An ordinary household mirror is white with a faint green cast. Rendered on your screen, the color it turns a sheet of white paper is #F1F8F7. It reflects 92.87 percent of daylight, it peaks at 520 nm, and the tint sits 3.69 CIEDE2000 away from a perfectly neutral surface of the same lightness. It is not silver, and it is not colorless.
That is the short answer. The reason it took a measurement to get there is that almost every version of this question online stops at either “white, because a mirror reflects everything” or “actually slightly green”, and neither one commits to a number. Slightly green is not an answer. Green by how much, and green because of what?
So I built a mirror out of published optical data and rendered it.
How I measured it
A common bathroom mirror is a second surface mirror, which means the reflective coating is on the back. Light does not hit metal first. It goes through a few millimetres of window glass, bounces off a silver film, and comes back out through the same glass. Three things happen on that trip, and each one has published numbers behind it.
- The silver. Reflectance at the glass to silver boundary, from the Fresnel equations using the complex refractive index of silver measured by Johnson and Christy in 1972, still the standard reference for the noble metals.
- The glass. Absorption on the way in and the way out, from Beer's law using the extinction coefficient of clear soda-lime window glass measured by Rubin in 1985. I used a 4 mm sheet, which is typical for a wall mirror, and I ran 3 mm and 6 mm as well.
- The eye. The resulting reflectance spectrum from 380 to 780 nm, multiplied through CIE illuminant D65 and the CIE 1931 two degree observer, converted to CIELAB and to sRGB.
Two simplifications worth stating up front. I modelled light arriving straight on rather than at a slant, which is how you look at a mirror when you are using it. And I left out the four percent reflection off the front face of the glass, because that bounce never reaches the silver and shows up as the faint ghost image you can see doubled at the edge of a thick mirror rather than as part of the main reflection. Neither choice moves the color by more than a rounding error.
Finding one: the green is the window, not the silver
This is the part that surprised me, and it is the part every popular explanation gets backwards. People say mirrors are green because silver reflects green best. Silver does no such thing.
Run bare silver on its own and it comes out at 98.47 percent reflectance and 0.42 CIEDE2000 from neutral, which is as close to colorless as any real material gets. What tilt it has runs the other way, faintly warm, because silver gets better as you move toward red: 96.38 percent at 400 nm, 98.31 percent at 550 nm, 99.56 percent at 750 nm. Bare aluminium, the coating used on telescope and camera mirrors, is 0.38 from neutral with a slight cool lean. Both metals are, for practical purposes, perfect grays.
The glass is where the color comes from. Window glass is made with sand that carries iron, and the iron absorbs at the two ends of the visible range while leaving the middle alone. Through 4 mm of clear float glass and back out again, the double pass transmits 96.06 percent at 550 nm and 83.76 percent at 700 nm. Silver over the same interval goes from 97.65 percent to 99.02 percent. The metal is trying to make the reflection redder. The glass overrules it by a wide margin.
So the honest one-line version is that a mirror is green for the same reason the edge of a pane of window glass is green. You are looking through a window twice, and the tint of the window is what you get.
Finding two: it peaks at 520 nm
Here is the full reflectance of the modelled mirror, averaged in bands across the visible spectrum.
The single best wavelength is 520 nm at 94.79 percent, the worst is 780 nm at 74.69 percent, and everything from 460 to 550 nm sits within one percentage point of the peak. Green wins, and it wins mostly by red losing.
The 495 to 570 nm band coming out on top is a result I did not arrange. In 2004 Raymond Lee and Javier Hernández-Andrés took a spectroradiometer to the mirror tunnel at the science museum in Granada and measured the light coming back out of real mirrors. They reported that common mirrors reflect best between 495 and 570 nm. My model, built from first principles out of two unrelated datasets, lands on the same band and puts the dominant wavelength of the reflection at 495 nm. When a calculation and a spectroradiometer agree without being made to, that is usually a sign both are right.
Finding three: the tint is above threshold, and you still cannot see it
3.69 CIEDE2000 does not sound like much. It is more than it sounds. When we measured the just noticeable difference for ordinary observers, the threshold came out around 2.3, which is the figure the printing industry works to as well. The green of a bathroom mirror is roughly one and a half times that. Put a swatch of #F1F8F7 next to a swatch of neutral white of the same lightness and most people would see the difference immediately.
Nobody sees it in the mirror, and the reason is worth sitting with, because it is the whole business of color perception in one household object. A visible difference needs two things to compare. A mirror never gives you the second one. It shows you the reflected version of the room and it does not show you the room, so there is nothing to hold the reflection against. On top of that, your visual system is busy discounting the light source anyway, which is what color constancy does all day, and a uniform 3.69 shift across an entire field of view is exactly the kind of thing it is built to subtract.
This is also, more or less, why matching colors from memory is hard. The error people make when they recall a color and try to reproduce it is usually much larger than 3.69, and they do not notice that either, for the same structural reason: nothing is sitting next to it. You can test your own number on that in about a minute with the color memory game, which is the same comparison problem with the answer revealed at the end.
Finding four: stack the bounces and the tint stops hiding
There is one everyday arrangement that does supply a reference, and it is the reason anyone noticed this in the first place. Face two mirrors at each other and you get a tunnel of reflections, each one a bounce deeper than the last, all visible at once in the same picture. Every bounce multiplies the spectrum by itself again, so the tint compounds while the near rungs of the tunnel sit right there as a control.
Here is what a white wall becomes as it goes down the tunnel, computed by raising the mirror's reflectance spectrum to the power of the bounce count.
Three things fall out of that table that I had not expected.
The color saturates, then reverses
Chroma climbs steeply for the first ten bounces, then flattens, peaks around the fortieth bounce at 33.3, and starts falling again. The tint does not keep intensifying forever. The tunnel runs out of light before it runs out of green, and once the reflection is dark enough, dark is the dominant impression rather than green. Deep tunnel images go black, not radioactive.
It ends up cyan rather than green
The dominant wavelength drifts from 495 nm at one bounce to 500 nm by twenty five, and the rendered colors go from a barely tinted white to a distinctly blue-green teal. Calling the mirror tunnel green is a decent first approximation, but the far end is closer to teal than to anything you would pick out of a paint chart as green. That is because the glass eats red far harder than it eats blue, so what survives is green with the blue left mostly intact.
Each bounce warms the light in kelvin terms
Reflecting off the mirror shifts D65 daylight from 6504 K to 6703 K, and by ten bounces the light is at 8256 K. That is a real effect worth knowing about if you are shooting into a mirror or lighting a room through one, and it is a nice illustration of why correlated color temperature is a coarse tool. The tunnel is heading toward green, which is not on the blackbody locus at all, and the kelvin number keeps reporting obediently anyway.
The mirror you own is not the mirror I modelled
The tint is a property of the glass, so it scales with how much glass there is and how much iron is in it. Those are the only two knobs, and both are things you can buy differently.
- 3 mm clear glass: 94.09 percent reflectance, 2.80 from neutral.
- 4 mm clear glass: 92.87 percent, 3.69 from neutral. The default in this article.
- 6 mm clear glass: 90.48 percent, 5.33 from neutral. A heavy full length mirror is visibly greener than a small one, purely because the light spends longer inside it.
- 4 mm low iron glass: 96.71 percent, 0.79 from neutral. This is the fix. Low iron glass cuts the tint by a factor of 4.7 and drops it under the visible threshold. It is what gets sold as optical mirror or ultra clear mirror, and it costs more for exactly this reason.
- 4 mm body tinted green glass: 74.97 percent, 11.47 from neutral. Some cheap decorative mirrors use tinted stock, and at this point the mirror is unambiguously a colored object.
A front surface aluminium mirror, the kind inside a telescope or a projector, has no glass in the light path at all and comes in at 0.38 from neutral. If you want a genuinely colorless mirror, that is the one. It is also fragile in a way no bathroom would tolerate, since the coating is exposed and a wipe with a cloth will scratch it. Household mirrors are green because putting the metal behind glass is the only practical way to make a mirror that survives being cleaned.
So is a mirror silver?
No, and the confusion is a good example of how loosely we use color words. Silver as a color name, usually written #C0C0C0, is a mid gray. A mirror is nothing like a mid gray. It reflects 92.87 percent of the light that hits it, which puts it up with the brightest white paint you can buy, and well above white paper.
What people mean by silver is not a color at all. It is a description of how the surface scatters. A white wall sends light off in every direction, so you see the wall. A mirror sends every ray off at the angle it arrived, so you see whatever the light came from. Same brightness, completely different behaviour. When we draw a mirror as gray in a picture, we are drawing that behaviour, because a flat patch of #FFFFFF on paper does not read as a mirror, and we need it to.
Which gives the tidiest way I know of to state the answer: a mirror has the color of a very good white surface, and the manners of a window.
What the numbers add up to
- An ordinary 4 mm silvered mirror reflects 92.87 percent of daylight and renders as #F1F8F7, a white carrying a green cast of 3.69 CIEDE2000.
- The green is the glass, not the metal. Bare silver measures 0.42 from neutral and leans warm; the double pass through iron-bearing window glass drops the deep red to 83.76 percent while the green stays at 96.06 percent.
- Reflectance peaks at 520 nm, and the best band is 495 to 570 nm, which matches what Lee and Hernández-Andrés measured off real mirrors in 2004.
- The tint is above the visible threshold of about 2.3 and you still cannot see it, because a mirror never shows you a reference to compare against.
- A mirror tunnel supplies that reference. Ten bounces cost half the light and land on #8DC4BA, and by twenty five bounces the color is a teal at 22.27 from neutral.
- Low iron glass cuts the tint by 4.7 times, to 0.79, which is below the threshold. Thicker glass makes it worse: 6 mm measures 5.33.
The thing I keep coming back to is that this is a color sitting in plain view, above threshold, in an object almost everyone looks at every day, and essentially nobody sees it. Perception is comparative all the way down. Take away the comparison and a difference you would spot instantly on a paint chart simply stops existing for you. If you want the same lesson in game form, our spot the difference mode is that problem with the comparison put back in, the guide to training your eye is the long version, and the metamerism piece covers the related trick where two surfaces agree under one light and fall apart under another.