Complete darkness in the bedroom

What lies behind it

Light information for the internal clock does not travel via rods and cones but via dedicated light-sensitive ganglion cells of the retina, which project directly into the suprachiasmatic nucleus. They are tuned to short wavelengths: the action spectrum of melatonin suppression in humans identifies 446–477 nm as the most effective range.

What matters in practice is the shape of the dose-response curve. It is logistic, not linear. In a study with 23 healthy young men and women, dim room light of around 100 lux — slightly more than 1 % of the amount of light of bright evening light at around 9,000 lux — already produced half of the maximal phase delay. The steepest part of the curve lies in the range of ordinary room light.

At what brightness something happens

The laboratory studies define darkness as less than 3 lux, and that is the reference condition for all effects. The lowest documented threshold for sleep structure itself is 5 lux over the whole night: more wake time after sleep onset, more stage N1, less stage N2, more REM sleep. No difference was measurable between 5 and 10 lux.

One step higher, effects become measurable beyond sleep as well: a single night at 100 lux caused HOMA-IR to rise by around 15 %, while it fell by around 4 % under dim light. The glucose values themselves remained unchanged — the effect showed up in insulin demand, not in the sugar level.

What was measured in the population

In the prospective HEIJO-KYO cohort, bedroom light was measured objectively in 863 people without depressive symptoms, mean age 71.5 years. Over a median follow-up of 24 months, 73 new cases occurred. The group with an average of 5 lux or more had a hazard ratio of 1.89 compared with the dark group, and 1.72 in the multivariable model.

These are observational data in a specific group: older people in Nara, Japan. A second cohort finding shows in people with bipolar disorder that nighttime bedroom light predicts manic relapses. Both are associations, not causality.

What the studies show

Room light before bedtime shortens the biological night

Retrospective analysis of two laboratory studies, 116 healthy volunteers aged 18 to 30, at least five consecutive days as inpatients. Room light below 200 lux was compared with dim light below 3 lux in the 8 hours before bedtime; primary endpoints were the melatonin parameters. Result: later onset in 99.0 % of people, melatonin duration about 90 minutes shorter, suppression by more than 50 % in 85 % of trials. It is a retrospective analysis, not a randomized study designed for this purpose.

From 5 lux, sleep structure changes

Polysomnography in 23 analyzed healthy young men aged 21 to 29. Night 1 without light, night 2 randomized to 5 lux in 11 people or 10 lux in 12 people, dominant wavelength 501.4 nm. Under light, wake time after sleep onset (p = 0.014), stage N1 (p = 0.045) and REM sleep (p = 0.017) increased, and stage N2 decreased (p = 0.006). No difference was measurable between 5 and 10 lux.

One night at 100 lux worsens cardiometabolic measures

Laboratory study in a parallel-group design, 20 healthy adults, 10 per condition. Night 1 below 3 lux for both groups, night 2 at 100 lux for one group; an oral glucose tolerance test followed in the morning. HOMA-IR rose by around 15 % versus a decrease of around 4 % (p = 0.018), the Matsuda index fell by around 16 % versus an increase of around 3 % (p = 0.048), and heart rate during sleep rose more. The glucose values did not differ (p = 0.18).

Sleep mask: one finding and one counter-analysis

The only study on the practical measure. In experiment 1, 94 people aged 18 to 35 wore a sleep mask every night for one week and went through a control week without light blocking; after five habituation nights each, a cognitive test battery followed on days 6 and 7. Result: better episodic encoding and alertness with the mask, replicated in experiment 2 with 35 people. A reanalysis published in the same journal sees no clear and consistent evidence in this.

Where the data stop

For the standby light there are no measurement data. The human studies work with 5 lux, 100 lux or less than 200 lux; a single LED lies below that and has never been studied in isolation. Anyone who tapes it over is following a plausible extrapolation, not a measured result. Conversely, the gap is smaller than it sounds: around 100 lux already produces half of the maximal effect of 9,000 lux, and only the range below 5 lux is unsupported.

The study on the metabolic markers has 20 participants, a single night and exclusively surrogate markers; long-term consequences cannot be answered with it. And red light is not without effect: six hours of red light at 631 nm did not suppress melatonin in a study with 24 people, but shifted the phase by almost an hour on average.

How to do it

The dark condition in the laboratory was consistently below 3 lux. The evening window was generous: 8 hours of darkness before bedtime in one study, 6.5 hours in the early biological night in the other. For sleep itself, 5 lux over the whole night was enough; for metabolic markers, it took 100 lux. The only dosing tested under practical conditions is the sleep mask: every night for one week, with five habituation nights before measurement.

Safety

Complete darkness is physically harmless; the risk is mechanical: anyone who gets up at night finds no orientation in a completely darkened room. For older people, with limited mobility, visual impairment, nighttime urge to urinate or on blood pressure-lowering medication, dimmed orientation lighting is the safer option. Sleep masks can press and dry out the eyes and are unsuitable for eye diseases or after eye surgery. People with bipolar disorder should discuss changes to light and sleep routines with their treating practice: in their case, nighttime light is associated with manic relapses.

Sources

• Gooley et al., J Clin Endocrinol Metab 2011
• Zeitzer et al., J Physiol 2000
• Cho et al., Chronobiol Int 2016
• Mason et al., Proc Natl Acad Sci U S A 2022
• Obayashi et al., Am J Epidemiol 2018
• Greco et al., Sleep 2023
• Rhodes, Sleep 2023 (reanalysis of Greco et al.)
• Ho Mien et al., PLoS One 2014