Do 90-Minute Work Cycles Exist? Ultradian Rhythms and the Evidence

90-Minute Work

The claim is everywhere in productivity writing: the brain runs on 90-minute cycles during the day, so you should work in 90-minute blocks and rest when the cycle dips. It traces back to a real scientist, the sleep researcher Nathaniel Kleitman, and to a real finding, the roughly 90-minute cycle of sleep stages at night. But the step from night to day is where the evidence thins out. Daytime studies have found rhythms in alertness and performance, yet their periods range from about 80 minutes to several hours, they differ between tasks and people, and the link to the sleep cycle has never been confirmed. Ninety minutes is a reasonable size for a focused work block. It is not a biological clock you can schedule around.

What Kleitman Actually Proposed

Kleitman, who spent his career at the University of Chicago, is best known for the 1953 discovery, with his student Eugene Aserinsky, of rapid eye movement (REM) sleep. Later work in his lab showed that sleep is not one uniform state but a series of cycles: non-REM sleep followed by REM sleep, repeating roughly every 90 minutes through the night.

Kleitman then made a bolder suggestion. He proposed that the sleep cycle is only the night-time expression of a deeper rhythm that runs around the clock, which he called the basic rest-activity cycle (BRAC). In the 1963 edition of his book Sleep and Wakefulness and in a 1982 paper revisiting the idea, titled “Basic rest-activity cycle: 22 years later,” he described it as an ultradian rhythm, meaning one that repeats more than once a day, with a period of about 80 to 120 minutes in adults. Citing earlier observations of infants, he noted that the cycle appears shorter early in life, at around 50 minutes, and concluded that it lengthens with age.

Two details matter for how the idea is used today. First, the BRAC was a hypothesis about a general physiological rhythm in arousal, not a recommendation about work. Second, its strongest support came from sleep, where the cycle can be measured night after night with polysomnography. The daytime half of the hypothesis was always the less certain one.

The Night Cycle Is Solid; the Day Cycle Is the Leap

Question During sleep During waking hours
Does a repeating cycle exist? Yes: non-REM and REM stages alternate all night Some studies find rhythms in alertness, EEG and performance
How long is it? About 90 minutes on average, varying between cycles and people Reported periods from about 80 minutes to 4 hours
Is it easy to measure? Yes: sleep stages are scored from EEG all night No: needs many hours of frequent testing under controlled conditions
Is it the same mechanism? The defining case Not established; reviews describe the link as unconfirmed

Evidence against a single continuous clock appeared early. In 1977, Jerry Moses and colleagues reported in Nature that the REM cycle behaves as a sleep-dependent rhythm: in experiments with altered sleep and wake schedules, it did not carry on like an independent clock ticking through wakefulness. That does not rule out rhythms during the day. It suggests that if they exist, they are not simply the night-time sleep cycle continuing into the morning.

What Daytime Studies Actually Found

Researchers have looked for ultradian rhythms in waking people since the 1970s, mostly in small laboratory studies where participants repeat tests every few minutes for a whole day. The results are real but untidy.

Study Design What they found
Hayashi, Sato and Hori, 1994 (Perceptual and Motor Skills) 10 male students tested every 15 minutes for 9 hours: EEG, mood, task performance, self-ratings Behavior and self-ratings shared a cycle of about 2 hours; EEG showed a slower 4-hour component and faster components between about 80 and 144 minutes
Gordon, Stoffer and Lee, 1995 (International Journal of Neuroscience) 24 men performing verbal, spatial and perceptual tasks over 8 hours, with blood samples Verbal performance cycled at about 80 minutes and spatial at about 96 minutes; each rhythm vanished in one of the two seasonal groups
Lajambe and Brown, 2008 (review chapter, Ultradian Rhythms from Molecules to Mind) Review of studies using sleep loss to expose waking rhythms Some EEG studies show an arousal rhythm near 90 minutes, and others a slower rhythm of about 4 hours; the link to the sleep cycle through the BRAC has not been confirmed

Read together, these studies support a narrower claim than the popular one. Alertness and performance do fluctuate within the day on time scales of one to a few hours. But the period is not fixed at 90 minutes, different measures in the same person can cycle at different speeds, and some rhythms appear in one group of participants and not in another. Hayashi’s team interpreted their data as evidence for several independent oscillators rather than one master cycle. Lajambe and Brown also reported that the 90-minute and 4-hour rhythms were most visible in the morning and early afternoon, and seemed to be damped later in the day as the circadian rise in alertness took over.

Why the Numbers Never Settle

Part of the uncertainty is mathematical. Detecting a cycle reliably requires observing several repetitions of it, and laboratory days are short.

How many 90-minute cycles fit in a laboratory day
Cycles observed = recording length ÷ period
8-hour session: 480 ÷ 90 ≈ 5.3 cycles
9-hour session: 540 ÷ 90 = 6 cycles
Testing every 15 minutes: 6 data points per 90-minute cycle
Testing every 15 minutes over 9 hours: 37 data points in total

With only five or six cycles and a few dozen data points per person, a spectral analysis cannot pin a period down precisely. That is why results are reported as bands, such as 80 to 144 minutes, rather than as a single number, and why two studies of the same measure can land on different values. Repeated testing adds a general difficulty for this kind of research: doing the same cognitive task every 15 minutes produces practice, boredom and fatigue effects with their own time course, which have to be separated from any rhythm. These limits do not make the studies wrong. They explain why the honest summary is “rhythms of roughly one to two hours, and slower ones of about four,” not “a 90-minute cycle.”

How 90 Minutes Became a Productivity Rule

The modern form of the claim was popularized in management writing rather than in chronobiology. In a widely shared 2013 New York Times essay, the consultant Tony Schwartz described writing in three uninterrupted sessions a day of no more than 90 minutes, each followed by a break, and pointed to research by the psychologist K. Anders Ericsson showing that elite performers work in sessions of that length and rarely more than about 4.5 hours a day.

Ericsson’s best-known study on the subject, published with Ralf Krampe and Clemens Tesch-Römer in Psychological Review in 1993, is about how much deliberate practice experts accumulate and how much effortful practice per day is sustainable; it cites training studies that found little benefit from more than about four hours a day. It is not a study of rhythms and does not identify a 90-minute cycle. The 90-minute session length in popular accounts is therefore a blend: Kleitman’s sleep cycle, extended to daytime by hypothesis, combined with practice habits from expertise research, and presented as a single biological law.

Even If the Rhythm Exists, the Clock Doesn’t Know Your Phase

The scheduling advice has a second problem that would remain even if every person had a stable 90-minute cycle. A schedule of 90-minute blocks starting at 9:00 a.m. assumes everyone’s cycle starts at 9:00 a.m. Nothing in the research suggests that. The daytime studies report the length of fluctuations, not a shared starting time, and the phase would depend on when each person woke up, what they had been doing and how their circadian rhythm was positioned that day.

That makes a fixed block schedule a coin toss with respect to any underlying rhythm. A 90-minute block could begin at a peak, at a trough or anywhere between. This is our inference from how the studies were designed, not a tested result, but it explains why following the clock is less useful than noticing your own attention. Time since waking is a better-supported predictor of when focus is strongest, as covered in the best time of day for deep work.

What the Evidence Does Support About Work Blocks

Abandoning the 90-minute biology does not mean abandoning breaks. There is better evidence for why breaks help than for when they should come.

Finding Study What it shows
Attention declines during a long, unbroken task Ariga and Lleras, 2011 (n=84, 50-minute task) Two brief switches away from the task prevented the usual drop in vigilance
Short breaks restore energy Albulescu et al., 2022 (meta-analysis, 22 samples, 2,335 participants) Micro-breaks of up to 10 minutes raised energy and lowered fatigue; the effect on performance was small and not significant; longer breaks helped more on demanding tasks
Fixed break schedules: mixed results Biwer et al., 2023 (n=87); a 2025 study (n=94) In 2023, fixed breaks led to less fatigue and distraction than self-chosen ones; in 2025, 25-minute blocks with 5-minute breaks produced a faster rise in fatigue than flexible breaks; neither found a difference in overall output
Productive people take regular breaks DeskTime software data, 2014 The most productive users averaged 52 minutes of work and 17 minutes of rest; later reruns gave different numbers, and the data are observational

None of these studies identifies 90 minutes as special. They point instead to two robust effects: sustained attention to one task erodes with time on task, and short pauses help people recover energy and stay engaged. Which fixed interval works best, if any, remains unsettled, which is why lengths as different as 25, 52 and 90 minutes all have devoted users. How the 25-minute version came about, and what controlled trials show about it, is covered in the Pomodoro technique.

Using 90 Minutes Without the Myth

A 90-minute block can still be a useful unit of deep work, as a planning convention rather than a biological rhythm. Treated that way, it can be adjusted freely: shortened when attention fades early, extended when work is flowing, and placed where your own alertness tends to be highest rather than where the clock says a cycle should start. That flexibility is a judgment about how to apply the research, not a finding of it.

The most direct way to find your own interval is to measure it. For a week or two, start a timer at the beginning of each focused block, note when your attention first noticeably drifts and how long you keep going after that. If the drift points cluster around a particular length, that is a better guide than any published number, including 90 minutes. If they scatter widely, that is consistent with what the daytime studies found: rhythms in human alertness exist, but they are looser, more varied and more individual than a fixed cycle would suggest.

 


Sources and references

Hayashi, M., Sato, K. and Hori, T. (1994): Ultradian Rhythms in Task Performance, Self-Evaluation, and EEG Activity, Perceptual and Motor Skills
Laboratory study measuring EEG, mood and performance every 15 minutes for 9 hours
Lajambe, C. M. and Brown, F. M. (2008): Ultradian Cognitive Performance Rhythms During Sleep Deprivation
Review of waking ultradian rhythms in EEG and performance, and of the unconfirmed link to the sleep cycle
Moses, J., Lubin, A., Johnson, L. C. and Naitoh, P. (1977): Rapid Eye Movement Cycle Is a Sleep-Dependent Rhythm, Nature
Experimental evidence that the REM cycle does not continue as an independent clock during wakefulness
↑
✕Close Menu