Scientists Gave People a New Color to See

Somewhere in the space of all possible colors, there is a region that no human eye has ever reached. It lies just outside the boundary of the familiar “horseshoe” — the CIE chromaticity diagram that maps every hue a human can normally perceive. No display screen, no paint, no laser beam seen under ordinary conditions can produce it. The physics of color vision simply forbids it.

Until now.

On April 18, 2025, a team of researchers at the University of California, Berkeley and the University of Washington published a study in Science Advances describing a device that crossed that boundary. They named the new color “olo.” Their five human subjects described it as “blue-green of unprecedented saturation” — a teal so vivid, so ludicrously pure, that no ordinary source of light can produce anything close.

The device is called Oz.

Why Color Has a Boundary

To appreciate what happened, you need to understand why human color vision has limits at all.

Your eye has three kinds of cone photoreceptors: S-cones (sensitive to short, violet-blue wavelengths around 420 nm), M-cones (middle wavelengths, peaking around 530 nm), and L-cones (longer wavelengths, peaking around 560 nm). When light enters your eye, every photon that lands on the retina stimulates all three types to varying degrees, depending on its wavelength. A green-ish 543-nm laser stimulates your M-cones quite strongly, your L-cones somewhat, and your S-cones barely at all.

Your brain doesn’t perceive the wavelength of light directly. It perceives a ratio — a comparison of how much each cone type is excited relative to the others. This is why you can mix red, green, and blue light to make any color: you’re not replicating the wavelength, you’re replicating the cone-activation ratios.

Here’s the constraint: every physical light source stimulates all three cone types simultaneously, in proportions set by physics. The spectral response curves of the three cone types overlap substantially. No natural light can stimulate only the M-cones while leaving the L-cones and S-cones completely dark. If you want pure M-cone activation, with zero L and zero S, you’re asking for a color signal the brain has never received — because no light in the natural world can produce it.

That signal would be the color “olo.”

In the CIE chromaticity diagram — the standard horseshoe map of human color perception — every color you’ve ever seen in your life falls inside or on the boundary of that horseshoe. Theoretical colors outside the boundary exist in color-science mathematics but have never been experienced by human eyes. Olo would lie outside the horseshoe, in territory no eye has visited.

The Oz Device

The Berkeley team built a system they call Oz — a name borrowed from the Wizard of Oz, because it displays colors that aren’t supposed to be there.

Oz is built on a technology called adaptive optics scanning light ophthalmoscopy (AOSLO). AOSLO was originally developed to image the retina at cellular resolution — it uses deformable mirrors to compensate for the optical imperfections of the eye, achieving near-diffraction-limited imaging where individual cone cells are visible. James Fong, senior author Austin Roorda (a pioneer of adaptive optics retinal imaging), and their collaborators turned this imaging technology into a display.

The key innovation is cell-by-cell light delivery. Oz first creates a detailed map of the subject’s retinal cone mosaic — a classified atlas of every cone cell in a region near the fovea, with each cell tagged as S, M, or L type. This classification uses a technique called optoretinography and produces maps of typically 1,000 to 2,000 cones in a region about 1.8 degrees of visual angle across. The cone cells, packed like tiles in a mosaic, are roughly 1.5–1.8 arc-minutes apart near the 4-degree eccentricity where the experiments were conducted.

Then Oz begins to display imagery. It sweeps a laser beam across the retina at high speed in a raster pattern — just like an old CRT television, but aimed at your photoreceptors. Crucially, it modulates the intensity of each laser pulse in real time to hit each individual cone with precisely the intended dose. The system delivers about 100,000 laser microdoses per second per cone cell — intense, brief flashes of infrared light for imaging and visible-wavelength pulses for stimulation. Infrared eye tracking running at 60 frames per second monitors the eye’s inevitable tiny movements (fixational eye movements), correcting the targeting continuously so that each microdose lands on the intended cone even as the eye drifts.

The field of view is modest — a 0.9-degree square, centered 4 degrees from the gaze fixation target, falling on the classified region of the retinal map. But within that small patch of visual space, Oz has complete control over the activation level of each individual photoreceptor.

Seeing “Olo”

To demonstrate that Oz can produce colors outside the normal human gamut, the researchers did a formal color matching experiment — the gold-standard psychophysical method used to map color space since the 1920s.

In a color matching trial, the subject sees two colored squares side by side — one produced by Oz (trying to stimulate only M-cones), the other produced by conventional projector light. The subject adjusts the projector light until the two squares match as closely as possible. When the subject needs to add white light to the Oz color before they can match it, that’s mathematically equivalent to saying the color lies outside the gamut — only colors outside the gamut require negative components to be represented in standard color space.

Every single subject had to add white light to match olo. Not a little white — enough to pull the color firmly inside the gamut. The matched wavelengths, once the white was removed, fell between 501 and 512 nm — the most saturated teal-green hues that can be produced by ordinary light. But even those couldn’t match the raw purity of olo without dilution.

Subjects were also asked to describe what they saw. The color names they volunteered: “teal,” “green,” “blue-greenish,” “green, a little blue.” And saturation ratings on a 1-to-4 scale: olo consistently scored 4 out of 4 — as saturated as subjects could imagine — while the closest comparable colors produced by ordinary light averaged only 2.9 out of 4.

In total, five subjects participated in 222 color-matching trials. The measurements confirmed that Oz successfully produces colors lying outside the natural human gamut, forming a triangular region in color space (for 488 nm stimulation) and a line extending beyond the gamut boundary (for 543 nm stimulation).

The researchers then went further, demonstrating that subjects could perceive images and videos rendered in Oz colors. They ran image-recognition tasks: a four-alternative forced-choice test where subjects had to identify the orientation of a line, and a two-alternative forced-choice test where subjects had to detect the rotation direction of a moving dot — all rendered as red (pure L-cone) elements on an olo (pure M-cone) background. Subjects performed well in the experimental condition but dropped to chance-level guessing in the control condition, where the cone-targeting was deliberately scrambled.

The color olo wasn’t just a point of light. It was a functional, perceivable, image-forming color.

What It Feels Like (And Why That’s Hard to Explain)

There’s an obvious question here, and it deserves an honest answer: what does olo actually look like?

The honest answer is: there is no way to show you. By definition, no screen or printed image can display it. Any attempt to approximate it — a vivid teal, a hypersaturated cyan — inevitably falls short, because those approximations are produced by ordinary light and therefore lie inside the horseshoe. Olo is outside.

The closest analogy might be imagining a color of depth and purity that you’ve simply never encountered. Subjects consistently described it as more saturated than any color they could match — a kind of teal-green with the dial turned past 10. But even this description sells it short, because “more saturated” sounds like a difference of degree, when the reality is a difference of kind. Olo is a color signal the visual system has never received, carrying information from the M-cone channel with none of the cross-contamination from L and S cones that normally accompanies it. The brain, encountering this novel input, does something with it — and what subjects report is a color of almost aggressive vividness.

One small caveat: the prototype system achieves a partial expansion of color space, not a perfect one. Light aimed at M-cones inevitably leaks a little onto neighboring L and S cones — the cones are physically very close together, separated by about 1.5–1.8 arc-minutes of visual angle at 4 degrees. This “fractional leak” limits how pure the M-cone signal is in practice. Olo, as currently experienced, is a partial journey toward the theoretical extreme rather than the full destination. A perfect implementation — 100% pure M-cone activation with zero leak — would presumably be even more extreme.

Why This Matters

This is, first of all, a remarkable demonstration of precise biological control. The ability to classify individual photoreceptors in a living human eye, track their locations in real time as the eye moves, and then precisely dose them with light at cellular resolution is a genuine technical tour de force. Adaptive optics retinal imaging has been developing for decades; the Berkeley team has now turned it into a display technology.

The potential applications are genuinely interesting. One natural direction is vision science — Oz is an extremely powerful tool for probing how the visual system processes color at its lowest level. Questions that have been theoretically accessible but experimentally difficult (what exactly does the brain make of pure single-cone signals? how does adaptation work at the photoreceptor level?) are now experimentally tractable.

Another direction is therapeutic. If Oz can precisely control which photoreceptors are activated and by how much, it might eventually be possible to use it as a form of prosthetic vision — delivering visual information directly to remaining functional cones in people with retinal degeneration, bypassing damaged rods or non-functional cone subtypes.

And there’s the deeper, more philosophical angle. Humans have been constrained to the same color gamut since before Homo sapiens existed. The physics of light, the spectral overlap of the cone types, and the optics of the eye together defined a boundary that no experience ever crossed. This work crossed it, deliberately and measurably. Whether or not the experience of olo tells us something deep about consciousness or perception, the mere fact of its existence is genuinely remarkable.

Somewhere outside the familiar horseshoe, there is now a color. It has a name. Five people have seen it.

---

Paper: Fong J, Doyle HK, Wang C, et al. “Novel color via stimulation of individual photoreceptors at population scale.” Science Advances 11(16): eadu1052 (2025). DOI: 10.1126/sciadv.adu1052

Hero image: CIE 1931 chromaticity diagram comparing color gamuts, by BenRG and cmglee. Wikimedia Commons. CC BY-SA 3.0.