I like colours, but probably not for the reason you think. Colours present themselves as beauty, decoration, the subjective end of experience — and yet underneath that presentation, colour is a fully solved interface between three disciplines that rarely cooperate this cleanly. A colour is a physics result: a spectral power distribution, the intensity of light at every wavelength an object returns to the world. It is a biology result: three cone cells, each with a sensitivity curve, integrating that spectrum into a triple of responses. And it is a mathematics result: a projection — an infinite-dimensional spectrum compressed, without ceremony, into three numbers. When you say a wall is blue, you are reading out the output of a measurement pipeline that evolution engineered and that mathematics later reverse-engineered with such precision that we can now reproduce it with three dots of phosphor.
The pipeline hiding inside a colour
The full chain deserves a closer look, because every link in it is a design decision the body made before we knew design was involved. Photons arrive carrying a continuous spectrum; the eye does not record the spectrum — it cannot afford to — so it integrates it against three fixed sensitivity curves and keeps only the total excitation of each cone type. This is metamerism: two lights whose spectra differ at every single wavelength can produce identical cone responses and therefore identical perceived colours, all their physical difference discarded on the way in. What the brain receives is not the world’s light but a three-number summary of it, and everything the summary discards is unrecoverable. Early in the twentieth century, experimenters went further and standardized this living, varying instrument outright: the standard observer, the statistically averaged sensitivity curves of the majority, adopted as the canonical sensor — the single act of turning a population into one instrument on which the entire edifice of colour science, from the CIE spaces under your display’s hexcodes to the horseshoe diagram in every colorimetry textbook, still rests.
The question the eye was built to answer
Notice what this pipeline actually optimizes for. The eye’s compression is tuned to one deliverable: preserving object identity under changing light — fruit that looks like the same fruit at dawn and at noon, a face that reads as the same face under tungsten and overcast sky. Chromatic adaptation is the second compression stage doing exactly this work: the brain infers the illuminant, renormalizes against an assumed white point, and hands consciousness a colour corrected for the light it arrived in. It is an astonishing engineering feat, and it is also deliberately, structurally blind. Everything the visual system discards — the fine spectral detail, the absorption signatures of specific molecules suspended in tissue — is discarded by design, because for the purpose the eye serves, object constancy is worth more than spectral truth; and it raises a question the eye itself has no way to ask. The eye answers “what is this?” brilliantly — but how much of the body’s own activity is riding inside the light it sends back, invisible not because we lack the photons but because we asked the wrong question of them?
Asking the mirror a different question
This is where the inversion happens. Photoplethysmography — PPG — uses the identical physical substrate and the identical biological target, with the objective function reversed. Instead of asking “what colour is this tissue,” it asks “what did the tissue do to the light between sending and returning it.” When green light at roughly 525 nanometers is shone into skin, haemoglobin absorbs a measurable share of it, and that share rises and falls with each pulse of arterial blood; the returning light is a waveform, and the waveform is the heartbeat. Where the eye compresses toward object constancy, PPG decompresses toward physiology: the wavelength that best survives skin’s own absorption landscape becomes the carrier, red and infrared pair up to triangulate oxygen saturation from the differential absorption of two haemoglobin states, and the sensor’s emitter choices mirror — structurally, almost rhyme for rhyme — the cone triple that perception itself runs on. At Holosense the whole premise is contained in that mirror: evolution built one instrument from light to see the world, and the same physics, pointed backward, becomes an instrument to see the body. Non-invasive sensing works at all because biology is not hidden from light — it is written in light, in absorption spectra, and the sensor’s job is transcription.
The gamut of what light can reach
But the body, like every display ever built, has a gamut. A display’s gamut is the triangle its three primaries enclose — the region no increase in bit depth can exceed, which is why millions of hexcode values still tile only a slice of perceivable colour. Optical biosensing runs on the same architecture: the wavelengths that penetrate skin, survive melanin’s absorption, and return with enough signal-to-noise define a convex region of what is physically reachable, and every biomarker claim a device makes lives inside that triangle or not at all. I call this the physiological gamut — the honest boundary of a sensor’s reachable set, drawn by physics and skin before marketing is consulted. The precision-versus-range confusion that hexcodes taught us recurs here with clinical stakes attached: a device can resolve its pulse waveform to exquisite precision and still have no access whatsoever to continuous glucose, because glucose sits outside the optical window its LEDs enclose.
The colour sciences also hand over a warning. The standard observer quantified colour by averaging people, and eighty years later the approach still fails outliers — pulse oximeters run measurably worse on darker skin because melanin shifts the reflectance curve the device’s assumptions encode. A biosensor built for prevention cannot inherit an aggregate; the individual baseline is the signal, and calibration must run to the person, not the population — the same body-as-substrate logic I developed in The Body Is the Last Unvirtualized Interface, where the physical conditions of the person refuse every abstraction layer built above them. Getting this right turns a light sensor into something closer to what I sketched in What If Your Best Self Became Visible? — a mirror that reflects your own physiology back to you with your own baselines as the reference white.
Transcription, not magic
The crystallization is this: three disciplines that normally operate in isolation — physics describing the field, biology receiving it, mathematics compressing the exchange — already completed one full collaboration, and we stare at its output every waking hour without recognizing it as an engineered system. Colour perception was biology’s prototype of light-based measurement; colorimetry was mathematics recovering the blueprint; and photonic sensing is the same blueprint, deliberately recompiled with a different objective function. What was once an accident of evolution becomes a design space, and the design space has rules — spectra, absorption, gamuts, calibration — that reward honesty long before they reward ambition.
Stack Takeaway
- Colour perception is a functioning physics-to-biology-to-mathematics measurement pipeline that predates its own formalization — colorimetry reverse-engineered what evolution shipped, and the reverse-engineering is what makes every display possible.
- PPG works because perception’s compression is reversible at the source: the physiological information perception discards as noise remains present in the returned light, and a sensor is simply an instrument that asks the photons a different question.
- Every non-invasive biosensor has a physiological gamut set by emitter wavelengths and tissue optics, and “more precision inside the gamut” is routinely mistaken for “more range” — the hexcode fallacy with clinical stakes; honest design draws the triangle first and calibrates to the individual, not the aggregate.