Skip to content
NatureHQ

Claim check

Can a mantis shrimp see more colours than we can?

Not supportedNo good evidence supports this, or the evidence points the other way.

The opposite, as far as anyone has been able to test it. Asked to tell two colours apart, mantis shrimps needed the wavelengths much further apart than a human does — they discriminate colour considerably worse than you do, despite having four times as many receptor classes.

The claim as it circulates

“The mantis shrimp has twelve types of colour receptor to our three, so it sees colours humans cannot imagine.”

Where you may have met it: Viral illustrated explainers about animal vision; Aquarium and museum interpretation panels; Lists of animals with extraordinary senses

Taking it apart

What was claimed
That the number of photoreceptor classes sets the richness of colour experience, and that an animal with twelve therefore perceives a colour world vastly larger than a human one with three.
What was actually observed
Mantis shrimps were trained to choose one wavelength over another for a food reward, then tested with the two wavelengths progressively closer together. They separated colours reliably only when the wavelengths were far apart — a threshold several times coarser than a human observer’s, and coarser than that of many animals with three or four receptor classes.
What the evidence supports
That colour discrimination is produced by comparing receptor outputs, not by having receptors. A visual system with three channels and a great deal of neural comparison separates wavelengths a few nanometres apart. The proposal for the mantis shrimp is that it does very little comparing and instead reads which of its twelve narrow channels fired — a fast, coarse scan of the spectrum, which for an animal that strikes in milliseconds may be the better trade.
What it does not support
It does not support ranking animal vision by receptor count, which is what the popular claim does and what a great many headlines still do. Nor does it license the reverse overstatement: this result is about discriminating wavelengths, and mantis shrimp eyes do other things extremely well, including reading polarised light and moving each eye independently. The narrow finding is that one common shortcut for judging a sensory system does not work.

How we know

Asking the animal with twelve colour receptors to tell two colours apart

Mantis shrimp have twelve classes of colour receptor to a human’s three. Does that mean they see finer colour differences?

Mantis shrimp were trained to strike a fibre-optic target glowing at one particular wavelength, rewarded when they hit it. They were then offered pairs of targets whose wavelengths differed by progressively smaller amounts, and the smallest difference each animal could still get right was measured. This is the standard test of colour discrimination and it has been run on humans, bees, birds and fish, so the numbers are directly comparable.

What happened

The shrimp could distinguish wavelengths about 15 to 25 nanometres apart and failed below that. Humans, with three receptor types, manage differences of a few nanometres across much of the spectrum.

What it shows

That counting receptors does not tell you how well an animal discriminates colour, and that the popular version of this animal has it backwards. Colour vision comes from *comparing* channels, and comparing twelve of them is expensive; the evidence here suggests the shrimp does not compare at all, but reads a wavelength off directly from which receptor it excites. Fewer distinctions, made faster — which suits an animal whose entire hunting strategy is one strike measured in milliseconds.

What it does not show

It does not establish the mechanism, which remains an interpretation of the threshold rather than a recording from the brain. It does not mean mantis shrimp vision is poor — the same eyes do things no vertebrate eye can, including reading circular polarisation. And it is one species, in a trained task, with few individuals.

The controls — what makes this evidence rather than a story
  • A trained rewarded task, so the animal is reporting a discrimination rather than a preference.
  • Wavelength pairs narrowed step by step to find the threshold rather than assuming one.
  • Target brightness controlled, so the animals cannot succeed on intensity.

From A different form of color vision in mantis shrimp

The rest of the answer

The claims underneath

Each one carries its own evidence, scope and caveats. Expand any of them to reach the studies.

Mantis shrimp have twelve classes of colour receptor and can tell colours apart less finely than a human with three. Counting receptors does not measure colour vision.

Well supported

Good evidence backs this, though some details remain open.

Wavelength discrimination thresholds in Haptosquilla trispinosa are approximately 15–25 nm, substantially coarser than human thresholds of a few nanometres, despite twelve photoreceptor classes. The evidence is consistent with wavelength recognition by receptor identity rather than by opponent comparison between channels.

Who this applies to
Measured in one stomatopod; the general point about receptor count applies wherever colour vision is discussed.Do not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Haptosquilla trispinosa
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

A direct behavioural threshold measurement using the standard method, giving a number comparable with those from other species.

How far it can be extended

Stomatopods are diverse and the discrimination threshold has been measured in very few of them.

Caveats

  • This does not make mantis shrimp vision poor: the same eyes detect linear and circular polarisation, which no vertebrate eye does.
  • The proposed mechanism is an interpretation of the threshold, not a recording from the brain.
  • One species, few individuals, in a trained task.

Still unanswered

  • What the twelve channels are actually for, if not fine colour discrimination.

Last reviewed 2026-09-02

The evidence (2 studies)

Pictures captioned "what a bee sees" are translations into human colour of something a human eye cannot receive. They can be accurate about the information and cannot be accurate about the experience.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Rendering a non-human visual scene requires mapping receptor excitations from one colour space into another with a different dimensionality and different primaries. The mapping is not unique, and no rendering can represent a percept whose receptor basis the viewer lacks.

Who this applies to
A methodological point about visual simulations, applying wherever a non-human sensory world is depicted.
Studied in
Animalia
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

A four-dimensional colour space cannot be displayed on a three-primary screen without loss, and the choice of what to lose is made by the person producing the image. That is arithmetic, not opinion.

How far it can be extended

The constraint is mathematical rather than biological: it holds for any mapping between colour spaces of different dimensionality.

Caveats

  • False-colour images are genuinely useful: showing where ultraviolet reflectance sits on a flower conveys real information, and the objection is to the caption rather than the picture.
  • The same limit applies to every modality, not only vision — there is no honest rendering of what a magnetic compass feels like either.

Still unanswered

  • Whether receptor-based models predict animal colour discrimination well enough to be used as ground truth, which they do only for a handful of well-tested species.

Last reviewed 2026-09-02

The evidence (3 studies)

Go deeper

Ultraviolet vision

Two birds identical to a human eye can be obviously different to each other. Put a filter over the window and the choosing stops.

Related things people ask

Last reviewed 2026-09-02