Light is energy in packets. A photon of wavelength λ has energy E = hf = hc/λ. Violet photons (about 3 eV) carry nearly twice the energy of red ones (about 1.8 eV), but an eye cannot tell the energy of a single photon: it only knows whether the photon was absorbed.

E = hc/λ = 1240 eV nm / λ

1. Light to chemical energy. The light is focused on the retina and passes through the layers of nerve cells to the cones at the back. In each cone the pigment molecule contains retinal, which absorbs the photon and flips from its 11-cis shape to the all-trans shape. This is the only step that needs light.

2. Chemical to electrical energy. The change of shape sets off a chain of chemical reactions that closes ion channels in the cone’s membrane. In the dark a cone is partly depolarised (about −40 mV). In the light it becomes more negative (hyperpolarised, down to about −70 mV).

3. Electrical to a neural signal. The cone releases less of the neurotransmitter glutamate. This changes the signal in the bipolar cells and then the ganglion cells, which fire the nerve impulses along the optic nerve to the brain.

Three kinds of cone. The probability that a cone absorbs a photon depends on the wavelength. The S, M and L pigments absorb best near 420, 534 and 564 nm (the exact peaks depend on the source). The curves are the standard template for visual pigments, which is skewed, with a longer tail on the long-wavelength side. The graph shows the pigment sensitivity and does not include the lens or the yellow pigment of the retina.

Colour is a ratio. One cone cannot tell colours apart: a dim light at its best wavelength gives the same response as a bright light at another wavelength. The brain compares the responses of the three kinds. The ratio stays the same when the light gets brighter or dimmer. Different mixtures of wavelengths can give the same ratio, and so look the same: this is why a screen can make yellow from red and green light.

The picture is not to scale. In real life each cone absorbs millions of photons a second in daylight, and the animation shows only a few. S cones are only about 5–10% of the cones.