What Actually Determines Sleep Quality (and What Is Just Marketing)

Sleep advice has a signal problem. The genuinely well-supported findings are boring and few; the interesting-sounding ones are mostly underpowered studies or product marketing. Here is what the physiology actually says.

Sleep is not one state

A night of sleep cycles through distinct stages roughly every 90 minutes, and they do different jobs.

N1 is the brief transition into sleep — a few minutes, easily interrupted, and the stage where the falling sensation (a hypnic jerk) happens.

N2 occupies roughly half the night. Body temperature drops, heart rate slows, and the brain produces sleep spindles — short bursts of activity associated with consolidating motor learning.

N3, slow-wave sleep, is the deep stage. Growth hormone is released, the immune system does much of its maintenance work, and the glymphatic system clears metabolic waste from brain tissue at a substantially higher rate than during waking. This is the stage people mean by "deep sleep," and crucially, it is concentrated in the first half of the night.

REM sleep is where most vivid dreaming occurs, with muscle atonia preventing you from acting it out. It appears to be central to emotional processing and to consolidating complex, associative memory. REM periods lengthen through the night, so most REM occurs in the last few hours before waking.

That asymmetry has a practical consequence: going to bed two hours late costs you deep sleep, while waking two hours early costs you REM. They are not interchangeable losses, and "I got six hours either way" hides a real difference.

The factors with solid evidence

Consistent timing. Regularity of sleep and wake times predicts health outcomes at least as well as total duration — in some large cohort studies, better. The circadian system entrains to a schedule, and an irregular one keeps it perpetually mid-adjustment. If you change one thing, change this.

Light exposure, in both directions. Morning daylight — 10 to 30 minutes, outdoors, within an hour or two of waking — is the strongest signal for anchoring the circadian clock. Indoor lighting is roughly 100 times dimmer than an overcast sky and does not substitute. In the evening, bright light suppresses melatonin onset; this effect is real, though the specific fixation on blue light from screens is weaker than popular coverage suggests. Total brightness matters more than colour.

Temperature. Core body temperature must fall by about 1°C for sleep to initiate and to sustain deep sleep. A cool room — roughly 16–19°C for most people — supports this. A warm bath 1–2 hours before bed works by the same mechanism: peripheral blood flow increases, and core temperature drops afterwards.

Caffeine timing. Caffeine's half-life is about 5–6 hours, meaning a 2pm coffee leaves a meaningful dose in your system at 10pm. It blunts slow-wave sleep even when it does not prevent falling asleep — people who report "caffeine doesn't affect my sleep" are often correct about sleep onset and wrong about sleep depth.

Alcohol. Reliably shortens sleep onset and reliably degrades the night. It suppresses REM in the first half and causes fragmented, shallow sleep in the second as it metabolises. It is a sedative, not a sleep aid, and the distinction shows clearly on any objective measure.

What the evidence does not support

That everyone needs exactly eight hours. Adult requirements distribute around roughly 7–9 hours, with genuine individual variation. A small number of people carry rare variants that support comfortable short sleep; most people who believe they are in this group are not, but the population is not uniform either.

That you can meaningfully "train" yourself to need less. Performance on objective tests continues to degrade under restriction even after subjective sleepiness plateaus. You adapt to feeling tired, not to being unimpaired.

That most supplements do much. Melatonin is a timing signal rather than a sedative, and the doses commonly sold — 5 to 10 mg — are roughly ten to thirty times higher than what is needed to shift circadian phase. It is genuinely useful for jet lag and shift work at low doses taken early; it is not a general sleeping pill. Evidence for magnesium, valerian, and most sleep blends is thin.

About your sleep tracker

Consumer wearables estimate sleep stages from heart rate, heart rate variability, and movement. The reference standard — polysomnography — measures brain activity directly with EEG.

Validation studies consistently find that wearables are good at detecting sleep versus wake (often above 90% agreement) and considerably weaker at classifying stages, with deep and REM estimates frequently off by substantial margins. Your tracker's "deep sleep: 42 minutes" is an inference from wrist data, not a measurement.

This matters because of a documented effect sometimes called orthosomnia: anxiety about tracker scores that itself degrades sleep. Trackers are reasonable for spotting trends in timing and duration over weeks. They are not diagnostic, and a number that looks bad on a morning you feel fine is more likely a measurement artefact than a problem.

When it is not a habits problem

Persistent difficulty despite good conditions warrants a clinical answer rather than another article. Loud snoring with pauses in breathing, daytime sleepiness despite adequate hours, or insomnia lasting more than three months are all reasons to see a doctor. Obstructive sleep apnoea in particular is common, substantially underdiagnosed, and highly treatable — and no amount of sleep hygiene addresses it.

For chronic insomnia, cognitive behavioural therapy for insomnia (CBT-I) has a stronger and more durable evidence base than medication, and is the recommended first-line treatment in major clinical guidelines.

This article summarises published research for general information and is not medical advice.