Ultradian rhythm is a scientific term that now gets used to sell a very specific work schedule: focus for 90 minutes, then rest for 20. The first part is biology. The second is a hypothesis often repeated as a rule.
Short answer: An ultradian rhythm is a biological cycle that repeats multiple times within 24 hours, such as the NREM–REM sequence during sleep or pulsatile hormone release. Human sleep cycles have a median length of about 96 minutes (Cajochen et al. (2024)), but that doesn't establish a matching work cycle.
What is an ultradian rhythm?
An ultradian rhythm is a recurring biological pattern with a period shorter than a day, though the mechanisms and periods vary widely across biological processes (van der Veen and Gerkema (2024)). Sleep-stage cycling and pulsatile hormone release are two ultradian rhythm examples; neither means everyone's attention rises and falls on one shared timer.
| Example | What repeats | What the evidence says |
|---|---|---|
| Sleep | NREM and REM episodes alternate through the night. | In a lab-data review of 369 healthy participants, the median cycle was 96 minutes; the 95% interval was 60–150 minutes. Sleep cycle length varied between people and across a night. Cajochen et al. (2024) |
| Hormones | Growth hormone (GH) is released in pulses rather than at a steady level. | Endotext reports an average of 10 GH pulses per day, with 128 minutes between pulses; the major peak is late at night or early in the morning. That's a hormone pattern, not a work-rest prescription. Lim and Khoo (2025) |
The often-quoted 90-minute sleep cycle isn't a stopwatch setting, either: lengths varied across the night, and the first cycle was shorter than later ones (Cajochen et al. (2024)).
Ultradian rhythm vs. circadian rhythm: what's the difference?
A circadian rhythm runs on a roughly 24-hour pattern; an ultradian rhythm repeats within that day. Both can act on sleep at once: NREM–REM cycles repeat through the night, and the circadian pacemaker may also promote REM sleep toward the end of the night (Cajochen et al. (2024)).
| Rhythm | Period | Human example | Practical reading |
|---|---|---|---|
| Circadian | About 24 hours | Daily sleep-wake timing | Your sleepiness changes across the day. |
| Ultradian | Shorter than 24 hours | NREM–REM sleep cycling; hormone pulses | Some biological processes fluctuate more than once a day; the interval is specific to the process. |
The word names a timescale, not one universal biological clock. A sleep cycle, a hormone pulse, and a work session are different measurements.
Should you follow the 90 minute rule at work?
You can try a 90-minute block as a planning choice, but research doesn't establish it as the biologically correct work interval for everyone. The often-cited waking Basic Rest–Activity Cycle (BRAC) is a hypothesis; studies don't show that ordinary workdays reliably follow the same 90-minute cycle seen in sleep.
| Claim | Evidence | What it means for your workday |
|---|---|---|
| Sleep cycles average around 90 minutes. | Established sleep-stage cycling, with a median of 96 minutes and wide variation in a 2024 lab-data study. Cajochen et al. (2024) | A sleep pattern is real; it isn't evidence that you should set a 90-minute work timer. |
| Waking alertness follows the same 90-minute cycle. | Mixed evidence. A 2008 review found some approximate 90-minute EEG rhythms in sleep-deprivation experiments, but no confirmation of one BRAC connecting the sleep cycle to waking alertness and performance. Lajambe and Brown (2008) | Findings under sleep loss don't establish a schedule for a normal workday. |
| Cognitive performance reliably repeats every 90 minutes. | In a study of 60 people tested every 10 minutes for 9 hours, conservative spectral analysis found no significant 90-minute periodicity in task performance, alertness, mood, or heart period. Neubauer and Freudenthaler (1995) | The tested measures didn't support a fixed 90-minute work cycle. |
| Expert practice proves that 90 minutes is optimal. | Ericsson and colleagues recorded deliberate-practice sessions among violinists and pianists. Uninterrupted violin practice was limited to 1–1.5 hours; expert pianists' sessions averaged 87.9 minutes, likely an overestimate because diary entries missed breaks shorter than 15 minutes. Ericsson et al. (1993) | These observations concern selected musicians and deliberate practice. They didn't test a universal office schedule or prove a 90/20 rule. |
Kleitman's BRAC proposed that the sleep cycle's approximate 1.5-hour period might continue into waking activity (Neubauer and Freudenthaler (1995); Lajambe and Brown (2008)). That's different from proving a regular peak, a predictable dip, and a required 20-minute recovery break for every person. "Work 90, rest 20" is a practical extrapolation, not a result established by the studies above.
How can you find your own work interval?
Track when you naturally stop a demanding task for two work weeks, then look for a repeatable pattern. It's a personal observation, not a clinical measure or proof that your body runs on a biological cycle.
- For 10 workdays, work as usual. Don't set a 90-minute alarm. At the start of a demanding task, note the time and task. When you leave it for at least five minutes, record the end time and why: natural pause, task finished, or outside interruption.
- Keep interruptions separate. Mark meetings, messages, and other imposed stops so they don't masquerade as your natural limit. If you switch tasks without a real pause, keep the original session open only if you return to it within five minutes; otherwise, start a new entry when you resume.
- Add one simple rating. At the start and end, rate how focused you feel from 1 to 5. It's your own rating, not a validated fatigue test. Use the same scale each time.
- Compare the sessions. In a spreadsheet, calculate each eligible session's duration, then find the median. Compare morning with afternoon and one task type with another; don't let one unusually long day define your schedule.
- Try the pattern for a week. If similar tasks repeatedly end near the same interval, use that as a flexible calendar block and take a break when you reach a natural stopping point. If your intervals vary or imposed interruptions dominate, don't force a single number onto the day.
A simple row is enough: date | task | start | end | minutes | natural / finished / interrupted | focus before | focus after. For example, a 10:05–11:18 session is 73 minutes. One 73-minute session is a data point; a repeated pattern across comparable tasks is a reason to try that interval, not evidence of a universal rhythm.
The honest limits
A two-week log can show when you tend to stop; it can't tell you that a hormone pulse caused it. Workload, sleep, meetings, task difficulty, and interruptions can all shape the pattern. If your records don't repeat, the honest result is that this log found no dependable interval.
Use 90 minutes only if it fits what you observe or helps you plan. If your natural sessions are shorter, set a shorter block; if a useful task is flowing, finish at a sensible stopping point. I wrote more about that flow state and when not to interrupt it.
I make Timex. Its Energy estimate nudges after 90 demanding minutes without a 10-minute break, but that interval is a labeled assumption, not a validated biological cycle; the 1.5.0 release note explains the model. The break timer's default is five minutes every 30 minutes, with other intervals available. Those are scheduling choices, not a claim that your ultradian rhythm lasts 30 or 90 minutes. If a repeating reminder helps, here's how to choose a light-touch setup.
If you want a starting point, try 90 minutes for a week and compare it with your natural stopping times. Keep the interval only if it helps your work. Download Timex.
— DK