The Cuttlefish That Can’t See Colour

Cuttlefish are cephalopod molluscs — the group that holds octopus and squid. The species in every study behind this zine is Sepia officinalis, the common cuttlefish. The internal shell is the cuttlebone.

The skin patterns are made by chromatophores, pigment cells in the skin, along with other specialised skin cell types (Reiter, 2026).

The literature's word for the speed is rapidly. We have not printed a number, because none of the five papers here is a timing study, and a figure we could not source is worse than a comparison a child can picture.

The W-shaped pupil is not a flourish. It comes back on spread eight: an off-axis pupil is exactly what one of the four proposed explanations is built on.

Reiter (2026) on the hatchling point, in his own words: “Animals are capable of camouflage from hatching, without tutoring. This behavior therefore has a large innate component, ultimately encoded in the genome.”

The willingness to attempt camouflage on any substrate is what makes the animal testable at all, and it is the hinge of the experiment on spread five (Mäthger and colleagues, 2006, following Marshall & Messenger, 1996).

Sepia officinalis has a single visual pigment, with peak absorption near 492 nm. Colour vision requires at least two receptor classes with different spectral sensitivities, because the signal that carries colour is the comparison between them — a single class cannot distinguish a change of wavelength from a change of intensity.

One caveat worth having, because the zine's title rests on it. That pigment measurement traces to Brown & Brown (1958). Mäthger and colleagues note in 2006 that it is “the only study looking at the absorption of S. officinalis visual pigment,” with no modern microspectrophotometry since; Bellingham and colleagues (1998) modelled the same peak from the opsin sequence and agreed. So the anatomy rests on one old measurement and one model — which is why the behavioural test on the next spread matters so much.

And one cephalopod is a known exception. The deep-sea, bioluminescent firefly squid Watasenia scintillans expresses three photopigments with distinct peak sensitivities (Reiter, 2026). “Cephalopods are colour-blind” is a statement about most of them, not all.

Mäthger, Barbosa, Miner & Hanlon (2006). Sixteen checkerboards pairing sixteen grey shades with one green matched to the animal's 492 nm peak; then a blue-and-yellow board with intensities matched to the cuttlefish visual system. On the intensity-matched boards the animals produced non-disruptive body patterns, “suggesting that the substrates appeared to their eyes as uniform backgrounds.”

Their conclusion, quoted: “we show clearly that cuttlefish must be color blind.” The same assay showed the animals resolve contrast differences of about 15%, so the null is not a story about an animal that cannot see.

The design is the point of this spread. The authors call the body pattern “a quantifiable, neurally controlled motor response” — the animal's own body is the readout. No question was asked and none could have been. Spread eleven comes back to this.

Chiao, Wickiser, Allen, Genter & Hanlon (2011). Hyperspectral imaging records the full reflectance spectrum at every pixel rather than three channels, so the comparison does not inherit our own colour vision. Live camouflaged cuttlefish were imaged on natural substrates; most reflectance spectra of animal and background were similar.

They then modelled the vision of di- and trichromatic fish predators. The just-noticeable differences between animal and background were “relatively small and distributed randomly, an indication of good color match.” Pattern matched too — and largely in the predator's luminance channel rather than its chromatic ones.

Why this spread exists: without it, “the cuttlefish matches” would be our impression of a photograph. The paradox is only a paradox because both halves were measured.

These are two different experiments, and it is very easy to fuse them. The checkerboard was an artificial substrate in a tank, and the only observer whose vision it probed was the cuttlefish’s — no fish was involved at any point. This spread is natural substrates in the sea, and the vision being modelled is the predator’s. One asks what the animal can see; the other asks whether what it does works on the animal it is hiding from. The board and the sea floor are not the same scene, and a reader who carries the checkerboard into this spread will end up asking why fish cannot see squares that we can — a question neither paper poses.

Chiao and colleagues report it themselves, in the same paper this zine cites for the good match: “Curiously, most reflectance spectra of cuttlefish had a peak around 800 nm in the infrared range (IR), whereas natural substrates tested did not have this spectral characteristic.”

It is printed here because it is a measured imperfection inside the very result we are leaning on, and leaving it out would be the tidier version of the paper rather than the paper.

What we checked and what we did not. Vertebrate visual pigments are not known to extend to 800 nm, so the bump is not thought to be visible to the fish these models cover. We have not found a study that asks whether any cuttlefish predator can detect it — so “it does not matter” is an inference from the absence of such a sense, not a tested result. If you want a harder answer than that, you will have to go further than we did.

Reiter (2026) works through exactly these four and finds each “either insufficient or weakly supported by current data.”

One — passive skin reflectance. Colour matching is needed to avoid detection by predator and prey visual systems, and wavelength filtering does not remove that need.

Two — chromatic aberration. Stubbs & Stubbs (2016) modelled it: an off-axis pupil plus wavelength-dependent focus could let a single-receptor animal infer spectra by scanning through focus for the sharpest image. It is a computer model, and Reiter notes “there have been no physiological tests of this proposal”; it has also been challenged on ecological grounds, since most objects a cephalopod meets have broadband spectra, and turbidity and wavelength-dependent attenuation blur the chromatic focus further.

Three — polarized light. Cephalopods genuinely detect linear polarization. Using it for colour would require sensitivity to small changes in the degree of polarization arising from an object's material or spectral properties, which is a much stronger claim.

Four — light-sensing skin. Rhodopsin and retinochrome really are present outside the eyes, including within chromatophore organs (Kingston, Wardill, Hanlon & Cronin, 2015). But Reiter's assessment is that “extraocular photoreceptors appear better placed to mediate behavioral responses to global luminance changes, rather than color vision.”

Four live hypotheses, none carried over the line. That is the honest state of it in 2026.

This is Reiter's null hypothesis, and his framing of it: “it is necessary to take seriously the alternative possibility that accurate color matching can be achieved without color perception.”

The parallel he draws is to image-colourisation networks, which learn to predict an image's chromatic channels from its luminance channel alone. They work because natural scenes are statistically regular: luminance edges tend to coincide with colour boundaries, textures constrain likely reflectance, and scene structure constrains plausible colour. Crucially, such networks “do not recover the ‘true’ colors of objects in any physically unique sense; rather, they generate the most probable chromatic interpretation given a luminance pattern.”

On that reading a cephalopod is the generator and its predators are the discriminator, judging the output with the colour vision the cephalopod lacks.

This is a hypothesis in a review, not a result. It is the newest idea here and the least tested of all of them — the grass example is ours, not the paper's, and it is a way in rather than evidence.

Every paper behind this zine states the paradox as unresolved. Mäthger and colleagues close their own 2006 paper on it: “the question of how cuttlefish achieve ‘color-blind camouflage’ in chromatically rich environments still remains.”

Twenty years later Reiter's review arrives at the same place — having ruled more out, and with a better idea in hand. A question that stays open for twenty years while the evidence around it sharpens is what a healthy question looks like, and it is worth letting a child meet one without alarm.

This is a point about method, and it is deliberately not a claim about people drawn from a cuttlefish. The assay works because S. officinalis will attempt camouflage on any substrate it is placed on, which turns the animal's own body into what Mäthger and colleagues call “a quantifiable, neurally controlled motor response.” The experimenters built the answer channel; the animal did the rest.

Notice what nobody in this literature concluded. Every null here is reported as a fact about the reach of the test — not as an absence in the animal. Where that courtesy is not extended, Bone Song has the argument at length, and it is the argument this spread is standing next to rather than restating.

A rhyme, not a proof. Nothing about a cuttlefish establishes anything about a person. What transfers is the shape of a habit: build the channel before you conclude there is no signal.