The Number Everyone Believes
Roughly one in three American adults regularly sleeps fewer than seven hours, according to the CDC's 2024 National Health Interview Survey, which puts the number at approximately 83 million people who report feeling fine, hold jobs, exercise regularly, and drive on public roads without a second thought about what the lost hours might be costing their brains.
In 2003, a team at the University of Pennsylvania's Division of Sleep and Chronobiology put that assumption in a laboratory and measured what happened when "fine" was tested with actual cognitive instruments over two full weeks of controlled observation.
What Fourteen Nights Revealed
Hans Van Dongen, Greg Maislin, Janet Mullington, and David Dinges recruited 48 healthy adults between ages 21 and 38 and randomized them to one of four conditions: eight hours in bed per night for 14 days, six hours for 14 days, four hours for 14 days, or zero hours for three consecutive days. The experiments ran under continuous monitoring in a controlled laboratory setting, with all other sleep prohibited and waking activities standardized to eliminate confounds.
Every two hours during waking periods, participants completed a battery of neurobehavioral tests, the most important of which was the Psychomotor Vigilance Task, a reaction-time test that records how quickly a person responds to a visual stimulus. PVT lapses, defined as responses slower than 500 milliseconds, are a validated marker of cognitive impairment with high sensitivity to sleep loss and essentially no learning curve, meaning changes in lapse rate reflect genuine neurological deterioration.
The eight-hour group held steady. Not a single cognitive metric declined across all 14 days. The four-hour group declined rapidly, and by day 14 their performance had deteriorated to a level equivalent to three consecutive nights without sleep. But the six-hour group is where the finding turns unsettling: their PVT lapses accumulated in a near-linear trajectory that showed no sign of leveling off, and by day 14 their cognitive performance deficits were statistically equivalent to those of subjects who had been kept awake for two full days. Working memory followed the same dose-dependent erosion, with deficits accumulating night after night and no plateau in sight.
The Adaptation Illusion
If the six-hour sleepers had felt terrible, the finding would be alarming but self-correcting, because people who feel impaired seek more sleep. They did not.
Subjective sleepiness ratings showed a sharp initial response to sleep restriction and then flattened within days. Objective performance kept declining. By day 14, subjects performing at the cognitive level of someone awake for 48 hours rated themselves only slightly sleepy, and the gap between how impaired they were and how impaired they believed themselves to be had widened every passing night until the two measures occupied entirely different worlds.
Van Dongen's team identified this dissociation as the defining danger: "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."
The mathematical model was clean and startling in its universality. Regardless of whether sleep was lost chronically over many nights or acutely in a single marathon of wakefulness, performance lapses were near-linearly related to cumulative wakefulness exceeding roughly sixteen hours per day, regardless of whether the excess accumulated gradually or all at once. Below that threshold, the brain recovers; above it, every extra hour of wakefulness extracts a neurobiological cost that compounds across days without limit.
The Replication
That same year, Gregory Belenky and colleagues at the Walter Reed Army Institute of Research published an independent dose-response study with 66 volunteers restricted to three, five, seven, or nine hours in bed for seven days, confirming the core pattern: dose-dependent, cumulative impairment with no adaptation. Even one week of five-hour sleep produced significant deficits, and three days of recovery sleep failed to restore baseline performance.
Banks and Dinges reviewed both studies and the broader literature in a 2007 synthesis, concluding that chronic sleep restriction produces "the same magnitude of cognitive impairment" as acute total sleep deprivation and that subjects reliably underestimate their deficits. The collective evidence has been cited more than 3,000 times, and no serious challenge to the core findings has emerged in two decades.
The Strongest Counterargument
The most forceful objection is ecological validity. Van Dongen's experiment confined participants to a windowless laboratory for weeks, eliminating caffeine, exercise, sunlight, and naps โ all compensatory mechanisms that real-world sleepers use routinely. Jim Horne at Loughborough University has argued that a "core sleep" of about six hours is biologically sufficient for many adults and that laboratory impairments reflect the artificial removal of normal countermeasures rather than an inherent cost of short sleep.
This objection runs into the study's most troubling implication. Caffeine temporarily restores alertness but does not fully restore working memory or complex reasoning to baseline. And the adaptation illusion means the very people who would benefit most from compensatory strategies don't know they need them: the pilot who feels alert, the surgeon who feels sharp, and the driver who feels fine may all be operating with the reaction time of someone who hasn't slept in two days.
What We Didn't Prove
The sample is 48 adults, all healthy, ages 21 to 38, with no sleep disorders, and whether these dose-response curves apply to older adults, adolescents, or individuals with chronic conditions remains untested. Individual differences in vulnerability to sleep loss are substantial โ some participants showed far less impairment at six hours than others โ and the study was not powered to identify genetic or trait-based moderators. The 14-day window, while longer than most sleep restriction studies ever attempted, does not tell us whether the trajectory would continue declining indefinitely or stabilize over months of habitual short sleep.
One number the public health conversation has ignored: approximately 83 million American adults sleep fewer than seven hours per night. At the six-hour end, that is a population whose cumulative cognitive impairment matches someone who hasn't slept in two days, making medical decisions and driving on public roads at a deficit they cannot detect.
The Bottom Line
A landmark controlled experiment found that two weeks of six-hour sleep produces cognitive impairment equivalent to 48 hours of total sleep deprivation. The subjects couldn't tell. An independent military study replicated the pattern the same year, and no serious challenge has emerged in the two decades since. The finding implies that roughly 83 million Americans who routinely sleep six hours or fewer are making critical decisions at a level of impairment they cannot detect.
What You Can Do
If you sleep six hours and feel fine, that feeling is precisely the problem the study identified. Track your actual sleep duration and aim for at least seven hours in bed, because the cognitive cost is cumulative and weekend catch-up does not fully erase it. If your schedule makes seven hours impossible on weekdays, extending sleep by even 30 minutes per night slows the accumulation curve. Naps of 10 to 20 minutes provide temporary alertness gains but do not substitute for nocturnal sleep. And if you use caffeine to compensate, know that it masks subjective sleepiness without restoring the deeper cognitive functions the PVT measures.