How long is a sleep cycle across the night

When people ask how long is a sleep cycle, the standard answer is usually 90 minutes. A sleep cycle in the classical sense is an episode of non-REM sleep followed by an episode of REM sleep, and its start and end are taken from the hypnogram, the stage-by-stage chart of an overnight recording ^[2]. The commonly cited description of a night's sleep, which a 2025 paper quotes as what is known about classical sleep cycles, is 4 to 6 cycles lasting about 90 minutes each; the paper cites Feinberg and Floyd (1979) and Le Bon (2020) for it ^[2]. However, overnight recordings reveal that this duration is not a fixed constant.

What laboratory datasets measured for cycle duration

A retrospective analysis of 369 participants from laboratory studies run in Basel between 1994 and 2020 recorded 6,064 cycles by polysomnography, and the median cycle lasted 96 minutes ^[1]. Among healthy participants given an 8-hour sleep opportunity at their habitual bedtime, cycle duration varied widely from one person to another ^[1]. In the same dataset the number and the duration of cycles were not normally distributed, so a single average does not describe the spread well ^[1]. Later in the night the distribution of NREM sleep became narrower and that of REM sleep wider ^[1].

Other cohorts show similar spreads around that number. Across five datasets totalling 205 healthy adults, 961 classical cycles averaged 90 minutes with a standard deviation of 25 minutes, and participants had on average 4.7 cycles per night (standard deviation 0.9) ^[2]. Cycle durations in that sample were not normally distributed either ^[2].

Cycle structure and variations across the night

The first cycle of the night was consistently shorter than the cycles that followed it in the 369-participant laboratory dataset ^[1]. In a separate 2025 analysis of 205 healthy adults, cycle durations rose and then fell again across the night, an inverted U shape ^[2]. In 47 of the 205 healthy adults (23 percent) the first cycle contained no REM sleep at all, which the authors call a skipped cycle ^[2].

CIBDOL · Cycle structure and variations across the night

Cycle length also depends on how cycles are identified mathematically. When cycles were instead defined from the aperiodic, so-called fractal, component of the EEG, they averaged 89 minutes with a standard deviation of 34 minutes ^[2]. 81 percent of these fractal cycles could be matched to a classical cycle, and in 54 percent of participants all fractal cycles approximately coincided with classical ones, so two reasonable definitions do not always count the same cycles ^[2].

Factors influencing cycle components

Individual biology alters the internal timing of sleep stages. In the Basel dataset age and sex affected the length of the parts of a cycle: older participants had longer NREM and shorter REM episodes, particularly towards the end of the night, and women had longer NREM episodes ^[1]. After sleep deprivation the first cycle held longer NREM sleep, and when sleep pressure was low, after multiple naps, REM episodes were longer ^[1].

Environmental disruptions tested in the same protocol did not show the same shifts. In those laboratory studies, moderate light exposure before sleep and noise during the night had no significant effect on cycle duration ^[1]. The authors conclude that cycles are endogenous rhythms modulated by age, sex and sleep pressure but not directly responsive to moderate environmental cues in healthy good sleepers; the finding is limited to moderate levels, laboratory conditions and that population ^[1].

Sleep inertia and consumer tracking devices

Waking up at different points in a cycle influences how alert a person feels immediately afterwards. A review of sleep inertia, the transitional state of lowered arousal and reduced performance right after waking, reports that an abrupt awakening from slow wave sleep produces more sleep inertia than an awakening from stage 1 or 2, with REM sleep intermediate ^[3]. In the absence of major sleep deprivation it rarely lasts longer than 30 minutes, although individual studies report durations from 1 minute to 4 hours ^[3].

Many consumers use mobile devices to track these phases, yet consumer hardware performs differently from clinical instruments. When one smartphone sleep app was used by 20 volunteers during laboratory polysomnography, its epoch-by-epoch agreement with the recording was 45.9 percent because it separated the stages poorly, although it told sleep from wake with 85.9 percent accuracy ^[4]. The authors conclude that the app's sleep staging correlated poorly with polysomnography ^[4].