Segmented sleep

Jump to: navigation, search

This text is part of: "Science of sleep" by Piotr Wozniak (2017)

Wehr experiment

In 1992, Dr Thomas Wehr published the results of his interesting experiment on sleep in periods of prolonged darkness (e.g. as seen in Inuit during the arctic winter)(Wehr 1992[1])). He divided the experimental day into 10 hours of daylight (photoperiod) and 14 hours of darkness (scotoperiod). This type of artificially controlled environment resulted in segmented sleep that was often composed of two 4-5 hour segments separated by an hour of wakefulness. Wehr found that the onset of sleep was associated with an increase in melatonin, which is released in the periods of darkness (in both diurnal and nocturnal animals). He also noticed that the release period of nocturnal melatonin lasts longer in shorter photoperiods (Wehr et al. 1993[2]).

Fragment of a segmented sleep log produced in conditions of 10 hour photoperiod.

Fragment of a segmented sleep log produced in conditions of 10 hour photoperiod.

Interpretation of segmented sleep

When the results of the experiment were publicized, insomniacs rejoiced: Perhaps this is normal? Perhaps this is how we all should sleep? Those who tend to wake up in the night and spend an hour or so reading, watching TV or plundering the fridge, no longer had to feel abnormal. Indeed, the best criterion that should separate healthy sleep form unhealthy patterns is the refreshing power of sleep. Nocturnal awakenings should not matter as long as they did not contribute to morning misery.

A historian, Dr Roger Ekirch noticed that this prolonged two-part sleep is frequently mentioned in historical records that predate the advent of electricity: "Until the close of the early modern era, Western Europeans on most evenings experienced two major intervals of sleep bridged by up to an hour or more of quiet wakefulness. [...] The initial interval of slumber was usually referred to as "first sleep," or, less often, "first nap" or "dead sleep." (Ekirch 2001[3]).

Sleep researchers speculated that this is perhaps how healthy sleep should look like and that our sleep control models with a single nighttime circadian peak are wrong. Evolutionists speculated that this could be an adaptation to nighttime sex, or breastfeeding, or periods of extra vigilance. Dr Horne likes to refer to segmented sleep as an example of the human propensity to excessive sleeping.

Segmented sleep and Borbely model

I happen to disagree with most of the interpretations put forward thus far except those that stand in agreement with the mainstream sleep research. We need to observe that most of human and pre-human evolution took place in tropical areas with far shorter nights than those that characterize winters in the north, and, mathematically speaking, there should be no preference for waking up in the middle of the night for 1-2 hour as opposed to entering shallow sleep or waking at the end of each full NREM-REM cycle. The segmented sleep observed in Wehr's experiment can easily be accounted for with Borbély model of sleep. Even though Borbely model provides very specific mathematical conditions needed for initiating sleep, we must remember that it is only an approximation of reality that does not necessarily account for the level of lighting or external arousing stimuli. In periods of prolonged darkness and silence, lesser overall sleep propensity will be needed to initiate sleep. I will try to illustrate my claim using SleepChart's two-component model and some real life examples.

Segmented sleep and two-component model

When Wehr's data are processed using SleepChart's two-component model, we see that sleep is characteristically initiated at periods of relatively low sleep propensity:

Wehr's segmented sleep as interpreted with the help of the two-component model of sleep employed in SleepChart.

Wehr's segmented sleep as interpreted with the help of the two-component model of sleep employed in SleepChart.

In real life, the two-component model indicates that sleep is initiated when alertness levels drop to 33-40%. In segmented sleep, alertness at sleep onset is much higher: 40-50%.

Circadian graph shows that the favorite sleep initiation hour is the 15th from arising and it results in 5 hours of sleep on average. As waking comes close to the circadian acrophase, wakefulness cannot last long due to a rapidly ascending circadian sleepiness. The second bout of sleep then follows in the 21st hour and lasts 3 hours on average. Thus the sleep is segmented into a 5 hour long pre-sleep and 3 hour long "correction".

Circadian graph based on Wehr's data.

Circadian graph based on Wehr's data.

Sleep periodogram shows a typical frequency peak at period 24h, and two atypical peaks at 8 and 6 hours (instead of the usual 12h siesta peak).

Periodogram based on Wehr's data.

Periodogram based on Wehr's data.

Examples of segmented sleep

I scanned years of sleep logs in search for natural segmented sleep examples. I did not found that many. Without doubt, I can blame the modern lifestyle that rarely permits a leisurely early bedtime. Below I list three very different examples from real-life logs. Several characteristics seem to be associated with segmented sleep:

  • this sleep often results in high alertness, e.g. as testified by the absence of daytime napping
  • as predicted, this sleep is easier to find in winter periods
  • intense exercise may be a factor that helps induce segmented sleep (perhaps due to earlier bedtimes)
  • as in Wehr's data, SleepChart's two-component model of sleep propensity shows a very rapid decline in sleep propensity early in segmented sleep due to the fact that the sleep is initiated very early in reference to the circadian acrophase

Example: Premature bedtime

Typical long darkness premature bedtime segmented sleep. Sleep initiated too early, again with a very marked decline in sleep propensity resulting in a nocturnal awakening:

Typical long darkness premature bedtime segmented sleep.

Example: Nocturnal awakening caused by stress

Premature awakening caused by stress. Segmented night with a "correction":

Premature awakening caused by stress.

Example: Intense exercise

Prolonged sleep induced by intense exercise with increased demand for sleep. Over 12 hours of segmented sleep are initiated early with multiple harmless awakenings, fast decline in sleep propensity (inverse of the red line) in the first 3-4 hours of sleep that results in shallow prolonged sleep:

Prolonged sleep induced by intense exercise with increased demand for sleep.

Application of segmented sleep

This type of sleep results in very refreshing nights, however, it would be pretty hard to implement in agreement with the modern lifestyle. Certainly, it would not optimize the time use. I can only guess that matching sleep well with the circadian acrophase should also increase the efficiency of sleep. This should be verifiable with SuperMemo data, however, as of the moment of writing, I have not done the necessary computation. Considering the nice effect of this ancient sleep on mood and alertness, I would love to subject myself to an experiment with 14 hours of darkness, however, 10 hour working day would be pretty hard to stomach even in a short term. I also doubt I would be able to extinguish the thoughts of the day and initiate sleep early. Even the mere fact of collecting exciting data for the experiment would keep me up with my thoughts racing. I might try this in retirement when my vital powers decline sufficiently enough to make it truly enjoyable.


  1. Oren D.A., M.D., Wehr T.A., M.D., "Hypernyctohemeral Syndrome after Chronotherapy for Delayed Sleep Phase Syndrome," The New England Journal of Medicine / Volume 327 / Issue 24 (December 10, 1992): 1762
  2. T. A. Wehr, D. E. Moul, G. Barbato, H. A. Giesen, J. A. Seidel, C. Barker, and C. Bender, "Conservation of photoperiod-responsive mechanisms in humans," American Journal of Physiology - Regulatory, Integrative and Comparative Physiology / Volume 265 / Issue 4 (October 1993): 846-857
  3. Ekirch, A.R., "Sleep We Have Lost: Pre-industrial Slumber in the British Isles," The American Historical Review / Volume 106 / Issue 2 (2001): 343-387