Eye — Laboratorium

Laboratorium

Five stations: from the anatomy of the eye to animal eyes.

Keys: 1–5 modes · ←/→ wavelength (with Shift in 10 nm steps) · Esc stops animations

The eye is a camera made of living tissue — and also an outpost of the brain. Point at or tap a part of the eye to see what it does. How the eye bends light and where vision defects come from is shown on a neighboring page: Optics of the eye and vision defects → Lens.

Cross-section of the eye

diagram · right eye from above, nose at the bottom

Pick a part of the eye

Hover over the drawing or click one of the names below it. In order from the front: the cornea, aqueous humor, the iris with the pupil, the lens, the vitreous body, and on the wall of the eyeball — the retina, choroid and sclera.

The eye resembles a camera: it has a lens, an aperture, autofocus and a sensor. Except that this “sensor” pre-processes the image itself.

Pupillary reflex

model: Moon and Spencer, 1944

Ambient brightness— —

Diameter: —
Pupil area: —
To the retina: —
Latency: ≈ 0.25 s

brightness (log scale)pupil diameter

The pupil changes its diameter from about 2 to 8 mm, so the area of the opening changes only about 16-fold, while ambient brightness can vary a hundred million times. The rest is done by adaptation of the photoreceptors (mode 2). Constriction starts about 0.2–0.3 s after a flash and is faster than dilation. The reflex works in both eyes at once: shining light into one eye also constricts the other pupil.

From eye to brain

visual pathway · view from above

Each hemisphere “watches” the opposite half of the world. Fibers from the nasal halves of the retinas cross to the other side at the optic chiasm, while those from the temporal halves stay on their own side. Via the thalamus, the signal reaches the visual cortex at the back of the head, and from there dozens of further areas: motion, shape, color and faces are each analyzed separately. Some fibers lead to the midbrain and control the pupillary reflex. How the nerve impulse itself travels is shown on the Neuron page.

The retina is as thin as a sheet of paper, yet it holds about 100 million photoreceptors and several layers of neurons. In vertebrates it is “inverted”: light has to pass through the neurons before it reaches the rods and cones.

Layers of the retina

diagram · not to scale

Point at a layer

Hover over a layer (or reach it with the Tab key) to see what it does. Amber dots are photons; green ones are the nerve signal heading back toward the optic nerve.

	Rods	Cones
Number	approx. 90–120 million	approx. 4.5–6 million
Pigment	rhodopsin, peak approx. 498 nm	3 opsins: S, M and L
Sensitivity	respond to a single photon	need dozens of times more light
Vision	night, colorless, blurry	daytime, color, sharp
Where	none in the center of the fovea; densest about 15–20° from it	densest in the fovea

The numbers vary between people and sources. Measurements by Curcio et al. (1990) gave an average of about 92 million rods and 4.6 million cones; textbooks often quote 120 million and 6 million.

Where the rods and cones are

density per mm² · approximation

rodsconesoptic disc

Curve shapes after measurements by Østerberg (1935) and Curcio et al. (1990), simplified. The center of the fovea has about 200,000 cones per mm² and not a single rod, which is why a faint star is easier to see when you look slightly to one side of it.

Sensitivity of the visual pigments

approximate curves · peak = 1

S cones ≈ 420 nmrods ≈ 498 nmM cones ≈ 534 nmL cones ≈ 564 nm

Absorption peaks of the pigments measured in single cells (Bowmaker and Dartnall, 1980). Curve shape: the template of Govardovskii et al. (2000). The sensitivity of the whole eye is shifted slightly, because the lens and the macula absorb some of the blue light.

Dark adaptation

illustrative model

Time in the dark—

Purkinje effect

at dusk, red fades

Light—

Jan Evangelista Purkyně described this phenomenon in 1825. Rods are most sensitive at about 500 nm and barely respond to red, so when they take over from the cones at dusk, red flowers darken and blue ones seem brighter. At night the colors vanish (simplified model: we blend a “cone” brightness with a “rod” brightness).

The brain doesn't receive a wavelength, just three numbers: how strongly the S, M and L cones responded. Color arises from their proportions. Light spectra and how to measure them are shown on the Spectrum page.

Three cones, one color

monochromatic light

Wavelength—

—

S cones

—

M cones

—

L cones

—

The color of the circle is an approximation — a screen can't show colors as saturated as monochromatic light.

Metamers: a yellow that isn't there

as in an anomaloscope

top: 670 nm + 545 nmbottom: 589 nm

Red light 670 nm—

Green light 545 nm—

L

—

M

—

S

—

Set the sliders so that the top half of the circle looks like the bottom half. When the L and M cones receive the same signals, the two halves are indistinguishable to us, even though the light spectra are completely different — these are metamers. Every screen works this way: from red, green and blue pixels it builds colors that aren't really in it. Eye doctors use this matching (the Nagel anomaloscope) to test for color vision deficiencies: a person with protanomaly needs more red, one with deuteranomaly — more green.

The world through the eyes of people with color vision deficiencies

simulation · methods of Brettel and Viénot

Red–green deficiencies, men of European descent	approx. 8%
Red–green deficiencies, women of European descent	approx. 0.5%
of which deuteranomaly (weakened M cones), men	approx. 5%
Protanopia or deuteranopia (no L or M cones), men	approx. 1% each
Tritanopia (no S cones)	rarer than 1 in 10,000

The genes for the L and M cone opsins lie on the X chromosome. Men have only one such chromosome, so a single faulty copy is enough — hence the big difference between the sexes. The simulation shows dichromacy, i.e. the complete absence of one cone type; the more common mild forms (anomalies) look somewhere in between.

An Ishihara-style plate, generated before your eyes. The number and the background lie on a so-called confusion line: to a dichromat they are exactly the same color. Choose the simulation of the same type above, and the number disappears. This is not a diagnostic test — the screen's colors aren't calibrated.

What you see is the brain's interpretation. The same mechanisms that usually help can also lead us astray — and illusions reveal how the visual system works.

Find your blind spot

one eye · approx. 15° from center

The optic disc lies about 15° from the fovea toward the nose, so the “hole” in the visual field is on the temple side. It measures about 5° by 7°. The brain fills it in with whatever is visible around it, which is why the line and the grid look continuous. In everyday life, the other eye covers the gap.

Afterimage

tired cones

Stare steadily at the black dot in the middle. After 20 seconds, the image will turn into a light gray field — and you will see a meadow on it in its normal colors.

Cones that receive a strong stimulus for a long time (e.g. only “blue”) temporarily weaken. On a gray field the others then have the upper hand, so you see the complementary colors — which here are the original ones.

Hermann grid

lateral inhibition

Gray spots appear at the intersections — but not where you're looking. Retinal cells inhibit their neighbors, and at an intersection there is brightness on four sides, so the inhibition is stronger (that's the classic explanation — scientists still debate whether it's the whole story). In the variant with dots, the white dots “flash” black even though the image stands still.

Mach bands

difference-of-Gaussians model

actual brightnessretinal response (model)

Each band has a uniform shade, yet near its edges it looks lighter on one side and darker on the other. The model: the signal from the center minus part of the signal from the surround (the difference of two Gaussian blurs) produces exactly this kind of “overshoot” at the edges.

Simultaneous contrast

the same gray?

Which gray square is the lightest? Click it.

How many frames per second?

apparent motion

Bottom ball—

From about 10–12 frames per second we already see motion rather than a series of stills; at 24 fps a film looks smooth. A movie projector shows each frame 2–3 times (48–72 flashes per second), because getting the flashes to fuse takes a higher frequency than smooth motion does.

Flicker and flicker fusion

critical flicker frequency

Warning for people with photosensitive epilepsy. This field flickers. By default the flicker is slower than 3 times per second, covers only a small low-contrast square, and switches itself off after 15 s. Faster flicker can only be enabled after you tick the consent box below. If flashing light has ever affected you badly, skip this experiment. Esc stops it immediately.

I have read the warning and want to unlock flicker above 3 Hz.

flicker

steady

Frequency—

Saccades

jumps you can't see

Your eyes make about 3 jumps (saccades) per second; each lasts a few dozen milliseconds and reaches several hundred degrees per second. Yet the world doesn't blur: during a jump the brain suppresses vision (saccadic suppression). Try it in a mirror: look at your left eye, then your right — you won't see your own eyes move, though someone next to you will. Drawing: a diagram (green dots are fixations, lines are jumps).

Eyes evolved many times, independently. Animals see colors, ultraviolet, even the polarization of light — each one the way it needs to.

Evolution of the eye, step by step

every stage exists in living animals

In “On the Origin of Species” (1859), Charles Darwin admitted that the idea of such a perfect organ arising by natural selection seems, at first glance, absurd. But he immediately explained that the difficulty disappears if there is a series of intermediate steps from a simple to a complex eye, each useful to its owner — and he pointed to such steps in living animals. The Nilsson and Pelger model (1994) estimated that going from a flat patch to an eye with a lens takes fewer than 400,000 generations — the blink of an eye, geologically speaking.

The octopus has no blind spot

diagram of the retinas

The cephalopod eye is built much like ours, but it evolved independently — an example of convergent evolution. In the octopus the photoreceptors face the light and the nerve fibers leave from their back side, so they don't have to pierce the retina, and there is no blind spot.

A flower through a bee's eyes

diagram · false colors

A bee has three kinds of photoreceptors: for ultraviolet (about 340 nm), blue (about 440 nm) and green (about 540 nm), but it hardly sees red. Many flowers have patterns in UV — “signposts to the nectar”. In the drawing, ultraviolet is shown as blue, blue as green, and green as red; that's an arbitrary convention.

Mantis shrimp

12 types

It has 12 kinds of photoreceptors for color vision, including several for ultraviolet (we have 3), and about 16 counting the polarization receptors. Even so, in experiments by Thoen et al. (2014) it was poor at telling colors apart: it confused lights differing by about 12–25 nm, whereas humans can tell apart differences of a few nanometers.

It probably doesn't compare signals the way we do, but quickly recognizes a color by which receptor responded.

Eagle

≈ 2×

Large birds of prey see detail about 2 times (by some measurements up to about 2.5 times) more sharply than humans. Their cones are packed more densely, and each eye has two foveae: one for looking sideways, the other forward.

Cat

a mirror in the eye

Behind the retina, a cat has a reflective layer — the tapetum lucidum. It reflects light back through the photoreceptors, giving them a second chance; that's why cats' eyes glow in car headlights. Its vertical slit pupil changes its area as much as about 135 times; our round one, about 15 times (Banks et al., 2015).

Dog

2 cones

Dogs are dichromats: they have cones sensitive to about 430 and 555 nm. They confuse red and green much like a person with deuteranopia — a red ball on grass is hard for a dog to see, a blue one very easy. On the other hand, in dim light and at detecting motion, dogs do better than we do.

How it sees…

schematic comparison · these are not “photos through the eyes”

Human — three kinds of cones, a sharp image in the center of the visual field.

Dog — two kinds of cones (yellow–blue), several times lower acuity.

Bee — a compound eye of thousands of facets (a mosaic, low acuity); sees UV, doesn't see red. False colors.

Cat — a wider field of view (about 200°), less color and sharpness, but at dusk it sees much more than we do.

Challenges

They complete themselves when you do the task in the lab.

Find your blind spot
In mode 4, cover one eye, look at the cross and slowly move closer to the screen. When the dot disappears, click the button.
To do
Shrink the pupil to its minimum
In mode 1, bring the pupil down to a diameter of at most 2.05 mm — and wait for the reflex to kick in.
To do
M and L cones respond equally
In mode 3, find the wavelength at which the M and L cones give the same signal (a difference below 1%).
To do
The world through the eyes of a person with deuteranopia
In mode 3, turn on the deuteranopia simulation and see what happens to the apples, the ball and the dot plate.
To do
See a flower through a bee's eyes
In mode 5, switch the flower to bee vision and find the pattern visible only in ultraviolet.
To do
A gray field that only looks lighter
In mode 4, point to the square that looks the lightest, even though they are all identical.
To do

How does it work?

The biology of vision in simple words

From photon to signal

Rods and cones contain millions of molecules of visual pigment: an opsin protein bound to retinal (a derivative of vitamin A). A photon straightens the bent retinal molecule (from the 11-cis to the all-trans form). This sets off a cascade of proteins that closes ion channels in the membrane. Curiously, a photoreceptor is excited in the dark and constantly releases a neurotransmitter — light quiets it. Bipolar cells pass this signal on, and ganglion cells turn it into nerve impulses that travel along the optic nerve to the brain.

Different opsins absorb different wavelengths best. How much energy a single photon carries is given by Planck's formula:

E = h · cλ ≈ 1240 eV·nmλ

h = 6.626·10⁻³⁴ J·s (Planck's constant), c ≈ 3.00·10⁸ m/s, 1 eV = 1.602·10⁻¹⁹ J.

Pigment	Peak λ	Photon energy

Visual acuity: one arcminute

A healthy eye resolves details seen at an angle of about 1′ (one arcminute, or 1/60 of a degree). That corresponds to 20/20 vision (“1.0” on a metric eye chart). How large is such a detail at a distance d?

s = d · tan α ≈ d · α, α = 1′ = π180 · 60 rad ≈ 2.91·10⁻⁴ rad

From 6 m that's about 1.7 mm; from 40 cm (reading distance), about 0.12 mm. On the retina 1′ is about 5 µm, and the cones in the fovea are spaced about 2.5 µm apart — the detail has to cover roughly two cones.

Pupil and brightness

The approximate formula of Moon and Spencer describes the pupil diameter D (in mm) at an ambient luminance L (in cd/m²). The amount of light reaching the retina is proportional to luminance times pupil area; this quantity is measured in trolands (Td).

D = 4.9 − 3 · tanh(0.4 · log₁₀ L), E_ret. = L · πD²4

The formula gives diameters from about 1.9 to 7.9 mm. The real pupil also depends on age (it shrinks with age), the size of the bright field, emotions, and even whether we are looking at something up close.

Color is the brain's work

Three kinds of cones are only the beginning. Already in the retina the signals are subtracted from each other: “red versus green” and “blue versus yellow” (opponent processes). That's why there is no “reddish-green” color, and why the afterimage of red is greenish. Missing one kind of cone flattens this color space to two dimensions — that is dichromacy.

Did you know…?

The image on the retina is upside down, and the brain interprets it so that we see the world the right way up. In the 1890s George Stratton wore glasses that inverted the image; after a few days he began to move around in it surprisingly well.
In the dark, a few photons are enough. In the experiment of Hecht, Shlaer and Pirenne (1942), about 5–14 photons absorbed in rods were enough to see a flash, and a single rod responds to one photon.
We blink about 15–20 times a minute, and each blink lasts a fraction of a second. We don't see darkness, because the brain suppresses vision for that moment — and when reading or looking at a screen, we blink less often.
We see truly sharply only within a field of about 1–2° — roughly the width of your thumbnail at arm's length. The brain assembles the rest of the image from rapid eye jumps.

What we simplified

The cone curves are a pigment template (without filtering by the lens and the macula). The colors on screen are approximate.
Pupil: the Moon and Spencer formula plus a simple model with a 0.25 s delay and different rates of constriction and dilation.
The photoreceptor densities and the dark adaptation curve are smoothed approximations of typical measurements.
The color vision deficiency simulations cover dichromacy and assume a typical sRGB screen. Animal vision is shown schematically — nobody knows “what it's like” to see through a bee's eyes.
The blind spot test, the afterimage and the flicker depend on your screen and conditions; they are demonstrations, not an examination.

Optics of the eye and vision defects → Lens The nerve impulse → Neuron

Eyehow we see

Laboratorium

Cross-section of the eye

Pick a part of the eye

Pupillary reflex

From eye to brain

Layers of the retina

Point at a layer

Where the rods and cones are

Sensitivity of the visual pigments

Dark adaptation

Purkinje effect

Three cones, one color

Metamers: a yellow that isn't there

The world through the eyes of people with color vision deficiencies

Find your blind spot

Afterimage

Hermann grid

Mach bands

Simultaneous contrast

How many frames per second?

Flicker and flicker fusion

Saccades

Evolution of the eye, step by step

The octopus has no blind spot

A flower through a bee's eyes

Mantis shrimp

Eagle

Cat

Dog

How it sees…

Challenges

Find your blind spot

Shrink the pupil to its minimum

M and L cones respond equally

The world through the eyes of a person with deuteranopia

See a flower through a bee's eyes

A gray field that only looks lighter

How does it work?

From photon to signal

Visual acuity: one arcminute

Pupil and brightness

Color is the brain's work