How Do Animals See Colour? Dogs, Bees, Birds and the Mantis Shrimp

The retina at the back of an eye contains two kinds of light-sensing cell. Rods respond to dim light and see in shades of grey; cones work in bright light and come in types, each containing a pigment most sensitive to a particular band of wavelengths. A single cone type cannot tell colour from brightness, because a dim green light and a bright red one can excite it equally. Colour appears when the brain compares the output of two or more cone types: if the long-wavelength cones fire more than the medium ones, the light is reddish, and so on.

The number of cone types therefore sets how many dimensions of colour an animal can see. Two types give a colour space like a line, three give a plane, four give a volume. Each extra type lets the animal tell apart colours that looked identical before.

The ancestors of mammals spent the age of dinosaurs as small nocturnal animals, and somewhere in those hundred million years of night they lost two of the four cone types that other vertebrates keep. Most mammals still have only two, one sensitive to blue and one to yellow-green, which makes their colour vision like that of a red-green colour-blind person. A dog sees a red ball on green grass as two similar shades of yellow-brown, which is why so many dogs are slow to find red toys and quick to find blue ones.

Cats, horses, cattle and deer see the same way; the red cape at a bullfight is for the crowd, and the bull charges at the movement. Humans and the other Old World primates regained a third cone by duplicating a gene some thirty million years ago, most likely because telling ripe red fruit from green leaves was worth the cost, and the fact that the new pigment is tuned to exactly that distinction supports the idea.

Birds never lost their fourth cone, and it is tuned to ultraviolet. A bird's colour space has a whole dimension that ours lacks, and it uses it. Many species that look identical to us differ in ultraviolet plumage, so a male and female blue tit, which look alike to a birdwatcher, look plainly different to each other. Kestrels can see the ultraviolet reflected by the urine trails of voles in a field. Fruit and flowers advertise in ultraviolet as well as in visible colours.

Bees see a similar range shifted the other way: they have three cone types tuned to ultraviolet, blue and green, and cannot see red at all. Flowers have evolved to match. Many that look plain yellow to us carry ultraviolet patterns, bullseyes and landing stripes pointing to the nectar, that are invisible to us and vivid to a bee. A photograph of a dandelion under ultraviolet light shows a dark centre we never suspected.

The mantis shrimp, a small, spectacularly coloured predator of tropical reefs, has twelve types of colour receptor, plus others for polarised light, and for years it was assumed to have the richest colour vision on Earth. When researchers finally tested it in 2014, they found it could not distinguish colours that a human separates easily. The shrimp does not compare its receptors the way we do. Each one seems to act as a detector for its own narrow band, and the animal reads colour by which detector fires, like a bar code, rather than by mixing.

The reason may be speed. Comparing receptors takes brain processing, and the mantis shrimp, which strikes prey with the fastest punch in the animal kingdom, may have traded fine discrimination for a colour system that needs almost no computation. It is a reminder that more sensors do not mean better vision, only different vision. Colour vision across the animals:

• •Most mammals, including dogs and cats: two cone types, no red-green distinction
• •Humans and Old World monkeys: three, regained for finding fruit
• •Birds, many reptiles and fish: four, including ultraviolet
• •Honeybees: three, ultraviolet to green, blind to red
• •Mantis shrimp: twelve, read individually rather than compared

You cannot ask a bee what it sees, so the problem is designing a test where only colour could explain its behaviour. The classic solution came from the Austrian zoologist Karl von Frisch in the early twentieth century. He trained honeybees to feed from a dish of sugar water placed on a blue card, then set that card among cards in many shades of grey, from very light to very dark. If the bees were judging only brightness, at least one grey should have fooled them. They kept landing on the blue, which showed they were responding to hue itself.

The same experiment exposed their blind spot. When von Frisch used a red card instead, the trained bees confused it with dark grey and black, exactly what you would expect from an eye with no red-sensitive receptor. Modern researchers use variations of this logic with dogs, birds and fish: train an animal to choose a coloured target for a reward, then vary the brightness randomly so that only a true colour difference can guide it. They combine these behavioural tests with measurements of the pigments extracted from the animal's cones, and the two lines of evidence usually agree.

Human colour vision carries a trace of how it evolved. The genes for the red-sensitive and green-sensitive cone pigments sit side by side on the X chromosome, and they are so similar that they are easily mixed up when chromosomes swap pieces during the making of eggs and sperm, producing a missing or hybrid gene. A man has only one X chromosome, so a faulty copy leaves him red-green colour-blind. A woman has two, and the working copy usually covers for the other. That is why around one man in twelve of northern European descent has some red-green deficiency, against about one woman in two hundred.

Many monkeys of Central and South America show the same genetics in a more extreme form. They have a single pigment gene on the X chromosome, but it comes in several versions tuned to slightly different wavelengths. Every male, with one X, sees in two dimensions of colour, while a female who inherits two different versions gets three cone types and full trichromatic vision. Researchers have also looked for women with a fourth cone type, inherited the same way, and a very small number tested in the laboratory do seem to tell apart colours that others cannot, though such cases are rare and hard to confirm.

Octopuses, squid and cuttlefish pose a genuine puzzle. They can change the colour and pattern of their skin in a fraction of a second, matching rock, sand and weed so well that divers swim straight past them. Yet the eyes of most species contain only one type of light-sensitive pigment, which by the logic of comparing cones should leave them colour-blind. Behavioural tests agree. When cuttlefish have been placed on backgrounds whose patches differ in colour but are matched in brightness, they respond as if the patches were identical, though they react strongly to differences in brightness and contrast.

Several explanations are being argued over. One is that perfect colour matching is not needed: many predators judge mostly by brightness, pattern and texture, so getting those right may be enough. Another is that octopus skin contains light-sensitive proteins related to those in the eye, so the skin might sense light directly, although it is unclear whether it could tell colours apart. A third, more contested idea is that the unusual U-shaped and dumbbell-shaped pupils of some species could let them extract colour information from the way a lens focuses different wavelengths at slightly different depths. None of these has yet been confirmed.

Animals see colour by comparing the outputs of different cone types, and the number and tuning of those types decide what they can see. Dogs and most mammals have two and miss red, humans regained a third for fruit, birds and bees have ultraviolet, and the mantis shrimp shows that a dozen receptors read separately give less discrimination than three read together.