Is 17 Hours Awake Really Like Being Legally Drunk?

Williamson and Feyer (2000) studied 39 subjects, 30 employees from the transport industry and nine from the army. The same people were tested across 28 hours of sleep deprivation and, separately, after measured doses of alcohol up to about 0.1% blood alcohol concentration. Each person therefore acted as their own comparison, which removes the usual problem of comparing naturally alert people with naturally tired ones.

After 17 to 19 hours without sleep, corresponding to clock times of 2230 and 0100, performance on some tests was equivalent to or worse than performance at a BAC of 0.05%. Response speeds were up to 50% slower on some tests, and accuracy measures were significantly poorer than at that alcohol level. After longer periods without sleep, performance reached levels equivalent to the maximum alcohol dose the subjects were given, a BAC of 0.1%.

The phrase carrying the most weight in that sentence is "some tests". The equivalence was not a blanket statement about every measure in the battery, and the widely repeated version of this finding, that a worker 17 hours awake performs like a drunk driver, is a harder claim than the authors made.

Lamond and Dawson (1999) ran the same comparison differently. Twenty-two healthy subjects aged 19 to 26 completed three counterbalanced conditions. In the sustained wakefulness condition they were kept awake for 28 hours. In the alcohol and placebo conditions they drank either an alcoholic or a non-alcoholic beverage at 30 minute intervals until blood alcohol concentration reached 0.10%. Performance was measured hourly on four tasks from a standardized computer-based battery, and the placebo beverage did not significantly affect mean relative performance.

Their result was a range, not a point: depending on the task measured, approximately 20 to 25 hours of wakefulness produced performance decrements equivalent to those observed at a BAC of 0.10%. The dose-response patterns also differed by exposure. As BAC increased, performance on all tasks except one significantly decreased. As hours of wakefulness increased, four of six parameters significantly decreased. Alcohol degraded almost everything measured; wakefulness did not.

Put the two papers side by side and there is no single hour at which sleep loss becomes alcohol. Williamson and Feyer place a 0.05% equivalence in a 17 to 19 hour window on some tests. Lamond and Dawson place a 0.10% equivalence somewhere between 20 and 25 hours depending on the task. Any single-number version of this comparison is tidier than either study supports. The honest summary is a band that moves with the task you choose to measure.

Both experiments measured response speed and accuracy on laboratory test batteries. Neither measured real driving, real crashes, or real work output. The finding is that test performance degrades to a comparable degree, not that a fatigued worker is as likely to crash as an intoxicated one, and not that fatigue and alcohol impair through the same underlying mechanism. Treating the equivalence as a risk equivalence is the most common misreading of this literature.

The alcohol benchmark is also jurisdictional. A BAC of 0.05% is the legal driving limit in Australia, where both studies were run, and in much of Europe, but it sits below the general limit in the United States. Saying "legally drunk" without naming the country turns a specific number into a slogan. The samples are small and narrow as well: an occupational sample of transport and army personnel in one study, and 22 young healthy adults in the other.

Provenance matters here too. The comparison originates with Dawson and Reid (1997) in Nature, whose bibliographic record is confirmed but whose full text is not openly accessible, so every figure on this page comes from the two later papers instead. Note also that Lamond and Dawson (1999) shares an investigator with that originating study, so it is a continuation of one research program rather than an independent replication. Williamson and Feyer (2000) is the independent team. The evidence base runs from 1997 to 2000, is heavily cited and long established, and reports no confidence intervals or effect sizes that can be recovered from the accessible records.

The variable in these experiments is time since waking, not time on shift. A worker who woke at 0600 and finishes an evening shift at 0100 has crossed the tested window without working an unusual number of hours. Most rosters record shift length and rest between shifts; very few record how long someone has actually been awake, which is the quantity both studies manipulated.

Time awake and time of day are also entangled in this design. The 17 to 19 hour marker corresponds to 2230 and 0100, which sits in the circadian low, so these experiments cannot separate the effect of hours awake from the effect of the clock. That limitation does not weaken the practical implication: late finishes, call-outs, shift extensions, and long commutes all add to fatigue exposure whether or not the shift itself looks long on paper.

Use the window as a caution band, not as a certification. Nothing in either study establishes that 16 hours awake is safe or that a specific hour is a threshold. What the evidence supports is treating extended wakefulness as a measurable performance exposure that deserves the same planning attention as shift length and rest intervals.