Stand a color chart under a fluorescent tube and under sunlight, and tune the fluorescent spectrum until the chart's red patch reads the same RGB triplet under both. Your eyes agree: same red. Your camera agrees: same numbers. Load film behind each light and you get two different negatives, because the film never saw a color. It saw a spectrum, one photon energy at a time, and the two lights only match on the three-number summary your eye and your sensor both settle for.
That's metamerism: two spectral power distributions that differ, tuned to produce the same tristimulus response in a three-receptor system. Human vision has three cone types, so any two spectra that stimulate them identically look identical, no matter how different the underlying light actually is. A camera sensor is the same kind of device with three color filters instead of three cone pigments. Both are metamers of each other by construction — they're built to collapse a spectrum into three numbers, and any information that distinction required is gone the moment they do.
Where the Match Breaks
Film's color layers don't collapse anything. A color negative stacks three emulsion layers, each sensitized to a different band, and each layer's exposure depends on the actual energy arriving in its band, not on how a human viewer would later describe the color. Two lights that read identically to your eye can still hand the film's red layer two different exposures, if the energy in that specific band differs even while the three-receptor summary matches.
This is exactly the gap a LUT can't close. A 3D LUT is a function of RGB in to RGB out. It receives the same three numbers your camera or your eye would have produced under either light, because that's all a camera ever hands it. There's no fourth channel carrying "which spectrum was this." If two exposures are metameric at the sensor, they arrive at the LUT as the same input, and the LUT — having no way to tell them apart — returns the same output for both. Whatever gap existed at the film's red layer never had a chance to be corrected, because the data that would show it was discarded upstream, before the LUT ever ran.
Why This Isn't the Same as a White-Balance Miss
It's tempting to file this under color temperature and move on, but a white-balance error and a metameric mismatch are different failures. White balance corrects a known, single-number shift — the whole image reads too warm or too cool, and one gain adjustment per channel fixes it. Metamerism doesn't shift the whole image by a known amount. It's two specific spectra that happen to coincide in RGB while diverging elsewhere in the visible range, and the divergence can hit different objects differently depending on what each object reflects. A skin tone and a green leaf don't share a reflectance curve, so the same metameric light pair can leave one looking right and the other visibly wrong, in the same frame, under the same white balance setting.
That's the failure mode worth naming: not "the whole shot is a bit warm," but "this specific LUT, trained on one light's spectrum, quietly stops matching the film's actual layer response the moment the spectrum changes underneath an unchanged RGB reading."
What Would Actually Close the Gap
Closing it means keeping the spectral information a trichromatic sensor throws away. A multispectral or hyperspectral capture samples many more bands than three, so it can distinguish two lights that a normal sensor would report as identical. Spectral rendering pipelines in visual effects work this way for exactly this reason — they carry the full curve as far into the pipeline as they can afford to, because collapsing to RGB early means collapsing permanently.
For anyone not shooting multispectral, the practical fix isn't a better LUT. It's controlling or knowing the light a LUT was built from, and re-deriving the mapping when the light changes — which is a different problem than the one a single, static 3D LUT was ever built to solve. Our piece on why film LUTs break with exposure covers the sibling failure, where the mismatch comes from density instead of spectrum. You can see how sensitive the mapping actually is in the Cineon studio.
The three numbers your camera gives you were never the light. They were always a summary, agreed upon by eye and sensor alike, and a LUT can only ever work with the summary it's handed.