Ultradian Rhythms: The 90 Minute Work Cycle, Explained

An ultradian rhythm is a biological cycle shorter than a day. The one that matters for work is the roughly 90 minute oscillation in arousal and alertness that runs through waking and sleeping hours alike. It is why a task can feel easy at 9:30 and impossible at 11:00 without the task changing.

Most of what you have read about this is either overclaimed or vague. This page separates the physiology from the productivity folklore, then gives you a way to find your own cycle in three days and plan around what you actually measure.

What an ultradian rhythm actually is

Nathaniel Kleitman, who co-discovered REM sleep, proposed the basic rest and activity cycle in the 1960s: the body does not move through the day at a constant level of arousal, it cycles through periods of higher and lower alertness roughly every 90 minutes, during sleep and during waking.

During sleep this is easy to demonstrate. Sleep architecture moves in cycles of about 90 minutes, from light sleep into deep sleep and back out toward REM. That part is not controversial.

During waking hours the picture is messier, and this is where most productivity writing stops being careful. Alertness does oscillate through the day, and researchers including Peretz Lavie documented roughly 90 minute cycles in vigilance and performance on sustained attention tasks. But these daytime cycles are smaller in amplitude than the sleep cycles they are named after, they vary a lot between people, and they are easy to override with caffeine, stress, or a deadline.

The honest summary: alertness is cyclical rather than flat, the cycle is roughly 90 minutes for many people, and the exact period is individual. Anyone who tells you yours is 90.0 minutes is selling something.

What is well supported, and what is not

Advice in this area usually mixes findings of very different strength. It helps to sort them.

Well supported

Sustained attention degrades with time on task. This is one of the oldest findings in experimental psychology, going back to the vigilance research of the 1940s and 1950s. Holding attention on one thing reliably gets harder the longer you hold it.
Brief breaks restore performance. Ariga and Lleras showed in 2011 that even very short diversions, well under a minute, were enough to prevent the vigilance decrement on a long, boring task. What mattered was interruption of the task, not the length of the rest.
Interruptions carry a cost even when you handle them well. Gloria Mark and colleagues found that people resume an interrupted task faster but with measurably more stress and frustration. The time saved is paid for somewhere.
Capacity is limited and shared. This is the standard view of attention since Kahneman’s 1973 work. Difficult work draws on a finite pool, and everything competing for it draws from the same pool.

Not well supported

That a break must be exactly 20 minutes. No study establishes a universal break length. The evidence is about interrupting the task, and about restoration of attention over time. Twenty minutes is a plausible default, not a finding.
That everyone runs on 90 minutes. Daytime cycles differ between people and even within one person across a week. The 90 minute figure comes from sleep and from group averages.
That working in 90 minute blocks is optimal. This is a reasonable extrapolation from the physiology, not a directly tested scheduling rule. Nobody has run the trial where half the office works in 90 minute blocks and the other half does not.

None of this makes the practice useless. It means the practice should be built on what you measure about yourself rather than on a number you read.

Find your own cycle in three days

Group averages are a starting guess. A three day measurement is better, and it is not much work.

Set a repeating reminder every 45 minutes during your working hours. Not every 90. You are sampling, not scheduling.
Rate two things from 1 to 5. How sharp is your attention right now, and how much resistance do you feel toward the task in front of you. Two numbers, ten seconds.
At the end of each day, line up the numbers against the clock. Look for the shape: when does attention rise, when does it fall, how long is the gap between a peak and the next peak.
Repeat for three working days, then stop measuring. You are looking for the period and the two or three genuinely sharp windows, not for a precise curve.

What tends to show up: a sharp window in the first two to three hours after you actually start, a dip after lunch that is partly food, partly accumulated attention cost, and a second window in the late afternoon that is smaller and shorter than the morning one for most people, though not for everyone. The night owls reading this will find the reverse, and they are not broken.

How to plan a day around it

Once you have the shape, the planning rules are short.

Put judgment work in the sharp windows

Anything where a bad decision costs more than the time saved: writing, design decisions, hard debugging, anything with an irreversible step. These go in the windows you measured. Not in the first slot of the day because it is culturally normal to put them there.

Put presence work in the flat windows

Meetings, reviews, pairing, admin, and anything where you mostly need to be awake and responsive rather than sharp. A flat window is a resource, not a failure. Scheduling a meeting into a peak window is the most expensive mistake in this whole model, because meetings survive being done at 70 percent and deep work does not.

Put recovery between blocks, not at the end

The point of the break is to interrupt the task, so it needs to actually interrupt it. Scrolling a feed does not, because it loads the same attention system you were just using. Standing up, walking, looking at something far away, or doing something physical does.

End the block when the quality drops, not when the clock says so

If you measured 70 minute cycles and you are still working at 85 because the block was supposed to be 90, you are producing work you will edit later. The signal to stop is a second or third attempt at the same sentence, or reading the same paragraph twice, not a notification.

What breaks the cycle

Back to back meetings. They do not respect your peaks, so the peak gets spent in a status update and the recovery never happens.
Notifications during deep work. Each one costs more than the seconds it takes to dismiss, because the cost is the reload of context, not the interruption.
Starting the day with intake. An hour of email and news before the first real task spends the sharpest window you will get on other people’s priorities, then the rest of the day is spent in the aftermath.
Caffeine after the early afternoon. It does not create a new cycle, it flattens the one you have and borrows against tomorrow’s.

Common mistakes

Treating the number as a rule. If your measured cycle is 70 minutes, a 90 minute block is not aspirational, it is just wrong for you. Use your number.

Using the rhythm to skip breaks. The cycle is an argument for taking recovery seriously, not an argument for working harder inside the peak. Two sharp 70 minute blocks with real recovery between them beat four blurred hours.

Planning the cycle instead of the work. The rhythm tells you when to do judgment work. It does not tell you what the judgment work is, and no amount of energy optimisation substitutes for knowing your one priority. That is a separate decision, covered in how to plan your day.

Where a planner helps and where it cannot

The mechanical part of this is genuinely automatable: knowing your sharp windows, knowing which of today’s tasks need judgment, and rebuilding the day when a meeting lands inside a peak. Nityasha is built around exactly that, with your energy pattern kept in a profile rather than re-entered each morning. The day planner page shows the behaviour and AI scheduling assistant covers the rescheduling.

The part it cannot do is stop the meeting being scheduled at 9:30. A planner can move your deep work out of the way. It cannot defend the window against a colleague who does not know it exists, and pretending otherwise would be the same overclaim this page just spent three sections dismantling.

Frequently asked questions

Is the 90 minute rule real?

The underlying rhythm is real and was described in sleep research first. The specific claim that you should work in 90 minute blocks is an extrapolation from that physiology to work scheduling, and it has not been tested as a scheduling rule. The parts that are well evidenced are that sustained attention degrades over time and that interrupting the task restores it.

How long should a break be?

There is no established optimal length. The evidence points at interruption of the task mattering more than duration, and at genuine restoration mattering more than a fixed number. Five minutes of walking is likely to do more for the next block than twenty minutes of scrolling.

Does this work if my day is all meetings?

Partly. You cannot place deep work in a window that is already booked, but you can stop spending peak windows on meetings you could move. If you have no window longer than about 45 minutes, the constraint is the meeting load and the useful move is a calendar conversation, not a better rhythm.

Can I train myself onto a different cycle?

You can shift your sleep timing and, more slowly, your chronotype, and light exposure in the morning is the main lever. You cannot talk yourself into a 90 minute daytime cycle if you measure 70. The measurement is cheaper than the argument.

What about the afternoon dip?

It is real and it is not only lunch. Alertness drops in the early afternoon for most people regardless of what they ate, and it is a flat window rather than a broken one. Schedule presence work into it and stop expecting judgment work from yourself at 2pm.

Sources

Kleitman, N. Sleep and Wakefulness. University of Chicago Press, 1963.
Ariga, A. and Lleras, A. “Brief and rare mental breaks keep you focused: Deactivation and reactivation of task goals preempt vigilance decrements.” Cognition, 2011.
Mark, G., Gudith, D. and Klocke, U. “The cost of interrupted work: more speed and stress.” CHI, 2008.
Kahneman, D. Attention and Effort. Prentice Hall, 1973.

The short version: your capacity is cyclical, the cycle is roughly yours to measure, judgment work belongs in the peaks, presence work belongs in the troughs, and the break has to actually interrupt the task. Everything else in this category is a claim someone made without a control group.