Bedroom noise, air and heat: what actually costs you sleep

The thermostat is the one bedroom number people chase. In the only study that measured four bedroom exposures at once, in real homes, over two weeks, it came out in the middle. Basner and colleagues put continuous sensors in the bedrooms of 62 adults for 14 consecutive nights and tracked sleep with wrist actigraphy. Hourly sleep efficiency in the highest-exposure category was 4.7% lower for noise, 4.0% for carbon dioxide, 3.4% for temperature and 3.2% for fine particulate. Peer-reviewed — Basner et al., 2023.

That is not a league table, and the rest of this piece is mostly about why. But it is enough to raise a question worth asking: of the four things measured in those bedrooms, why is the one people buy gadgets for the only one anybody talks about?

What was actually measured

The design matters here more than the numbers. Sixty-two adults in Louisville, Kentucky, average age 48, each had an air-quality monitor and a decibel meter in the bedroom for 14 consecutive days in the summer of 2021. They wore an actigraph on the wrist. Sleep efficiency, the share of time in bed actually spent asleep, was calculated for each hour after sleep onset and modelled against what the sensors recorded.

Three caveats travel with those four percentages, and they are the difference between reporting this study and quoting it.

• The outcome is hourly, not whole-night. The paper models sleep efficiency computed for consecutive one-hour periods. "Your sleep efficiency drops 4.7%" overstates it.
• The ranking is descriptive. The authors do not test whether the four estimates differ from each other. Temperature at 3.4% and particulate at 3.2% should be read as the same answer, not as fourth place beating third.
• Actigraphy is not polysomnography. A wrist monitor infers sleep from movement. It is the right tool for 14 nights in someone's own bed, and it is not an EEG.

The part that argues against my own headline

I went looking for evidence that carbon dioxide beats temperature. The same paper contains the reason to be careful about saying so.

In the high-exposure contrast, CO₂ looks like the second-largest effect at 4.0%. But when the same authors modelled each exposure as a continuous dose-response, CO₂ was the one that vanished: particulate, noise and temperature all held up, and CO₂ did not (Peer-reviewed — Basner et al., 2023). An effect that appears at the extreme and disappears across the range is a weaker finding than a straight line, and anyone quoting the 4.0% without that sentence is quoting half a result.

A separate field study points the same direction with the same limits. Yan and colleagues followed 45 elderly subjects in Shanghai for six consecutive summer nights and reported that as CO₂ rose by 100 ppm, total sleep time fell by 11 minutes. Peer-reviewed — Yan et al., 2022. That is a real measurement in real bedrooms. It is also 45 older adults in a humid subtropical summer, and the published abstract states the association without stating what it was adjusted for. It is a reason to ventilate, not a number to put on a chart.

Why the temperature number cannot be fixed

Here is the part that changed how I read every bedroom-temperature recommendation, including the one this site used to publish.

Lin and Deng measured the insulation of bedding systems on a heated manikin and found they span 0.90 to 4.89 clo. Feeding those into a comfort equation, the thermally neutral room temperature runs from 30.1 °C at the light end to 8.9 °C at the heavy end. Peer-reviewed — Lin & Deng, 2008. That is a 21-degree spread produced by nothing but what is on the bed.

Akimoto and colleagues showed the same thing from the other direction. They took a room at 18.6 °C and a room at 26.4 °C and brought both to the same equivalent temperature, purely by changing posture, sleepwear and how much of the body the blanket covered. Peer-reviewed — Akimoto et al., 2025.

Both of those are thermal manikins, not sleeping people. Neither measured sleep. What they establish is narrower and more useful than a sleep result would be: bedding moves the comfortable room temperature further than any thermostat setting you would plausibly choose. A recommendation given as a room number, with no mention of what you sleep under, is missing the larger variable.

Which is why the measured optima disagree without contradicting each other. Baniassadi and colleagues followed 50 older adults in Boston and found sleep most efficient between 20 and 25 °C. Peer-reviewed — Baniassadi et al., 2023. Yan and colleagues followed 104 older adults through a Shanghai winter, across bedrooms ranging from 8 to 22 °C, and found that humidity and CO₂ were the factors significantly influencing sleep quality, while sleep quality "was less sensitive to bedroom temperature", which they attributed to the widespread use of bed heating. Peer-reviewed — Yan et al., 2025. Warm the bed and the room matters less. That is the clo range doing its work in the field.

What our own data says about the one compound on the list

Of the four exposures, exactly one is a compound in the Embr Exposure Ledger: fine particulate matter. Its verdict there was written before this article and it happens to line up: real, and the one a HEPA purifier genuinely fixes, with the bedroom sources being cooking, candles, smoke and outdoor haze rather than the bed.

Noise is not a compound and never will be. And carbon dioxide is not in the Ledger at all, because the Ledger tracks compounds of concern and CO₂ at bedroom concentrations is not one. It is a proxy: what a CO₂ reading tells you is whether the room is exchanging air with outside. That is worth saying plainly rather than quietly widening the definition to make our own dataset look more relevant than it is.

It also points at the practical conclusion. Ventilation is the single action that moves more than one of these at once: it lowers CO₂ by definition, and it is one of the few things that clears the compounds we do track, the ones that come off a new mattress and settle in household dust.

What this does not establish

These are small effects. A few percent of hourly sleep efficiency is not the difference between rest and exhaustion, and none of these studies followed anyone long enough to say what a few percent does over years.

The populations do not line up. Basner's cohort averaged 48 and was studied in summer; both Yan studies and Baniassadi's are older adults, in two different climates and two different seasons. Older people thermoregulate differently, and Yan's 2022 paper says outright that its subjects were more affected by heat than younger subjects have been.

Group averages hide people. In Baniassadi's within-subject analysis, of the 21 participants with a statistically significant temperature relationship, eight ran in the opposite direction to the group. The authors' own conclusion is a call to personalise, not a number to adopt.

And none of this is causal. Every study here is observational. Nobody was randomised to a noisy bedroom.

What I would actually do

Ventilate before you optimise. It is the cheapest of these levers, it is the only one that moves several at once, and a stuffy room is the one condition all three field studies flagged.

Match bedding to the room instead of chasing a setpoint. Twenty-one degrees of neutral temperature sit inside that choice, which is more range than your thermostat will give you.

Take noise seriously, because it had the largest and statistically firmest association of the four and it is the one nobody sells you a mattress for.

Watch the warm side rather than the cold. The measured penalties climb as rooms get hotter, and if you overshoot cold you can add a blanket, which is precisely the 21 degrees of adjustment the manikin studies describe.

Frequently asked questions

References

1. Basner M, Smith MG, Jones CW, et al. Associations of bedroom PM2.5, CO2, temperature, humidity, and noise with sleep: An observational actigraphy study. Sleep Health. 2023;9(3):253-263. doi:10.1016/j.sleh.2023.02.010 (PMID 37076419) Peer-reviewed
2. Yan Y, Lan L, Zhang H, et al. Association of bedroom environment with the sleep quality of elderly subjects in summer: A field measurement in Shanghai, China. Building and Environment. 2022;208:108572. doi:10.1016/j.buildenv.2021.108572 Peer-reviewed
3. Yan Y, Lan L, Gong P, Guo C, Hou Z. Association between bedroom environment and sleep quality of older adults: A winter field study. Journal of Building Engineering. 2025;99:111497. doi:10.1016/j.jobe.2024.111497 Peer-reviewed
4. Baniassadi A, Manor B, Yu W, Travison T, Lipsitz L. Nighttime ambient temperature and sleep in community-dwelling older adults. Science of the Total Environment. 2023;899:165623. doi:10.1016/j.scitotenv.2023.165623 (PMID 37474050) Peer-reviewed
5. Lin Z, Deng S. A study on the thermal comfort in sleeping environments in the subtropics: Measuring the total insulation values for the bedding systems commonly used in the subtropics. Building and Environment. 2008;43(5):905-916. doi:10.1016/j.buildenv.2007.01.027 Peer-reviewed
6. Akimoto M, Shinoda J, Bivolarova MP, et al. Effect of bedding on total thermal insulation in different sleeping postures measured with thermal manikin and modelled with JOS-3. Building and Environment. 2025;279:113074. doi:10.1016/j.buildenv.2025.113074 Peer-reviewed

Related reading: why you sleep hot, brain fog and your bedroom, mattress off-gassing.