What your dog sees of you

The popular picture

Ask around and many people will tell you that dogs see in black and white. If that were true, this is all your dog would see.

What the tests found

It is not. In 1989 Neitz and colleagues trained three dogs to pick out coloured lights, and worked out from their choices that their eyes hold two kinds of cone, not the three we have: one most sensitive to violet-blue light, around 429 nm, one to yellow-green, around 555 nm. Four years later, recordings from the eyes of dogs and foxes found the same two.

So a dog sees colour, along one line instead of two: from blue, through grey, to yellow. This is you on that line.

And dogs use it. In 2013 Anna Kasparson and colleagues taught 8 dogs to find food at a yellow or a blue box, one dark and one light, then reversed which was dark and which was light. All of them leaned towards the colour rather than the brightness; for 7 of the 8 the preference was statistically clear.

Hold the button at the bottom (or press on the picture with a mouse) to see it with your own eyes, then let go.

Look at the bow

In the painting, the girl's red bow is the brightest thing in the garden, and the red flowers behind her burn against the green. Through the dog's eye they sink into the lawn.

This is the colour a dog is missing. Red and green differ for us because we have a third cone that tells them apart. For a dog, the whole range from green to red falls on the yellow side of the line, and a red bow and green grass differ mostly in brightness. In this painting both come out a dull yellow; the bow is only a little darker and a little more yellow than the grass.

That is the red ball thrown onto a lawn: bright for you, a yellowish shape on yellowish grass for your dog, told apart by a small difference in brightness and colour, and by its shape. A blue ball on the same lawn stays blue. We found no study that tested toys on grass; this follows from the two cones alone.

There is also one colour that a dog cannot tell from grey at all: a blue-green, around 480 nm in the tests. Here it comes out at 487.

How sharp

In 2017 Olle Lind and colleagues in Sweden tested dogs and people in the same set-up, with striped patterns at a distance. In bright light, the dogs saw stripes of 5.5 to 19.5 cycles per degree; the people, 32.1 to 44.2. A dog sees detail roughly three times coarser than you do.

Move the slider to step away. At 2 metres, the finest stripe a typical dog from that range can see (10.4 cycles per degree, the geometric middle) is about 1.7 mm wide; for you, about 0.5 mm.

Dogs differ a lot. In Lind's tests the best was a pug, at 19.5; the lowest scores were about 5.5. Choose one.

Up close, a dog's eye has its own limit: it probably cannot focus much nearer than a third to half a metre, according to a review of dog vision. The face pushed into your dog's nose is a blur.

Very dim light

Lind's team also turned the light right down, to 0.0087 candela per square metre, about five thousand times dimmer than the first test. The dogs' limit fell to 1.8 to 3.5 cycles per degree. The people's fell too, to 5.9 to 9.9, and they were still ahead.

At this light most vision runs on rods, which do not tell colours apart, so the colour is gone here too. Dogs' eyes are built for dim light, with far more rods than cones (Mowat and colleagues counted about 23 rods to every cone at the centre of a beagle's retina) and a mirror behind the retina. But the one head-to-head test found people still seeing finer detail.

You, in the same light

This is the same room at the same light, through your own eyes as Lind measured them.

1.5 metres

This is your head as your dog sees it from 1.5 metres, the distance used in the study that follows. Hair, the outline of the head against the wall, the dark and light of it: all there. The eyes and mouth are softer.

In 2013 Ludwig Huber and colleagues in Vienna taught dogs to walk up to one of two people and touch their face: the owner, or another person the dog knew well. The two knelt behind a screen, only their heads showing through two holes, about 1.5 metres from the dog.

Of 15 dogs that were trained, 14 learned to pick the right head. Then the real heads were replaced by life-sized photographs of them; 10 dogs managed that too.

Then they hid the hair

In the last stage the photographs got a digital balaclava. It covered the hair and the outline of the head, leaving eyebrows, eyes, nose, cheeks and mouth. Of the 10 dogs, 2 still picked the right person.

This is your face in balaclava mode. Put your face inside the oval.

The authors do not know which cues the dogs had used, but suggest that those who failed had been relying on the whole head: its colour, its brightness, the hairstyle. Two dogs managed every stage, so a dog can learn a face. In 2017 Paolo Mongillo and colleagues found the same limit with real people: dogs went to their owner more often than chance only when the outline of the head was visible.

Not every study agrees. A 2024 preprint on 416 free-ranging dogs reports that they used the inner features. And brain scans disagree about whether dogs have a region for faces at all: two small studies found one; a larger one, of 20 dogs, found none.

So this is you, for your dog: a figure on a blue-to-yellow line, softer than you know yourself, recognised in these experiments by the shape of your head more than by the face you see in the mirror. Dogs also connect your face to your voice: in 2007 Ikuma Adachi and colleagues found that dogs looked longer when their owner's voice was followed by a stranger's face.

Your dog has never seen your face the way your mirror shows it. It learned you anyway. Go and stand in the doorway, where it can see you.