The Shape of LightAn interactive exhibition about light

The fourth room · Behind the eyeChapter 18 of 21

Computed: a page as one glance delivers it, if the page fills thirty degrees of your view and you are looking at its centre. Blur stands in here for a loss that is not quite blur.

You seem to see a whole scene at once, sharp from edge to edge, with no gaps. Almost none of that is in what the eye sends. The retina is sharp in one small place, blind in another, and reports mostly what has changed.

Chapter 18 · Behind the eye

The Edited Picture

What the retina leaves out before anything reaches the brain: most of the detail, one whole patch of the scene, and every surface that is merely even.

Look first

A hundred receptors for every fibre

The image of chapter 6 lands on the retina’s receptors: 4.6 million cones and 92 million rods in an average eye. The only way out of the eye is the optic nerve, and the optic nerve has about a million fibres.

So the picture cannot simply be passed along. A hundred receptors must share each fibre, and something inside the eye has to decide what is worth sending. The retina is not a sensor with a cable attached. It is the first piece of the brain, and it has already begun to edit.

schematic
To scale in number only: 97 receptors, 5 of them cones, sharing one fibre of the optic nerve.

Commit to a prediction

How much is sharp?

Hold your thumb out at arm’s length and look at the nail. You can see the fine ridges in it. You can also see the room around it, and the room looks as clear as the nail.

At this moment, without moving your eyes, how much of the scene are you seeing as sharply as the nail?

Why the first guesses made sense

Wherever you look, it is sharp

The room looks sharp all over because you cannot catch it being otherwise. The moment you wonder about the corner, your eyes have jumped there, and the corner is on the fovea. Your eyes make several such jumps every second, and you notice almost none of them.

So the sharp, wide scene is not a thing you see at any instant. It is something kept, and patched together from one small sharp sample after another.

Try it with a mirror

Stand close to a mirror. Look at your left eye, then your right, then your left. You will never see your eyes move, though anyone watching you can. Benno Erdmann and Raymond Dodge noticed this in 1898. During each jump, seeing is turned down, and the gap is closed over afterwards.

your eyes

A case you have not seen

The hole in each eye

The million fibres have to leave the eye somewhere. They gather at one place and pass out through the retina together, and at that place there are no receptors at all. It lies about 15 degrees to the outer side of where you are looking, and it is 5.5 degrees wide: more than ten full Moons, side by side, would fit across it.

Close one eye. What do you see in that part of the scene?

The rule

The retina reports differences

What does one fibre of the optic nerve say? In 1953 Stephen Kuffler put a fine electrode beside single cells in a cat’s retina and tried small spots of light. A cell would fire briskly for a spot in one small place. A spot just beside that place did the opposite: it quieted the cell. Light over the whole region, centre and surround together, did very little.

So the fibre does not report how much light there is. It reports how much more light there is in one place than around it. An even surface, however bright, is hardly news. An edge is.

calculated
A model of the cell Kuffler found, and its response to a spot of light as the spot grows.

One more case

A staircase of flat greys

Paint five broad stripes side by side, each a perfectly even grey, each a little lighter than the one to its left. Every stripe is flat: the same value in every pixel from one edge to the other.

What does each stripe look like?

The bench

One glance

The page below is drawn as a single glance would deliver it. Move the point of gaze with your finger, your mouse or the arrow keys.

Things to try

  1. Rest the gaze on one word. How many words either side could be read?
  2. Sweep the gaze along a line, as in reading. Is there any moment when the whole page is clear?
  3. Switch to the stripes and look at the real thing, with your own eyes. Where do the stripes fade into grey: along a level line, or along a curve?

What you can now say

The eye does not send the brain a picture. A hundred receptors share each fibre of the optic nerve. Only the centre of gaze is sharp, one patch of each eye is blind, and the fibres report differences between a place and its surroundings. The wide, sharp, seamless scene is put together afterwards.

You have now seen two things with your own eyes that were not on the screen: shading in flat stripes, and a line across a gap.

For the determined reader: the laws, the model cell and the limits

Sharpness across the field

With E the angle from the centre of gaze, the finest detail that can be resolved follows a straight line:

size(E) = size(0) × (1 + E / E2)

For fine stripes E2 is about 2.5 degrees. For other tasks it is different: about 1 degree for telling which way a broken ring faces, and under 1 degree for judging whether two lines are in step. The page uses the value for stripes. The picture of a glance is blurred so that a printed letter becomes unreadable where Anstis found that letters of its size do.

The model cell

The cell adds up light with a bell-shaped weighting over its centre and subtracts a share of the light under a wider bell. Its responses to an edge, a spot and a set of stripes then have exact forms, and the tests repeat each of them by direct numerical summation. The widths and the share are chosen for the drawing. In a real retina the cells are smallest at the fovea and grow outward, some are excited by light at their centre and others by darkness there, and there are many more kinds than this one.

Where this stops

Blur is the nearest a picture can come to showing the edge of vision, and it is misleading. Away from the centre, what is lost first is the order and position of things, not their focus, as the three letters showed. And the model cell explains the fluted stripes too neatly. The same fluting changes with how the stripes are arranged in ways no single cell in the eye accounts for, and some of it is certainly made in the brain.

The edge of a slide

In 1931 Keffer Hartline and Clarence Graham teased a single fibre out of the optic nerve of a horseshoe crab and recorded what it carried. Hartline called the crab “a fortunate choice of experimental animal”: its eye is coarse, its nerve is long, and its fibres can be split apart by hand. Brighter light gave faster firing. Light on the neighbours slowed it.

Kuffler’s cat came in 1953, and in 1958 two young men in his laboratory, David Hubel and Torsten Wiesel, moved the electrode from the eye to the visual cortex at the back of the brain. For a month the cells there ignored every spot of light they were shown. Then, during one long recording, a cell that had been silent for hours suddenly, in Hubel’s words, “went off like a machine gun”.

It was not the spot. They had been sliding a glass slide into the projector, and its edge had cast a faint straight shadow across the retina. “That was what the cell wanted, and it wanted it, moreover, in just one narrow range of orientations.” The cortex takes the retina’s spots of difference and assembles them into lines and edges at particular angles.

Hartline shared the Nobel Prize of 1967 with Wald and Granit. Hubel and Wiesel received theirs in 1981, “for their discoveries concerning information processing in the visual system”.

Where this explanation stops

There was never a picture

It is natural to imagine the eye as a camera and the brain as someone looking at the photograph. This chapter leaves nothing for them to look at. What travels up the optic nerve is a million streams of differences, sharp in one spot and coarse elsewhere, with a patch missing and the even parts hardly mentioned.

From that, something builds the room you are sitting in. It fills the hole, flattens the stripes it has no reason to doubt, and keeps the whole scene steady while the eyes jump. Nobody has yet found the place where the finished picture is.

End of chapter 18

Added to your notebook
The eye does not send the brain a picture. A hundred receptors share each fibre of the optic nerve. Only the centre of gaze is sharp, one patch of each eye is blind, and the fibres report differences between a place and its surroundings. The wide, sharp, seamless scene is put together afterwards.

Signals this sparse could have come from many different worlds. How does the brain settle on one?

Chapter 19 · The Best Guess