There Is No White

colour science  ·  the colour your eye keeps re-deciding

Ask the internet whether white is a colour and it answers twice, at once, in opposite directions: white is all the colours mixed (true of light); white is no colour at all (true of paint). Both miss the deeper thing. There is no white wavelength. You don't need all the colours to make it. And it isn't even a fixed colour — it is the brightest neutral your eye has currently decided on.

Here is the question stated cleanly, because the usual answers are a muddle: is white all the colours, or none? The "all colours" answer comes from light — combine every wavelength and you get white. The "no colour" answer comes from pigment — a white surface is the one that absorbs nothing, reflecting all of it back. Both are real facts about two different systems, and both quietly assume white is a property of the stuff. It isn't. White is a property of you. This page builds the case live, from the same CIE 1931 colour-matching data that underpins every screen you own, in three moves: white has no wavelength; white needs only two of them; and white is not fixed at all.

I · White has no wavelength

Split sunlight with a prism and the colours fall out in order — red, orange, yellow, green, blue, violet — each a single wavelength of light. Drag the slider and watch one pure colour walk the rim of the diagram below, the spectral locus, the curve of everything a single wavelength can be. Now look for white. It isn't on the rim. It sits inside, alone in the middle, marked E — and the closest any single wavelength ever comes to it is 0.24 away, a yellow near 578 nm that is plainly not white. No beam of light is white. White is a place on this diagram that no wavelength can reach.

Instrument I · the interiortrace a wavelength → never white
spectral locus (pure wavelengths) the whites (E · D65 · A)
Every point on the rim is one wavelength. The whites live in the middle — no single wavelength touches them.
this wavelength wavelength575 nm chromaticity (x, y) distance to white E is it white?
380 nm — violet700 nm — red
A single wavelength is always a saturated colour on the rim — never the neutral centre.

White behaves exactly like magenta in this one respect: it is a colour with no wavelength, a colour the eye assembles rather than receives. (There Is No Magenta is its sibling — magenta is the hue the eye invents to close the open spectrum; white is the neutral the eye builds at its centre.) But magenta at least needs the two ends of the spectrum to make it. White, it turns out, needs almost nothing.

II · You don't need all the colours

"White is all the colours mixed together" is the line everyone learns, and it is true — but it badly oversells what's required. White is not a roll-call of every wavelength; it's a balance across your three cone types. Anything that lands the balance at the centre is white, and there are infinitely many ways to do it. Below are three recipes for the identical white. The first uses every wavelength. The second uses three. The third uses two.

Instrument II · recipes for whitetwo complementary lights
the lights in this recipe700 nm
the result lands at (x, y) distance to white recipe
Two single wavelengths, mixed in the right ratio, land dead on white.

Drag the slider in two lights mode. For almost any wavelength you pick, the machine finds its complementary — one other single wavelength such that the two, added in the right proportion, fall exactly on the white point. Two lines of light, nothing between them, and your eye reports white. The schoolbook picture — white as a dense sum of all the colours — is one recipe out of infinitely many, and far from the cheapest.

There's an honest wrinkle worth seeing: slide into the greens, roughly 495 to 568 nm, and the machine reports no spectral complement. A pure green has no single-wavelength partner that makes white; its complement is a magenta, which (as the sibling page shows) is itself not a wavelength. So even "two lights" has a gap — but everywhere else, two is enough.

White is not all the colours. It is any balance that lands in the middle — and two lights can carry it there.

And the three whites you just built — the full spectrum, the three screen-primaries, the two complementary lines — are metamers: three utterly different distributions of physical light, collapsing to one identical white, because your eye reports light through only three numbers. The white of the midday sky, the white of this screen, and the white of two laser lines are, to you, the same colour and a different universe of light.

III · White is not even fixed

Here is the move that retires the question entirely. White is not a property of the light at all — it's a property of your adaptation. The same physical surface, lit by different lamps, sends your eye wildly different spectra, and you call it white every time, because your visual system continuously re-estimates "what counts as neutral here" and subtracts the lamp. Turn the lamp from cold daylight to warm firelight below, and watch the sheet of paper refuse to change.

Instrument III · the moving anchordaylight
A sheet of paper and three swatches, lit by the lamp. The paper is a perfectly neutral surface.
the paper reads its raw signal (x, y) after your eye adapts lamp
Adaptation on: the paper holds white while the raw signal slides toward the lamp.

With adaptation on — the way you actually see — the paper holds white across the whole range, while the readout shows its raw signal sliding all the way out to the lamp's own colour. Your eye is silently undoing a shift of 0.16 in chromaticity, a correction larger than the distance from white to a clear blue sky. Switch adaptation off and you see the raw retinal truth: the paper is the colour of the fire. This is white balance — the thing your phone camera fakes in software and your eye does for free, every waking second, without telling you.

One more, with no numbers at all — just two grey squares. They have byte-identical colour; only their surrounds differ. Yet the one on the dark field looks white, and the one on the light field looks grey. Press prove it to draw a bridge of the same grey between them.

on black → "white"
on white → "grey"
both patches are #b9b9b9 — check with any colour picker.

"White" is the brightest thing your visual system has decided is neutral. Change what surrounds it, and the decision changes — without one photon changing. That is the whole of it: white is not in the light, and not on the page. It's a verdict your eye is always quietly re-issuing.

IV · The check

Everything above is recomputed in your browser from one embedded table — the CIE 1931 2° standard-observer colour-matching functions — plus the standard Bradford chromatic-adaptation matrix. The same checks run offline in research/there-is-no-white/verify.mjs (15/15).

live · recomputed in your browser from the CIE 1931 table + Bradford CAT
running…
what's exact, and what's a standard The geometry is exact: the spectral locus, the complementary pairs, the green gap, the metameric whites, and the adaptation are recomputed from the embedded data to the digit, offline and live. What's conventional is the machinery itself — the CIE 1931 2° observer is a standard average human, not a law of physics (real observers vary; the 1964 10° observer differs); D65, illuminant A, and the Bradford transform are agreed standards, not nature. In Instrument III the neutral paper is the exact, verified claim — a perfectly neutral surface reflects the lamp's spectrum, and the transform maps that onto the eye's white to ~1e-8. The three coloured swatches beside it are the von Kries model's prediction, shown to make the scene legible, not a measurement of specific pigments. "White is not a wavelength" is unconditional; the rest is a fact about a standardised eye, stated as exactly that.

V · What you can point at

So the next time the question comes up — is white a colour? is it all of them, or none? — you have a better answer than the two that cancel out. White is not a wavelength: no light is white. It is not all the colours: two will do, and the recipe is never unique. And it is not a fixed colour at all: it is the neutral your visual system is forever re-deriving from whatever light it finds itself in. You can watch each of those, above, and check the arithmetic yourself. White is the most ordinary colour there is, and it is the one that was never really out there.

VI · Sources & data

CIE 1931 2° standard-observer colour-matching functions (x̄, ȳ, z̄)The empirical curves underlying every chromaticity here. Digitised tables from the Colour & Vision Research Laboratory, UCL — the standard machine-readable form of the CIE 1931 standard observer. Same table as the sibling page.cvrl.ioo.ucl.ac.uk/cmfs.htm
CIE standard illuminants — D65, D50, A, and equal-energy EThe reference whites. D65 (0.31272, 0.32903) and D50 (daylight phases), illuminant A (0.44758, 0.40745; tungsten, ~2856 K), and equal-energy E at (1/3, 1/3). CIE 15:2004, Colorimetry.
Bradford chromatic-adaptation transform (CAT)The standard von-Kries-type transform used in Instrument III: XYZ → sharpened cone space, scale each channel by the white-point ratio, transform back. Matrix from S. Bianco & R. Schettini / Lam (1985), as standardised in ICC profile connection and widely tabulated.
sRGB / Rec. ITU-R BT.709 primariesThe "three lines" recipe and the on-screen rendering. R (0.640, 0.330), G (0.300, 0.600), B (0.150, 0.060), white D65. IEC 61966-2-1:1999; the linear-RGB↔XYZ matrices are the standard ones.
Simultaneous lightness contrastThe two-greys demo: a classic perceptual effect (a patch's apparent lightness depends on its surround). The patches here are byte-identical by construction — verifiable with any colour picker.
research/there-is-no-white/verify.mjsThe offline ground truth (15/15): white's interior position, the complementary pairs and the green gap, the three-line and metameric whites, and the Bradford adaptation mapping the lamp onto the eye's white. The page recomputes the same from the same data.