The fourth room · Behind the eyeChapter 20 of 21
People who see things that are not there, in fever, in migraine, in darkness, under drugs, on the edge of sleep, have for centuries drawn the same few shapes. The shapes are not in the world or in the eye. They are the shape of the machinery.
Chapter 20 · Behind the eye
Seeing Without Light
A colour no light is sending, a slow wave that crosses the brain, and why hallucinations come as tunnels and spirals.
Look first
The map behind the eye
The optic nerve’s fibres end, after one relay, on a sheet of cortex at the very back of the head. The sheet holds a map of the visual field: neighbouring points in the view go to neighbouring points on the sheet. It is this map that Hubel and Wiesel’s electrode was in.
It is a strange map. One degree at the centre of your gaze gets 23 millimetres of cortex. One degree at the edge of your view gets a fifth of a millimetre. The middle 10 degrees are about 1% of what you see, and take up nearly half of the sheet.
Jonathan Horton and William Hoyt measured it in 1991, from patients with small injuries to that part of the brain: each had a blind patch in the view, and a scan showed where on the sheet the injury was.
The same five rings in the visual field, left, and where each lies on the cortex, right, from Horton and Hoyt’s measurements.
Commit to a prediction
After staring at red
Begin with the smallest case. You will stare at a red disc for twenty seconds, without moving your eyes. Then the disc will be taken away, and the whole panel will be white.
What will you see where the disc was?
Try it with your own eyes
Press the button, then look steadily at the small cross in the middle of the red disc. Keep your eyes on the cross when the disc goes.
your eyes calculated
When the disc has gone, every pixel of the panel is the same white. The small patch that appears beside it afterwards is the colour the cone model predicts.
Most people see a disc of pale blue-green, floating on the white, which drifts when they move their eyes and fades in a few seconds. There is no blue-green light on the screen. The panel is sending the same white from every point.
Twenty seconds of red drove your L cones hard and the others hardly at all. A cone that has been working hard answers less for a while. So under the place where the disc was, white light now gets a full answer from two kinds of cone and a weak one from the third. Chapter 17 says what three numbers like that look like: white with the red taken out.
Why the first guesses made sense
A signal is a signal
If you said nothing, you were trusting the light: white in, white seen. But nothing behind the cones can tell a signal that light caused from a signal with some other cause. Whatever sets the machinery going, the result is seeing.
Newton knew a way to do it with a finger. “When a Man in the dark presses either corner of his Eye with his Finger, and turns his Eye away from his Finger, he will see a Circle of Colours like those in the Feather of a Peacock’s Tail.”
The crews of the last seven Apollo flights saw flashes of light that no lamp was making. The cause, as far as could be told, was the particles of chapter 12: heavy cosmic rays passing through their heads and eyes. And in 1968 Giles Brindley and Walpole Lewin placed a grid of electrodes on the visual cortex of a blind woman of 52. A small current through an electrode, and she saw points and spots of light.
A case you have not seen
Stripes on the cortex
Suppose the sheet of cortex becomes active by itself, with no signal from the eyes at all, and suppose the activity takes the simplest form a sheet can show: plain parallel stripes, like ripples on sand, a few millimetres apart.
What does the person see?
Left: stripes of activity on a strip of cortex, 8 mm from crest to crest. Right: the same activity as its owner would see it, out to 40 degrees from the centre of gaze.
Equal steps along the cortex are equal ratios in the view. So stripes that lie across the map are seen as rings, each a fixed number of times larger than the one inside it: a tunnel. Stripes that lie along the map are seen as rays from the centre. Stripes at a slant are seen as a spiral.
In 1926 Heinrich Klüver studied what people saw under mescaline, and found the same geometric patterns recurring from one person to the next. He sorted them into four kinds: lattices, cobwebs, tunnels and spirals, and called them form constants. They turn up far beyond that drug: in fever, in migraine, on the edge of sleep, in long darkness, and when the closed eye is pressed. In 1979 Bard Ermentrout and Jack Cowan showed that they are what the simplest patterns a sheet of cells can fall into would look like through the cortex’s map.
One more case
A slow wave
Before some migraines, a disturbance creeps across the visual cortex like a slow tide, at a few millimetres a minute. The person sees a shimmering, zig-zag arc near the centre of gaze, with a blind patch behind it. Take the tide to move at a steady 3 millimetres a minute.
How does the arc move in the person’s view?
Left: the wave on the cortex, moving at a steady 3 mm a minute. Right: where its owner sees it, in the right half of the view.
It starts as a small flicker beside the point of gaze and takes ten minutes to creep out to 10 degrees. Six minutes later it is at 30, and four minutes after that at 60, a great arc sweeping off the edge of the view. By then it is moving 35 times faster than at the start, and nothing about the wave has changed.
In 1995 Otto-Joachim Grüsser published drawings of such arcs, made every minute or two while they were happening, and plotted their distance from the centre of gaze against time. The points lay on an exponential curve: the curve this map gives for a wave of constant speed. A migraine draws the map of the cortex on the view from the inside.
The bench
A sheet that stripes by itself
Why would a cortex make stripes? In this sheet every cell excites its near neighbours and quiets those a little further off. Below a certain gain the sheet stays at rest. Raise the gain and run it.
Things to try
- Run the sheet with the gain below the threshold. What happens to the faint noise it starts with?
- Raise the gain just past the threshold and run it again. How far apart are the stripes that form?
- Stir the sheet and run it once more. Is the pattern the same? Is its spacing?
What you can now say
Seeing does not need light. It needs the machinery to be active, and anything that sets it going is seen: tired cones, a pressed eye, a cosmic ray, a wave crossing the cortex. What is then seen has the shape of the machinery, above all of the map that gives the centre of gaze half the cortex.
Ordinary seeing uses the same machinery. The difference is that the eyes are steering it.
For the determined reader: the map, the sheet and the limits
The map
With E the angle from the centre of gaze, M the millimetres of cortex for one degree, and x the distance along the cortex:
M(E) = 23.07 / (1 + E / 0.75), x(E) = 17.3 ln(1 + E / 0.75)
The second follows from the first by adding up. Round the centre of gaze the page takes distance on the cortex to be 17.3 mm for each radian, which is a good approximation except within a degree or so of the centre. Each half of the brain maps the opposite half of the view, and the pictures here join the two halves as mirror images.
The sheet
Each cell’s activity decays toward rest and is driven by its neighbours: excited by those within about 0.4 mm, quieted by those within about 0.8 mm, in balance. A ripple of a given spacing is fed back by a known amount, and rest is stable only while the gain times the largest such amount is less than one. The tests take the feedback by direct numerical integration, find the threshold with an optimiser, and check that the engine’s own sheet grows a pattern of the predicted spacing. The two ranges are a model’s. They were chosen to give a spacing of 2.7 mm, between one and two times the width of the cortex’s repeating unit, which is put at 1.3 to 2 mm.
Where this stops
This sheet makes stripes. A real cortex also codes the angle of edges at every point, and the fuller theory of Paul Bressloff, Cowan and colleagues, which includes that, is needed for lattices and cobwebs. Nobody has watched a human cortex form stripes during a hallucination. The theory explains the shapes people report, and that is the evidence for it.
The grandfather’s visitors
In 1760 the Swiss naturalist Charles Bonnet described the case of his grandfather, who was 87 and nearly blind from cataracts. The old man had begun to see men, women, birds, carriages, buildings and tapestries, none of them there. People in his position generally know that what they see is not real. It is the eyes that have failed, and a visual system with little coming in begins to supply its own.
The same machinery runs when you picture something on purpose. In 1995 Stephen Kosslyn and colleagues scanned people who had closed their eyes and were visualising objects, and found the primary visual cortex active: the very sheet this chapter has been about. Where on the sheet depended on how large the imagined object was, as the map would require.
Not everyone can do it. Francis Galton noticed in 1880 that some people report no mental pictures at all, and in 2015 Adam Zeman and colleagues gave the condition a name, aphantasia. Such people see perfectly well. What they lack is the ability to run the machinery from the inside at will.
Where this explanation stops
Why activity should be light
This chapter explains the shapes. It does not explain the seeing. A wave of activity crossing a sheet of cells is, as physics, ions moving through membranes. Why that should be, for the person, a shimmering arc of light, and not nothing at all, is not touched by anything here.
Seeing can happen without light. The last chapter asks the opposite question, and it is the stranger one: can light reach the eye, be caught, be signalled, be processed, and still not be seen?
End of chapter 20
Added to your notebook
Seeing does not need light. It needs the machinery to be active, and anything that sets it going is seen: tired cones, a pressed eye, a cosmic ray, a wave crossing the cortex. What is then seen has the shape of the machinery, above all of the map that gives the centre of gaze half the cortex.
Seeing without light. Is there light without seeing?
Chapter 21 · Light Without Seeing