After several short nights, feeling reasonably awake can seem like evidence that you have adapted. Sleep researchers test that assumption by measuring experience and performance separately. Three experiments, spanning a classic laboratory study and two 2026 papers, illustrate why the answer depends on what is measured.
How you feel may change less than how you perform
A 2003 study involving 48 healthy adults aged 21–38 compared restricted sleep schedules with total sleep deprivation. In the restriction experiment, participants were assigned four, six or eight hours in bed nightly for 14 days. The separate total-deprivation experiment allowed no sleep for three nights.
Performance worsened cumulatively with the shorter nightly schedules. Reported sleepiness initially increased but changed much less over subsequent days. The mismatch matters: becoming accustomed to a feeling did not demonstrate preserved performance on the tested tasks.
These were controlled laboratory schedules in young adults. They do not provide a precise prediction for a particular person's commute, workday or sleep requirement.
Eye movements reveal another kind of performance
In a 2026 randomized crossover study, 12 participants, six female, completed a four-week protocol. Researchers compared a week with five hours in bed nightly against a week with nine; each experimental week followed a washout week with nine hours in bed each night.
During laboratory visits after each experimental week, a tracking task every two hours measured how the eyes followed a moving target. Restricted sleep produced modest deterioration in smooth tracking and changes in the rapid corrective movements called saccades. Some features changed while others were comparatively preserved.
The study illustrates why passing one task cannot settle every question about performance. Its small sample also leaves substantial uncertainty about individual assessment. NASA and the San José State University Research Foundation funded the work. One author disclosed being an inventor on relevant NASA-owned patents, without involvement in their commercialization.
A brain pattern answers a different question again
A July 2026 imaging study developed a machine-learning signature using a discovery sample of 151 healthy adults assigned to normal sleep, half-night sleep or a night awake. The classifier distinguished sleep-deprived from rested participants using patterns of connections between brain regions, and was tested in independent datasets.
In the discovery sample, a difference remained after one recovery night. That finding identifies a measurable residual pattern; it does not by itself establish permanent damage, a diagnosis or an individual threshold for safe activity.
Funding included the National Institute of Mental Health and military research support. Disclosures included one author's pharmaceutical research or consulting relationships and another's ownership of NOCTEM Health.
Keep the question attached to the measurement
For readers, the useful distinction is between a report of experience, a demonstrated ability on a task and a statistical pattern in physiology. Each can add information. None automatically supplies the answers provided by the others.
When evaluating a sleep claim, look for the schedule, the population and the actual outcome. Ask whether researchers tested attention, visual tracking, memory or subjective sleepiness. A paper about one of those outcomes should retain that specificity when it becomes a headline—or a personal interpretation.