An extra hour of revision costs about eight minutes of sleep

What the diaries showed

In 2013 Cari Gillen-O’Neel, Virginia Huynh and Andrew Fuligni published a study with a design that most research on this question lacks. Five hundred and thirty-five secondary students in Los Angeles kept a daily diary for fourteen days, recording how long they had studied, how long they had slept, and whether they had had academic trouble that day: not understanding something taught in class, or doing badly on a test, quiz or piece of homework. They did it again a year later, and again two years after that.¹

The design matters because of what it removes. Comparing students against each other confounds everything: the students who study more are different students, and their marks are different for a hundred reasons. A diary lets each student be compared against their own normal. The question becomes narrower and answerable. When this student studies more than they usually do, what happens to them tomorrow?

The answer in the authors’ own summary: regardless of how much a student generally studies, if they sacrifice sleep to study more than usual, they will have more trouble understanding material taught in class and be more likely to struggle on an assignment or test the following day. And because students are increasingly likely to make that trade in the later years of school, the pattern gets more common as the exams get closer.¹

Where the hour comes from

The part that is rarely reported is how small the trade usually is. A day is a fixed budget, so an extra hour has to come out of something, and it comes out of whatever is least anchored. Bedtime is held in place by tiredness and by what time you have to be up. An evening of television, phone and other people is held in place by nothing at all.

So for the ordinary night, the all-or-nothing framing of study against sleep is simply wrong. Most of an extra hour is taken from the loose part of the evening. What the study then shows is where the damage lives: the next-day trouble is carried by the sleep that was lost, not by the extra revision itself. When the hour came out of the evening, it was close to free. When it came out of the night, it cost.¹

That is a licence more than a restriction. It does not say study less. It says that where you take the hour from decides whether it costs you anything, and that the hour is usually there to be taken from somewhere other than sleep.

Why you cannot feel the cost

The reason this is worth an article rather than a sentence is that the cost is invisible from the inside. In 2003 Van Dongen and colleagues kept forty-eight healthy adults in a laboratory for two weeks, allowed four, six or eight hours in bed a night, and tested them every two hours; a fourth group was kept awake for three nights straight.²

The four-hour and six-hour groups got steadily worse for the whole fortnight, in what the paper calls cumulative, dose-dependent deficits on every task. Their own ratings of how sleepy they felt did something different: an initial rise, and then almost nothing. Two weeks in, on six hours a night, people were making many more attention errors than they had at the start, and reporting that they felt about the same as they had after the first few days. The authors’ phrase is that the subjects were largely unaware of the deficits.²

That is the whole problem with deciding this by feel. The instrument you would use to judge whether you are short of sleep stops moving before the damage does. It is not that the gauge reads zero; it is that it has a short scale, pinned near the top, attached to something that keeps rising past the end of it.

So how many hours

There is no number here, and the absence is the honest answer. What the two studies say between them is this. Extra revision on its own did not hurt the next day. Extra revision paid for with sleep did, and the effect grows with the year group. And you cannot tell from the inside when you have crossed from the first case into the second, because the feeling of being fine is one of the first things a short night takes.

• Take the hour from the evening. That is the study’s own practical reading. The loose part of the day is usually longer than it feels.
• Count, do not consult. Hours in bed is a number you can check without asking a faculty that has been switched off. If the count says short and you feel fine, the count is the better instrument.
• Protect the night before a test more than the night before a test week. The trouble measured was next-day trouble. A short night lands on tomorrow.
• Do not budget by hours revised. The diaries did not find that more hours predicted a better day. They found that hours taken from sleep predicted a worse one.

What this study cannot say

The sleep in the diary study is self-reported, and adolescents overstate their sleep by a good margin when a device is recording alongside them. A diary cannot rule out a third thing driving both the long evening and the short night, such as a deadline that would have wrecked the evening either way. And the students were in Los Angeles, in a different school system, which is not a reason to disbelieve the direction and is a reason not to quote the minutes as though they were yours.¹

The laboratory study was on adults aged twenty-one to thirty-eight, and its authors say explicitly that they do not know whether the same accumulation applies to adolescents, or whether even their eight-hour baseline was enough.² Everything here is about the shape of the effect, not its exact size for a seventeen-year-old in May.

And nothing here is a claim about Kerfox. The app has a schedule and a memory model; it has no data on anyone’s sleep and no study behind it. This is what two careful papers found, at the size they found it, and what a person might sensibly do with that.