Why indoor lighting is hard to get right and how to fix it

The days are getting shorter in the northern hemisphere, and with the ongoing pandemic, most of us expect to be spending more time in our homes than normal. Creating a healthy home environment is more important than usual, and the light inside your home is an often underappreciated part of this.

There has already been some explicit discussion[1] about the importance of lighting for health and productivity, as well as many mentions of it in other places. Nonetheless, based on discussions I’ve had recently within the community, I get the impression that it is helpful for me to write up the results and tinkering that have done over the past few years.

First, I will cover some of the research on how our bodies respond to light, and which particular characteristics of natural light we want to mimic. Then I will explain solving this problem is hard and my overall strategy for solving it. Finally, I will give some specific advice on what to buy and how to arrange things.

I give quite a lot of background before offering any specific advice. Although I think the background information might help you make good decisions, you should feel free to skip the next section if you’re in a hurry or if it seems uninteresting.

Background

Note: My background is in optics, not physiology or psychology and I began researching and writing this document almost four years ago. My original draft, as well as many of my sources, have been lost in the intervening years, so what you’re seeing here is based on a combination of my notes that survived, my recollection of the research, and a partial duplication of the research. To make matters worse, it does seem that new research has come along since I began this project, so this is likely out of date. My guess is that most or all of the practical conclusions still stand, but I am only moderately confident of this. As much as I would like to take the time to update the research, past experience suggests that I will never actually publish it if I try to put too much more work into it. I welcome corrections, and if there is sufficient enthusiasm around this topic, I may try to write an updated version.

Your body uses light to synchronize its internal clock and to modulate your mood and alertness[2]. While the particulars of the lighting in your environment are important, your only perceptual access to information about lighting in the moment comes from your visual system, which is poorly adapted to solving the problem of determining the intensity and spectrum of a light source. This is mainly because our vision is optimized more for accurately identifying materials, textures, colors, and other properties of our surroundings than it is for knowing details about sources of light. This has the consequence that some of our default intuitions about the nature of ambient light are wrong, so when we’re building our lighting environment, it can be difficult to make accurate judgments just by looking at things. We can do better if we use quantitative measures and our scientific understanding of how things work to solve the problem.

Physiology

In addition to visual photoreceptors that are used for seeing things, your eyes contain non-visual photoreceptors which serve non-visual functions. Melanopsin is a photopigment that is found in cells in the retina[3]. It is sensitive to blue light, and when activated, these cells send signals that help with things such as regulating our internal clocks[4]. Unlike the our visual photoreceptors, which are more densely packed in the center of our retina[5] than in the periphery, these photosensitive cells are distributed relatively evenly throughout the retina[6], so that light coming from both the periphery and the center of our vision is important. Although I have found contradictory accounts of how visual photoreceptors and melanopsin-sensitive cells interact, there does seem to be some evidence that activation of our visual photoreceptors can influence the functioning of our melanopsin-containing cells. This seems consistent with many people’s experience that blue LED lighting is more relaxing than white light. On the other hand, it may just be that it takes a lot of light to activate these cells[7].

I find the literature on how our mood and circadian rhythm respond to light exposure throughout the day to be somewhat confusing, and I have not tried to unravel it in detail. As far as I can tell, it depends on the full history of exposure over a course of days, not something simple like “total light exposure over the past N hours” or “Exponentially-weighted average of light exposure”. For this reason, I am not particularly optimistic about figuring out the details of the algorithm used by our bodies to decide when to be tired or when to be happy, or any other aspect of our lives that we might want to improve.

Given this, as well as my priors in favor of just letting our bodies have the regulatory mechanisms they’re “designed” for, it seems that we should try to create lighting that is similar to natural light and well-synchronized to our preferred schedules for sleeping, working, and relaxing. This means that, in order to signal to our bodies that it is daytime, and therefore time to be wakeful, we should use light that is bright, white, and distributed over our full field of view.

When we want to signal to our bodies that it is time to start winding down for the day, we want to avoid bright, white light. I have not done as much research on the best timing for switching over or the optimum spectrum for this, but based on the fact that it is dark at night, sunsets are yellow, and on my personal experience and that of people I know, there should be several hours between bright daytime lighting and bedtime, and the difference between the best lighting for the late evening and the daytime is quite large, both in terms of intensity and color.

Natural vs Artificial Light

Creating light that effectively mimics natural light is difficult. The first problem is that natural light is almost always much brighter than artificial light. Even on a very cloudy day, you might notice that the insides of buildings look dark from the outside, the brightest parts of most indoor spaces are next to windows, and if you use a camera or a light meter to measure the amount of light, you’ll find that things are more brightly illuminated outdoors than indoors.

The second problem is that natural light has a broad, continuous spectrum, which contains only a few gaps within the visual range of wavelengths, while artificial light almost always has large portions missing, especially at shorter wavelengths. This has to do with how artificial light is generated. Incandescent bulbs are blackbody sources, but they’re not hot enough to contain very much light at shorter wavelengths. Incandescent bulbs are usually around 3000K while the sun is 5800K. Fluorescent lamps and LEDs generate UV and blue light, which is then absorbed and re-emitted at longer wavelengths by phosphors. This technique produces light that can have a correlated color temperature (CCT) anywhere from 2000K or less to over 6000K[8]. Unfortunately, this results in a very spiky spectrum in the case of fluorescent bulbs and a large dip between blue-ish and yellow-ish wavelenths for most LEDs. Here are some spectra from Wikipedia[9]:

All this missing light can be easy to overlook. Our visual system projects all the spectral content onto three dimensions (or two, depending on how you look at it), and then does a remarkable job of giving us accurate information about the color of objects in a variety of lighting conditions (the blue-black/​white-gold dress phenomenon is remarkable because of our failure to reliably solve this problem), which is great for many tasks, but terrible for evaluating the quality of a source of light.

Nonetheless, this missing light is important. In addition to failing to provide the appropriate signals for time of day, poorly filled out spectra can make things look unappealing, give us inaccurate information about the colors of objects, and make an environment feel generally less pleasant.

What I am not trying to replicate (for now)

There are some properties of natural light that are less important or more difficult to replicate. Although these might be desirable in certain contexts, we’ll ignore them for now.

  1. Non-visible light: We’re not trying to add in ultraviolet or infrared light. Although getting some UV might be good for people with a vitamin D deficiency, doing this safely and without other consequences (like degrading plastics or bleaching colors in your office or bedroom) seems hard, and I don’t recommend trying to do this.

  2. Spectrum and intensity that changes continuously throughout the day. There are lamps that do this, and they seem especially pleasant to me for waking up in the morning, but I think this is particularly hard to get right and less important than other things

  3. All the little details. Natural light has lots of subtleties that might contribute to making it seem pleasant. For example, the sky is blue and polarized and direct sunlight is a bit yellow. The anisotropy of the color of the light makes for pleasant photographs. Things like clouds and leaves can make soft, moving shadows. Again, it would be neat to replicate this, but it seems hard to get right and I think it’s a much lower priority than other things.

Why this is a hard problem to solve

To illustrate how much your visual system can correct for variations in lighting, I took photos and illuminance measurements of a photographic color calibration card in a variety of environments, shown here:

The camera and software are calibrated to the photo in the top-left with the green border around it, which I took during a typical cloudy day in San Francisco, so the lighting is reasonably close to a 6500 kelvin blackbody spectrum. I tried to get the brightness of each image the same, so the apparent brightness of the display on the lux meter should give you a sense of how much brighter some of the environments are than others.

The yellowest lighting in these photos has a red-to-blue ratio 15 times higher than the bluest lighting, and the intensity varies by five orders of magnitude. Nonetheless, the card mostly looked the same to me in nearly every context. I was aware that the incandescent bulb was dimmer and yellower than the light from the cloudy sky, but the effect is nowhere near as extreme as the photos suggest. The two main exceptions to this were the LCD photo, for which the light was very dim (though I did not try allowing my eyes to adjust), and the low-quality fluorescent lamp, in which many of the colors just didn’t look quite right.

I hope this will give you a sense for why we should not rely on our visual systems to tell us how bright or how blue a light source is. Fortunately, we can measure these things in other ways.

General approach to solving the problem

Given everything that we have so far, we want light that is:

  1. As bright as is practical during the day, preferably at least 1000 lux over the full room, with a CCT close to that of the sun (5500K)

  2. Much dimmer in the evening, with a much lower CCT (2700K)

  3. Covers all or most of our visual field

  4. Has a full spectrum, with few holes at visible wavelengths

  5. Does not have other annoying characteristics like sharp shadows, flickering, or the generation of unnecessary heat or noise

Natural light meets all of these criteria, except in some circumstances being annoying (direct sunlight in your eyes) or being too bright in the evening (if you live very far from the equator). I always try to work near a window if possible, and get as much natural light as I can. That said, at the moment I’m writing this, it has been dark outside for over an hour and it’s only 6:25pm. The whole reason this is a problem to be solved is because we can’t just use natural light all the time.

Getting enough light

As I have mentioned already, getting your indoor daytime light to be anywhere near as bright as outdoor lighting takes effort. If you install some lighting, and you still experience seasonal depression, the first thing I would try is adding more light. Most people will be limited by how many bulbs they can reasonably install in a room, either due to space, power, or heat, before they’ll have too much light. Or they’ll have a bunch of lights and think “Surely this is enough! Look at how many bulbs I have!”, but until it is sufficiently bright to make your computer screen look dim, it is probably not as bright as natural light. In my small living room, I find that my setup which provides 20,000 lumens, and gives me a relatively uniform 1000 lux[10] is sufficient. I wouldn’t mind more light (and I intend to add more once I’m in a more permanent living situation), but it is enough for me to feel alert during the day and avoid or at least reduce the effects of seasonal depression, so long as I’m able to spend time in there later in the day. For a larger room, you’ll need more light, in a way that should scale roughly with the inside area of the walls, ceiling, and floor.

Getting the right spectrum

The spectrum of your daytime light is the trickiest part, but I have found two strategies that seem to work:

  1. Combining lots of different bulbs so they can fill in each other’s gaps.

  2. Finding bulbs with a very high color rendering index, especially those marketed for photography.

For a while during grad school, my office had, in addition to the fluorescent tubes installed in the ceiling fixtures, three different kinds of LED bulbs and two different kinds of very high lumen compact fluorescent (CFL) bulbs. A disadvantage of variety is that you can’t just buy a huge pack of all one kind of bulb, which is usually the most convenient and inexpensive option. An advantage of ordering many different bulbs is that if you get a particular one that you don’t like, you only bought a few of them (for example, I had some which made an annoying buzzing sound and others that had an annoying green tinge). For high CRI bulbs marketed to photographers look for CRI of at least 85-90. A safer bet is to go with 95+. Once you know what you like, when you need to replace bulbs or build a lighting system in a new room, you be able to get it right more quickly.

For daytime color temperature, I recommend 5300-6000K. Many people dislike the light from high color temperature light bulbs, but I think this is because there is a lot of terrible indoor lighting out there, including light from bulbs that have a high color temperature but poor color rendering. A CCT of 6000K is very close to the color temperature of natural daylight through a window. If you get it right, people might even mistake the light coming out of your room for daylight! Still, some people seem to do alright with significantly lower color temperature. I’ve heard of some people doing well with as low as 4000K.

Evening light should be less than 3000K. Incandescent and LED bulbs both work well without much need to combine bulbs to fill in gaps at these color temperatures. If you do find low CCT bulbs to be unpleasant somehow, you might try either incandescent bulbs or high CRI bulbs.

My preferred light for the last hour before bedtime is an orange LED bulb with so little blue light that I can’t tell blue objects apart from black.

Covering your full visual field

It might be tempting to get a therapy lamp or to get fewer bulbs and just put them in the corner of the room where you’re sitting all the time. But this can be hard to get right. For starters, remember that the photoreceptors you’re trying to activate cover your entire retina, so you want to illuminate your full visual field if possible. This covers a very wide angle of 210 degrees, which extends slightly behind you. Once you account for turning your head some of the time, this can easily extend to well over half the room. Another problem, at least for me personally, is feeling confined to one part of the room. This is more of a problem at home than it is at my office. When possible, I recommend just illuminating the whole room.

Still, this may not be an option for everyone. If I’m working in a shared space, I don’t want to impose my preferences on everyone else, nor do I want to spend hundreds of dollars on lights that I’ll only use for a couple months. Sometimes I have lights arranged just to illuminate my own part of the shared space. Similarly, some people use therapy lamps in these settings.

Evening and nighttime lighting

Evening lighting is easier. The main difficulty is in having multiple sets of lights for different parts of the day. Some people will only need daytime lights for their workplace. I use inexpensive low CCT LED and fluorescent bulbs, usually 2700K, and I switch from my bright daytime lights to those at 7pm. In my bedroom, I also have an orange LED bulb that is very dim, but sufficient for reading, and I switch over to this for the last 30-60 min before going to sleep.

The little things

I said I’m not trying to solve all of the problems with reproducing natural light indoors, but there are a few small things that can get it a little closer, and which can be nice.

Lightbulbs can cast annoying, sharp shadows, and if they are in your visual field, they can be annoyingly bright. Diffusers help with this. Most light fixtures have a lampshade or some other device to increase the effective size of the source, softening shadows and reducing the intensity of the light fixture itself. If you want diffusers that are spectrally neutral (that is, they do not absorb some colors more than others), you can find those in photography stores.

Having a bit of variety in the spectrum can be nice. Outdoors on a sunny day, shadows are blueish, and reflected sunlight is yellowish, for example. Because of this, even if you had bulbs with a perfect 6000K blackbody spectrum, it may still seem unnatural. One thing that can help with this is to add in some lower color temperature lights. Personally, I find that one ~3000K bulb per three ~5500K bulbs is very pleasant and less artificial-feeling than just 5500K[11].

Specific recommendations on what to buy

Since links for buying things online are only useful geographically and for a limited time, I will recommend brands that I have been happy with and explain my overall strategy, along with links to specific products.

Lux meter

Being able to measure how much light is in your house is useful and inexpensive. I use a $30 Uceri meter that I ordered from Amazon.

Light bulbs

I have found that fluorescent bulbs that are designed for photography are usually a good bet for color rendering. In particular, I have been using these very large bulbs from Alzo (note that these bulbs are huge and fragile, will not fit in most fixtures, and probably should not be used with strings of light sockets). For LED bulbs, I have been reasonably happy with bulbs from Alzo and Cree. Phillips and GE make bulbs that I have been happy with, but they also make bulbs that I have been quite disappointed with, so be careful.

For LEDs, I know one person[12] who has had success with Yuji corncob bulbs, for both high and low color temperature. He was kind enough to give me one of the 3200K bulbs, which I find to create very pleasant light, and which I now mix in with my 5500K bulbs during the daytime.

As I mentioned earlier, evening bulbs are generally easier, since high CRI, low CCT bulbs are simpler to make. Any incandescent should have a nice spectrum, and most LEDs with high CRI should be okay.

Bedtime bulbs

I’ve been satisfied with my orange LED bulb from Sunlite. One person I know prefers red bulbs over orange. I do not recommend incandescent bulbs with red/​orange filters, as they are less efficient, get hot, and have a filament that still looks white-ish through the filter, which find I to be annoying.

Fixtures

Installing these bulbs can be a bit tough, especially the gigantic CFLs. Usually you’ll want to have them above eye level, or at least have a diffuser. They need to be somewhere that they don’t get too hot, and where they cast light in a way that illuminates the room evenly.

If you want to build your own fixtures with neutral diffusers, I purchased a couple of generic diffuser “socks” from Amazon several years ago, and I’ve been happy with them. They are no longer available, but if you want to see what I’m talking about they’re here.

If you have many fixtures or fixtures in inconvenient locations, you may want to get a remote like this one from IKEA. I would imagine that a smart lighting setup could help, but I have not tried one myself.

Photography stands

I like using photography lighting stands because they’re inexpensive and tall enough to get the bulbs well-above eye level. If you’re worried about them being ugly, they can probably be decorated (I intend to add some fake ivy to mine, for example). I use these from Amazon

DIY fixtures

It is also easy to build something that can sit on top of a shelf. One way to do this is to use cable ties to attach a power strip to a wooden board, and stick socket adapters in the power strip. Unforunately, I cannot share a photo of the last one that I built, because the pandemic ate it.

Wrapping up

I hope this has been helpful. The most important things to remember are:

  1. More light is good, and it is difficult to have too much light.

  2. Try to cover your full visual field

  3. The spectrum of your light matters a lot, and during the day, you probably want >5000K

  4. During the evening, you probably want <3000K

  5. If you normally find high color temperature (bluish) light to be unpleasant, try using higher quality lighting. Mixing bulbs or using stuff made for photographers helps with color rendering and overall pleasantness

I welcome any comments on your experiences! I know that others have spent time researching and experimenting, and many have probably done a better job of tracking down or constructing good lighting equipment than I have. I am also happy to add links to other people’s write-ups on this.


Acknowledgements: Many people helped me along the way while I was researching and writing this. In particular, thanks to my former lab mates at UT Austin for putting up with my lighting experiments, Meredith Johnson for her encouragement to start the project in 2016, Katja Grace for reminding me that I should just post stuff instead of worrying so much about whether it is bad, and everyone that I’ve shared notes with over the past few years, especially those at the FHI office.


  1. See:

    https://​​www.lesswrong.com/​​posts/​​Ag7oQifJQM5AnMCrR/​​my-simple-hack-for-increased-alertness-and-improved

    https://​​www.lesswrong.com/​​posts/​​BTXdajWzoN2YRbgjG/​​rational-health-optimization

    https://​​www.benkuhn.net/​​lux/​​

    http://​​www.lincolnquirk.com/​​2019/​​11/​​26/​​lumenator.html

    https://​​ryan.abel.space/​​blog/​​adventures-in-interior-lighting ↩︎

  2. https://​​www.ncbi.nlm.nih.gov/​​pmc/​​articles/​​PMC6751071/​​ ↩︎

  3. The retina is the photosensitive part of the eye, analogous to a CCD or CMOS sensor in a camera ↩︎

  4. https://​​science.sciencemag.org/​​content/​​295/​​5557/​​1065 ↩︎

  5. https://​​en.wikipedia.org/​​wiki/​​Fovea_centralis ↩︎

  6. See fig 1E: https://​​science.sciencemag.org/​​content/​​295/​​5557/​​1065 This diagram seems to suggest there are actually more melanopsin-sensitive cells in some parts of the periphery of our visual field. It is plausible to me that you could look at this map, and determine where to put lights in your house, for example, ensuring that there is more light in the upper portion of your visual field, but I have not taken the time to decipher that diagram into such a map of where the illumination should be in our visual field. ↩︎

  7. A potentially stupid and probably not very informative experiment I did once: I used to work in a lab where I had to wear laser safety glasses all the time. These glasses effectively cut out all of the blue light, in addition to severely reducing overall light transmission. Wearing these all day is really bad for people who suffer from seasonal depression, so I tried attaching some blue LEDs to the temples of the glasses so that the light would reflect off the inner surface of the lenses and into my eyes. When I tried them in the lab, the first thing I noticed was that the lab seemed way less depressing than usual. After maybe 30 minutes, I noticed that I was feeling somewhat less drowsy and less compulsion to go outside than I often did in the lab. Unfortunately, the lights gave me a terrible headache, so I never used them enough to see if this was a real effect or not.

    Note that although melanopsin-sensitive cells are referred to as “non-visual” there does seem to be more recent research suggesting that melanopsin may play a role in visual perception as well. ↩︎

  8. Remember that this is the temperature of a blackbody source that is most similar to the the light source in question. Be careful, because the spectrum is often very different from a blackbody! ↩︎

  9. https://​​www.ncbi.nlm.nih.gov/​​pmc/​​articles/​​PMC6751071/​​ has a nice diagram showing spectra for various light sources ↩︎

  10. Lumens are a unit for luminous intensity (the amount of light per time, weighted by how bright it appears to humans)

    Lux is lumens per area, or how brightly-lit a surface is, according to human vision. ↩︎

  11. H/​T to Ben Weinstein-Raun for bringing this to my attention and giving me a bulb to try it out ↩︎

  12. These are the bulbs that Ben recommended to me which I found useful for adding some more yellow light to my daytime lighting. ↩︎