If you stayed awake for 72 hours, your attention would fail first, then your perception, then your grip on reality. By 17 to 19 hours you test as badly as someone over 0.05% blood alcohol, by 24 to 48 hours you get perceptual distortions, by 48 to 90 hours complex hallucinations, and from about 72 hours delusions appear. A normal night or two of sleep usually reverses it, but the risk on the way, especially behind a wheel, is real.
- Beyond 15.84 hours awake a day, attention lapses start to pile up (Van Dongen 2003).
- At 17-19 hours awake, some test scores match or fall below 0.05% blood alcohol, worse than India's 0.03% driving limit.
- 24-48 h: perceptual distortions; 48-90 h: complex hallucinations; 72 h+: delusions (Waters 2018).
- Two weeks of 4-hour nights builds more excess wakefulness than 72 hours straight, in the model.
- Recovery takes one or two long sleeps for most, but not always all symptoms.
Seventy-two hours is a useful line because it is roughly the longest anyone now studies on purpose. Research ethics committees treat more than 48 hours of total sleep deprivation as hard to justify.
So most of what we know about day three comes from older experiments, military studies and a few famous stunts. The picture is still remarkably consistent.
Why does staying awake get harder by the hour?
Staying awake gets harder because sleep pressure builds continuously while you are awake and only sleep clears it. The most famous person to test that limit was a teenager in San Diego.
From December 1963 to January 1964, Randy Gardner, a 17-year-old student at Point Loma High School in San Diego, stayed awake for 264.4 hours: 11 days and 25 minutes. Two classmates, Bruce McAllister and Joe Marciano Jr., kept a log. Stanford sleep researcher William Dement came to observe, and a US Navy officer, Lt. Cmdr. John J. Ross, monitored his health.
On day 10 Gardner could still beat Dement at pinball. But Ross recorded moodiness, trouble concentrating, short-term memory problems, paranoia and hallucinations. On day 11, asked to keep subtracting 7 from 100, Gardner stopped at 65. Asked why, he said he had forgotten what he was doing.
Wikipedia: Randy Gardner (record holder)The two-process model
Sleep scientists describe this with the two-process model:
- a homeostatic "Process S" that rises with every waking hour and only sleep clears;
- a circadian "Process C", a roughly 24-hour clock that pushes alertness up in the day and down at night.
You feel worst when high pressure meets the circadian low in the small hours of the night, which is why night two of a 72-hour vigil is far harder than evening one.
Putting numbers on the pressure
The best quantitative version of Process S comes from a 2003 study in Sleep by Hans Van Dongen, David Dinges and colleagues at the University of Pennsylvania. They kept people in the lab on 4, 6 or 8 hours in bed for 14 days, or with no sleep for 3 nights, and tested them every two hours on the psychomotor vigilance task (PVT), a 10-minute reaction-time test that counts "lapses", responses slower than half a second. Their model found that lapses rise almost linearly with cumulative wakefulness beyond a critical daily wake period of 15.84 hours (standard error 0.73 h).
Here \( y \) is the rise in PVT lapses over baseline, \( \gamma \) is a person-specific rate, and \( \theta = 0.78 \pm 0.04 \) is the curvature they estimated for PVT lapses. Run it for 72 hours straight:
- Excess wake for 72 hours awake. \( \Sigma_{72} = 72 - 15.84 = 56.2\ \text{h} \).
- Excess wake for two weeks on 6 hours a night. Each day adds \( 24 - 6 - 15.84 = 2.16\ \text{h} \), so \( \Sigma = 14 \times 2.16 = 30.2\ \text{h} \).
- Relative lapses. \( (30.2/56.2)^{0.78} = 0.54^{0.78} \approx 0.62 \): two weeks of 6-hour nights produce about 60% of the lapse increase of 72 hours awake, in this model.
Calculated in this post from the Van Dongen et al. 2003 model (critical wake period 15.84 h, curvature 0.78): 14 nights × 2.16 h = 30.2 h against 56.2 h for 72 hours awake.
- Two weeks on 4 hours a night. \( \Sigma = 14 \times 4.16 = 58.2\ \text{h} \), slightly more than 72 hours awake.
Calculated in this post from the Van Dongen et al. 2003 critical wake period of 15.84 h: 14 nights × (24 − 4 − 15.84) h.
That last line is the uncomfortable result. The paper itself reports that the 4-hour group reached lapse levels seen after two nights without sleep, and the 6-hour group levels seen after one, while rating themselves only slightly sleepy. The model is a fit to healthy young adults in a lab, not a law of nature, and the day-averaged comparison blurs the exact hour, so read the outputs below as rough equivalents.
Your sleep debt as hours of continuous wakefulness
What happens to your brain in the first 24 hours awake?
In the first 24 hours the main casualty is sustained attention: you start missing things in short gaps. A 2010 meta-analysis by June Lim and David Dinges in Psychological Bulletin pooled 70 articles, 147 datasets and 1,533 participants deprived of sleep for up to 48 hours. The biggest effects were on simple attention, the ability to keep responding to a boring signal. Reasoning held up best.
See the data as a table
| Effect size (Hedges g (absolute)) | |
|---|---|
| Simple attention: lapses | 0.762 |
| Simple attention: reaction time | 0.732 |
| Working memory: accuracy | 0.555 |
| Working memory: reaction time | 0.515 |
| Complex attention: accuracy | 0.479 |
| Short-term memory: recall | 0.383 |
| Complex attention: reaction time | 0.312 |
| Processing speed: reaction time | 0.302 |
| Reasoning: accuracy | 0.125 |
Source: Lim & Dinges, Psychological Bulletin 2010, Table 2 · All effects are impairments. Reasoning accuracy was not statistically significant.
Pooled 147 datasets from 70 articles (1,533 participants) on total sleep deprivation of 48 hours or less. Simple attention lapses showed the largest impairment (g = −0.76), working memory accuracy was moderately hit (g = −0.56), and reasoning accuracy barely moved (g = −0.13, not significant). Limit: only up to 48 hours, mostly young lab volunteers, and many different tests lumped into domains.
Read the paper (free)Why attention and not reasoning? The leading explanation is "state instability": the sleep-deprived brain keeps slipping into brief sleep-like states and back. A 2025 study in Nature Neuroscience from Laura Lewis's lab at MIT, with 26 volunteers scanned after a sleepless night and after a normal one, found that attention lapses lined up with waves of cerebrospinal fluid flowing out of the brain, slower breathing and heart rate and constricted pupils, which normally happen during sleep. In plain terms, parts of the waking brain start doing night-time housekeeping on their own, and your attention goes missing while they do.

Is being awake for 24 hours like being drunk?
On some tests, yes, and earlier than 24 hours. Ann Williamson and Anne-Marie Feyer tested 39 people (30 transport workers and 9 army personnel) in both a sleep-deprivation session and a graded-alcohol session, published in Occupational and Environmental Medicine in 2000. After 17 to 19 hours awake, roughly 10:30 pm to 1 am for someone who woke at 6, performance on some tests was equivalent to or worse than at 0.05% blood alcohol, and response speeds were up to 50% slower on some tests. With longer wakefulness, performance reached the level of the highest dose given, 0.1%.
Within-subject comparison of 39 people across a 28-hour wake period and graded alcohol doses. After 17-19 hours awake, some measures matched or fell below 0.05% BAC; with longer wakefulness, some reached the 0.1% equivalent. Limit: computer-based tests, not driving or crash data, and a small sample.
Read the paper (free)By the end of a long day, before any all-nighter has even started, a driver's test performance can be worse than what India would prosecute if it came from a drink.
Against India's drink-driving limit
Put that next to Indian law. Section 185 of the Motor Vehicles Act sets the limit at 30 mg of alcohol per 100 ml of blood, that is 0.03%, which is lower than the 0.05% benchmark in the study. By the end of a long day, before any all-nighter has even started, a driver's test performance can be worse than what India would prosecute if it came from a drink. No breathalyser measures it.
| Hours awake | What the studies report | Main source |
|---|---|---|
| 16 | The critical wake period: beyond about 15.84 h, lapses start to accumulate | Van Dongen 2003 |
| 17-19 | Some test scores equal or worse than 0.05% blood alcohol | Williamson & Feyer 2000 |
| 24-48 | Large attention lapses; perceptual distortions, anxiety, irritability, depersonalisation, losing track of time | Lim & Dinges 2010; Waters 2018 |
| 48-90 | Complex hallucinations and disordered thinking | Waters 2018 |
| 72+ | Delusions; hallucinations reported in vision, touch and hearing by day 3 | Waters 2018 |
| ~120 (day 5) | A "turning point" towards a picture resembling acute psychosis or toxic delirium | Waters 2018 |
When do you start hallucinating without sleep?
Hallucinations usually start between 24 and 48 hours, and complex ones appear after 48. The most complete account is a 2018 systematic review in Frontiers in Psychiatry by Flavie Waters, Vivian Chiu, Amanda Atkinson and Jan Dirk Blom. They pooled 21 studies with about 760 participants kept awake for 24 hours up to 11 nights, mostly 72 to 92 hours. Vision was the most affected sense, in 90% of studies, followed by touch and body sensations (52%) and hearing (33%).
The slide, step by step
The sequence they describe is a gradual slide. It starts with misperceptions: edges that seem to move, a floor that seems to tilt, a coat on a hook that looks like a person. It moves to full hallucinations, then to delusions, beliefs that something false is true and resistant to argument. They found that a period of normal sleep resolved the psychotic symptoms in many cases, but not all.

A systematic review of 21 studies (about 760 healthy participants, 24 hours to 11 nights awake). Perceptual distortions, anxiety and depersonalisation appeared at 24-48 h, complex hallucinations at 48-90 h and delusions after 72 h, with day 5 marking a shift to a psychosis-like state. Limit: most studies are old, protocols and scales varied, and stress may have contributed; modern ethics rules prevent replication.
Read the paper (free)What happened to people who stayed awake much longer?
The best-known case is Randy Gardner, whose 264.4 hours awake open this post. His record matters less for what happened during the 11 days than for how it was read afterwards.
Gardner then slept 14 hours and 46 minutes and recovered quickly, which made the experiment famous as proof that sleep loss was harmless. That reading ignores what Ross recorded and the fact that a healthy teenager under constant supervision is a best case. In a 2017 interview, Gardner himself said he developed serious insomnia decades later and blamed the stunt. One case cannot prove cause, but it is not a reassurance either.

Could you physically stay awake for 72 hours if you wanted to?
Most healthy adults could, with help, but not cleanly: the brain inserts microsleeps, brief lapses of a few seconds, whatever you intend. The 72-hour mark is within the range of the studies in the Waters review and the 3-night arm of the Van Dongen experiment, where participants were kept awake under supervision. Unsupervised, the main danger is not the third night itself but what you do during it: drive, ride, cook with gas, climb stairs or make a decision you can't undo.
What caffeine can and cannot do
Caffeine delays but does not cancel the debt. It blocks adenosine receptors, and adenosine is one of the chemical signals of sleep pressure, but the PVT lapses that predict accidents keep rising. The US National Heart, Lung, and Blood Institute lists sleep deficiency as a factor in serious car crashes and in human errors behind industrial accidents, and it calls the belief that people can function well on little sleep a myth.
How long does it take to recover from 72 hours awake?
Recovery is quicker than the loss: one or two long nights restore most functions, but not always all of them. Total sleep lost over 72 hours is far less than 72 hours, because the body does not need to repay hour for hour. In the Van Dongen study, three nights of total deprivation meant cumulative sleep loss of only about 23.1 hours relative to each subject's normal sleep, even though wakefulness had been extended by 72 hours.
Gardner's first recovery sleep was under 15 hours. The Waters review notes that symptoms resolved after sleep in many but not all participants, and that the recovery dose needed is not well studied.
A healthy adult can stay awake for 72 hours under supervision. The cost is predictable: test performance worse than India's 0.03% drink-driving limit before the first night is over, hallucinations by the second day and delusions by the third, with a recovery of one to two long sleeps in most people. The deciding number is 15.84 hours: every waking hour past it adds to a debt that only sleep repays.
What's new
- MIT study links attention lapses after a sleepless night to waves of fluid leaving the brain MIT News, Oct 2025
- Going to bed 90 minutes later for six weeks led to weight gain and less activity in 95 adults Sci.News, Jul 2026
- A Delhi traffic court applies the 30 mg per 100 ml limit under Section 185; there is no equivalent test for fatigue Indian Kanoon, Oct 2025
For the memory side of sleep, see how sleep turns learning into memory; for the practical fix, the CBT-I guide to sleeping better. If exam season is pushing you towards all-nighters, read study techniques ranked by evidence first, and the working memory limit explains which of your mental capacities sleep loss shrinks first.
What if next?
- What if you meditated 10 hours a day for 10 days? The opposite experiment: ten days of deliberate attention, and why some people also see things that aren't there.
- What if a thought had weight? If tiredness feels heavy, physics can tell you what a thought actually weighs.
- What if you could stop time for everyone but yourself? The ultimate way to get more hours in the day, and why you'd suffocate in the dark.
Questions people ask
Is it safe to stay awake for 72 hours?
No. It is survivable for most healthy adults under supervision, but impairment begins within the first day and hallucinations are common by the second. It is especially risky for anyone with epilepsy, bipolar disorder or a history of psychosis, and for anyone who drives. See a doctor if you are going days without sleep.
Can you die from staying awake for 3 days?
Three days of sleep loss is not known to kill a healthy person directly. The danger is indirect: accidents from microsleeps and poor judgement, and triggering conditions such as seizures or mania in people who are vulnerable.
How much sleep do you need to recover after an all-nighter?
Usually one or two long nights. You do not repay hour for hour; in the Van Dongen study, three nights awake amounted to only about 23 hours of lost sleep, and recovery sleep tends to be deeper.
Why do you see things when you are very sleep deprived?
The visual system is the first to misfire: 90% of studies in the Waters review report visual changes. Brief sleep-like episodes intrude into waking, and the brain fills gaps with predictions, so shadows and patterns get misread as objects or people.
Sources
- Van Dongen H.P.A. et al.: The cumulative cost of additional wakefulness, Sleep 26(2) (2003) (source for the 15.84 h critical wake period, θ = 0.78 and the 23.1 h figure)
- Lim J., Dinges D.F.: A meta-analysis of the impact of short-term sleep deprivation on cognitive variables, Psychological Bulletin (2010) (source for the effect-size chart)
- Waters F. et al.: Severe sleep deprivation causes hallucinations and a gradual progression toward psychosis, Frontiers in Psychiatry (2018) (source for the symptom timeline and modality percentages)
- Williamson A.M., Feyer A.M.: Moderate sleep deprivation produces impairments equivalent to alcohol intoxication, Occup Environ Med (2000) (source for the 17-19 h and 0.05% equivalence)
- MIT News: This is your brain without sleep (Yang, Lewis et al., Nature Neuroscience) (2025)
- Wikipedia: Randy Gardner (record holder) (accessed 2026) (source for dates, 264.4 h, recovery sleep and Guinness)
- NHLBI: Sleep deprivation and deficiency (accessed 2026)
- Indian Kanoon: State vs Fateh Bahadur, quoting Section 185 of the Motor Vehicles Act (2025) (source for the 30 mg per 100 ml limit)
- Sci.News: Losing even 80 minutes of sleep a night linked to weight gain (Annals of Internal Medicine) (2026)
- Ministry of Health and Family Welfare: Tele MANAS
More from the world of the mind on the Manthan world page.
