically.
- How long you would have to run it, given how long that outcome takes to appear.
- How many people, roughly, and why more than forty-eight.
- What would go wrong with it. Who drops out, what they do at home when nobody is watching, and what you would be forced to accept instead of control.
The fourth answer is why that study mostly does not exist, and why lesson 3's evidence is observational. Write two lines on that before you move on.
Take 15 minutes now. Take 10 more in a fortnight, which sit outside this lesson's stated time.
This is preparation for the project's log section and it works on its own.
Write down, today, three things you'd be able to notice if your own performance dropped by a small amount. Not "I would feel tired": something you could point at. A task you time, a number of errors you count, a thing somebody else would remark on.
Then, in two weeks, come back and answer one question: could you actually have noticed any of the three, if it had happened?
Most people find that at least two of their three were unfalsifiable when they wrote them, which
is this course's expectation rather than something it has measured,2 and finding that out is the
exercise. The project in assessments/ is where this gets done properly, with two instruments and a week of
nights.
Connections
Back. Lesson 1 gave you four instruments; this study used polysomnography overnight and a set of performance tasks by day, and the two did not move together. Focus and Deep Work is where attention as a measured quantity was taught, and this is what taking it away on purpose looks like.
Forward. Lesson 3 asks the same question with an instrument that could not be more different: a questionnaire, given to more than a million people, with death as the outcome.
Go deeper
- The Cumulative Cost of Additional Wakefulness, Van Dongen, Maislin, Mullington and Dinges, 2003. This course read the abstract and nothing else. The abstract is long, structured and unusually informative, and reading it is a good exercise in its own right: every quotation in this lesson comes from it, so you can check the lot in ten minutes.
Sources
- Hans P. A. Van Dongen, Greg Maislin, Janet M. Mullington and David F. Dinges, "The Cumulative Cost of Additional Wakefulness: Dose-Response Effects on Neurobehavioral Functions and Sleep Physiology From Chronic Sleep Restriction and Total Sleep Deprivation", SLEEP 26(2), 2003, pp. 117 to 126. Read at abstract level: the structured abstract verbatim and in full, from the journal's own abstract page. The paper was not opened, which the body says twice. Supports: the design, the participants, the setting, the prohibition on other sleep, the cumulative dose-dependent deficits, the subjective sleepiness result, the polysomnographic and delta power result, the comparison with total sleep deprivation, the "up to 2 nights" conclusion, the fitted 15.84 hour parameter, and the sleep debt proposal. No task in the study is named anywhere in this lesson, because the abstract names none.
- The course's own constructions, each labelled where it appears in the body. Nadia's case is invented and the body says so where it begins. The reading of the sleepiness finding as meaning that feeling fine is not evidence either way is this course's, said as such in the checkpoint rather than attributed to the paper. The expectation that most readers' three signals will turn out to be unfalsifiable is an expectation, and the exercise says so. The one-line description of what delta power is is background this course supplied rather than anything the abstract says, and the body says so at the point of use.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.