Screens and sleep

"No screens before bed" is the most repeated item on any sleep advice list, and the version most people have absorbed — that blue light from phones is uniquely harmful — is a compressed and slightly distorted account of what the research shows.

The distortion matters, because it sends people to the wrong lever. They buy amber glasses and enable night mode and keep the phone in bed at full brightness for two hours, and then conclude that the evidence must be nonsense because nothing changed.

The stance here, stated up front: brightness and timing are doing most of the work, what is on the screen is doing a lot of the rest, and the wavelength is the part with the least practical leverage. The most commercially successful intervention — blue-blocking lenses — is also the one with the weakest evidence.

The core physiology is not in dispute. The eye contains receptors most sensitive to short-wavelength light that feed the body clock rather than vision, and light reaching them in the evening suppresses melatonin and pushes the clock later.

Two findings put usable numbers on it.

The threshold is far lower than people assume. Phillips and colleagues' 2019 study found that melatonin was suppressed by fifty percent at under thirty lux on average — dimmer than ordinary indoor lighting, and comfortably within what a living room provides. The screen is not the only light source in the room, and often not the brightest.

Individual variation is enormous. In that same study, the level required for half-suppression ranged from around six lux in the most sensitive participant to around three hundred and fifty in the least. That is a fifty-fold difference between two people in the same room. It means the question "does evening light affect sleep" has no single answer, and that the friend who scrolls until midnight and sleeps fine may genuinely be built differently.

For screens specifically, Chang and colleagues' well-controlled 2015 study had people read on a backlit tablet or a paper book for four hours before bed across several nights. The tablet readers took longer to fall asleep, had suppressed melatonin, had their clock delayed by over an hour, and were sleepier the next morning despite equal time in bed. Cajochen's earlier work found similar effects from an LED-backlit computer screen. The effects are real and they are more about the next morning than the night itself.

If short-wavelength light is the mechanism, filtering it should help. This is exactly the kind of reasonable inference that needs testing, and it has been tested.

Lawrenson and colleagues' 2017 systematic review found the evidence for blue-blocking spectacle lenses weak across all the outcomes claimed for them. The 2023 Cochrane review by Singh and colleagues — the most rigorous synthesis available — reached a similar conclusion: the trials do not support a clear benefit, the evidence on sleep outcomes is inconclusive and of low certainty, and the lenses do not reduce eye strain from computer work either.

Why the gap between mechanism and product? Probably because the filters remove only a fraction of the relevant light while total brightness stays the same, because the effect of a screen at arm's length is smaller than a laboratory light box, and because what people do on the screen was never the thing being filtered.

We flag this because it is the sort of claim that has become almost universal by repetition. Dimming the room does more than filtering the wavelength does. Night mode on a phone is not useless, but it is a small lever presented as a large one.

The variable most consistently associated with worse sleep in real-world studies is not light at all.

Exelmans and Van den Bulck found in a general adult population that bedtime phone use predicted later bedtimes, longer time to fall asleep, poorer sleep quality and more daytime fatigue. Phones are worse than televisions in most such comparisons, and the obvious differences are not photometric: a phone is interactive, it is responsive, it is social, and it has no ending.

Three mechanisms sit behind that, and each has a different answer:

• Arousal. A message, an argument, a work email at eleven at night activates the system described in tired but wired, and no filter touches that.
• Displacement. The hour spent scrolling is an hour not spent asleep. This is arithmetic, not physiology, and it is probably the largest single effect — see revenge bedtime procrastination for why that hour is so hard to give up.
• No natural end. A book has a chapter, an episode used to have a credit sequence. An infinite feed has no point at which stopping is the obvious move, so the stopping has to come from outside it.

In rough order of how much they are likely to do:

Move the hour, not the wavelength. If the phone is costing you an hour of sleep, that hour is the intervention. Nothing else on this list competes with it.

Dim the whole room, not just the screen. Given the thirty-lux figure, overhead lighting matters at least as much as the device. Low, warm lamps for the last hour is a change you can feel within a few days.

Pick the content deliberately. Same screen, different night: a slow documentary and a work inbox are not the same stimulus, and the arousal difference outweighs the light difference.

Get bright light in the morning. This is the part of the advice nobody sells anything for, and it is the strongest lever on the clock. Morning light pulls it earlier, which makes evening sleepiness arrive on time.

Run your own test before believing any of it. Given fifty-fold variation in sensitivity, general advice here is weaker than personal evidence. Two weeks with the phone outside the bedroom will tell you more about your own system than this page can — and the Sleep Quality test gives you something to compare before and after. If the nights stay bad after the screens are gone, the problem was not the screens, and the entry on sleep hygiene explains what usually is.