Is an all-nighter worth it?

These were laboratory experiments, mostly in young adults, in which one group stayed awake all night under supervision while the other slept normally, and both then learned new material such as word lists or pictures and were tested on it. They were not students cramming for a real exam, and the tests measured memory for the lab material, not grades. Most studies assigned people at random or in a balanced order, but not every one did. Most were small, and the review found signs that studies with weaker results were less likely to be published; after the authors adjusted for that, the difference was smaller (g 0.463) but still clear. Few studies checked that the sleep-deprived group did not nap during the day before learning. The review gave no overall rating of how certain the evidence is.

The difference was larger when people were tested the day after the sleepless night, and smaller, though still present, when they had slept for one night or more before the test - so part of the difference may come from being tired at the test rather than from memory alone. Most individual studies were too small to detect a difference this size reliably, and after the authors adjusted for studies that seem to have gone unpublished the estimate shrank (g 0.166) but stayed above zero. In the review's quality check, the studies that described their methods most completely - mainly how they chose and excluded volunteers - did not show a clear difference on their own; the authors think this may partly reflect less complete reporting by the other studies rather than no effect, but it is a reason to treat the size of this result with caution. Most studies compared two separate groups of people rather than the same people twice. Most participants were young adults, and the tests were lab memory tasks, not exams.

Most studies compared the same people's scores when rested and again after the sleepless night, rather than comparing two groups assigned at random. Time awake ranged from about one day to two days, so this is not an estimate for exactly one night. The studies were small - about 21 people on average - and were published up to 2008. Simple attention tests are repetitive tasks such as pressing a button whenever a light appears; they show how attention slips, not how well someone does at a real task like driving or sitting an exam. The review's summary gives a slightly different figure for this result (g -0.776, 95% CI -0.96 to -0.60); the figure here is the one in its results table and results text.