The Shape of LightAn interactive exhibition about light

The fourth room · Behind the eyeChapter 21 of 21

A still from the test below: a grid of crosses that will turn, and three yellow dots that will not move, change or dim.

The last chapter found seeing where there was no light. This one looks for the reverse: light that reaches the eye, is caught by the cones, is signalled to the brain, and is not seen.

Chapter 21 · Behind the eye

Light Without Seeing

Dots that vanish while they stay where they are, a glimpse that is gone before it can be read, and a gorilla that half of an audience never saw.

Commit to a prediction

Dots that stay put

Three bright yellow dots sit on a dark screen. They do not move, flicker or fade: the same pixels are lit the whole time. Behind them a grid of blue crosses turns slowly. You look steadily at a point in the middle.

What happens to the yellow dots?

Why the first guesses made sense

In plain view is not the same as seen

If you said nothing would happen, you were relying on the plainest rule there is: if a thing is in front of my open eyes, in good light, I see it. Nothing in ordinary life contradicts it, because you can never notice the things you are not noticing.

The rule is false. Catching the light is the first step of several, and a signal can fail at any of the later ones.

A case you have not seen

Twelve letters

In 1960 George Sperling showed people rows of letters, as many as twelve at once, for a twentieth of a second, and asked them to write down all they had seen. Of twelve they managed three to five: about a third.

How much of the display was really available to them at the moment it vanished?

One more case

The gorilla

In 1999 Daniel Simons and Christopher Chabris showed people a short film of two teams passing basketballs, and asked them to count the passes made by one team. During the film a person in a gorilla suit walked through the scene, in full view.

Afterwards, what share of the viewers said they had seen anything unusual?

The bench

Something will change

Five things on a shelf. When you press Start, one of them will change, steadily, over fifteen seconds, in full view. Then you will be asked which.

Things to try

  1. Run it once and answer. Were you sure?
  2. Play the same change again at speed. Could you have missed it then?
  3. Run it again while counting backwards from two hundred in sevens. What is different?

What you can now say

Light reaching the eye is not yet seeing. A thing in plain view can vanish while its light still arrives. A glimpse holds about twice what can be reported, and the extra is gone within a second. And much of what happens in front of us goes unseen while we attend to something else. Why any of it is experienced at all, nobody knows.

The picture that reaches awareness is a small selection, and it comes with no mark to show what was left out.

For the determined reader: the sums, the timing and the limits

Sperling’s sum

If a person can give c letters of a row of r, chosen at random after the display has gone, then the share c / r of the whole display was available. The curve in the figure falls from Sperling’s 75% toward 35% with a time constant of 0.3 seconds, which is this page’s choice to fit his finding that the advantage is gone after about a second. The account read here gives his 35% for displays of twelve letters and his 75% for displays of nine. This page shows twelve, which he also used, and compares shares, not counts.

Timing on your screen

A screen redraws a fixed number of times a second, so the letters are shown for a whole number of frames: 3 frames on a screen that redraws 60 times a second, which is 50 milliseconds. The page measures your screen’s rate and tells you the time it achieved. A browser is not a laboratory, and your figures will be rougher than Sperling’s. He named the row with a tone, where this page uses an arrow. And this page counts a letter as right wherever in the row it stood, which flatters you a little: someone who picked four of the twenty letters blindly would get 0.8 of a row right on average.

Where the tests stop

Not everyone loses the yellow dots, and the letters test is harder on a small screen. The share of people who miss the gorilla varies from one version of the film to another. Each of these is a real and repeatable finding. None of them is a law with a fixed number in it.

Guessing right

In 1974 Lawrence Weiskrantz and colleagues reported on a patient, known as D.B., who had lost part of his primary visual cortex and with it part of his field of view. In that part he was not aware of seeing anything. Made to guess about things shown there, their shape or whether they moved, he was right far more often than chance allows, and put it down to coincidence. Weiskrantz and his colleagues called it blindsight.

Guessing is how such a thing is shown. Someone choosing between two answers at random will get 27 or more of 30 right about 4 times in a million. Whether blindsight is sight with no experience at all, or very faint experience hard to put into words, is still argued over.

What is taken up is small, and what is kept is smaller. Steven Luck and Edward Vogel found in 1997 that people can hold only about four colours or four tilted lines in mind at once. Yet over hours the total is vast: in 2008 Timothy Brady and colleagues showed people 2,500 objects, one after another, and afterwards they could pick the one they had seen from a pair 92% of the time.

And what is kept is rebuilt each time it is used. In 1974 Elizabeth Loftus and John Palmer showed 150 people a film of a car accident. Some were then asked how fast the cars were going when they hit each other, and others when they smashed into each other. A week later, those who had heard “smashed” were more likely to remember broken glass. There had been none.

Where this room stops

The part nobody can explain

This room has followed the light inward as far as anyone can follow it. Three numbers from each cone. An edited report from the retina. A best guess at the world, with priors behind it. A map on the cortex with its own geometry. A narrow gate, and a store that fades in a second. Every one of those can be measured, modelled, and tested on your own eyes, and you have done so.

What none of it explains is why there is anything it is like to see. A camera does three numbers. A program can guess, and attend, and remember. The philosopher David Chalmers named the gap in a talk in 1994, published the following year: explaining what the brain does with light is the easy problem, hard as it is, and explaining why doing it is accompanied by experience is the hard one.

There are theories. One holds that a signal becomes conscious when it is broadcast across the whole brain. Another, that it belongs to any system whose parts act on one another in a sufficiently integrated way, and it remains controversial. In 2025 the supporters of both, with a neutral group between them, published a test of one against the other on 256 people, with the predictions written down beforehand. The results agreed with some predictions of each, and challenged key claims of both.

You began with a question on a door: when you look at something, what are you actually seeing? This room’s answer is that you are seeing your brain’s best account of the world, built from very little, in the dark. How that account comes to be seen by someone is the oldest open question in science.

End of the fourth room

Added to your notebook
Light reaching the eye is not yet seeing. A thing in plain view can vanish while its light still arrives. A glimpse holds about twice what can be reported, and the extra is gone within a second. And much of what happens in front of us goes unseen while we attend to something else. Why any of it is experienced at all, nobody knows.

You have followed light from the lamp to the eye, past the edges of sight, back inside the lamp, and into the dark behind the eye. So what is it?