The sleep duration guide is about what cohorts observe. This one is about what experiments do: take healthy adults, assign them a number of hours in bed, keep them there for days or weeks, and measure what changes. Those experiments are the reason short sleep is treated as a cause of harm rather than a correlate of it.

The two landmark studies were both published in 2003 and have been replicated in pieces many times since. Their central finding is uncomfortable: the impairment accumulates, and the impaired person stops noticing it after a few days.

Sleep pressure builds with time awake and is discharged by sleep. When sleep is cut, the debt carries over, and performance on vigilance tasks degrades in proportion to the cumulative hours of wakefulness beyond what the brain can absorb. Subjective sleepiness, meanwhile, adapts: people rate themselves only slightly sleepier on day fourteen than on day three while their lapses keep climbing.

The physiological side runs through the stress and metabolic systems. Restriction raises evening cortisol, blunts the insulin response and raises hunger, and in athletes it raises perceived effort for the same work. Whether the short-term metabolic changes drive long-term disease is the link that is still being tested.

Cognition: dose-dependent and cumulative. In the 2003 Pennsylvania study, 48 healthy adults were randomized to four, six or eight hours in bed for 14 days, or to three nights of total sleep deprivation. The four- and six-hour groups showed steadily accumulating deficits on every cognitive task; by the end, six hours a night had produced impairment equivalent to up to two nights without sleep, and the participants were largely unaware of it. The Walter Reed study the same year assigned 66 volunteers to three, five, seven or nine hours for a week: performance in the three-hour group declined steadily, the five- and seven-hour groups declined then stabilized at a lower level, and three nights of recovery sleep did not restore the five- and seven-hour groups to baseline.

A 2017 meta-analysis of 61 studies confirmed the picture: sleep restriction impairs cognition across domains, with the clearest effects on sustained attention and executive function and smaller effects on long-term memory.

Physical performance. Nine nights of five hours in bed, in resistance-trained women, slowed bar speed in lower-body lifts by up to 15%, raised cortisol by 42% and raised perceived exertion, with only trivial loss of total volume. The quality of training suffers before the quantity does.

Extension: sleep improves, metabolism may not. A 2026 trial gave 14 short-sleeping adults with overweight or obesity about an hour more sleep a night for six weeks. Sleep health improved. Insulin sensitivity measured by clamp, in every organ system tested, did not, nor did 24-hour glucose control. One trial, small, but a direct test of a widely assumed benefit that did not appear.

The experiments describe the harm rather than cause a separate one. Acute sleep loss impairs driving and safety-critical work at a level comparable to alcohol, and the unawareness documented above is why self-assessment is unreliable. People with insomnia should note that deliberate sleep restriction, under supervision, is a core component of cognitive behavioral therapy for insomnia and works precisely because it consolidates sleep; the same tool is therapeutic there and harmful in people who can sleep but choose not to.

No product, no regulator. Work-hour limits for drivers, pilots and medical trainees are the regulatory expression of this literature, and they vary by country and industry.