Colour Contrast Illusions: Why Identical Greys Look Different

Three illusions, one cause

None of your colour perception is absolute. Every patch you look at is judged relative to what surrounds it, because the retina does not report brightness — it reports contrast. This page demonstrates three classic illustrations of that fact: simultaneous lightness contrast, Mach bands, and Edward Adelson's 1995 checker-shadow illusion, all traceable to the same piece of retinal wiring.

Centre-surround receptive fields

Retinal ganglion cells, the neurons that carry the eye's output down the optic nerve, have centre-surround receptive fields: light in a small central zone excites the cell, light in a concentric ring around it inhibits the same cell (or vice versa for off-centre cells). A uniform grey field falling equally on both zones produces almost no net response — the cell reports change, not level. This is exactly why a mid-grey square looks darker on a white background and lighter on a black one: the surround's inhibitory or disinhibitory contribution shifts the reported brightness even though the photon count reaching your eye from the square itself never changes.

response ≈ k · (L_centre − w · L_surround)

L_centre    luminance falling on the receptive field centre
L_surround  luminance falling on the surrounding ring
w           surround weight, roughly 0.6–1.0
k           a sensitivity gain

Mach bands: contrast at a slope, not just a step

Ernst Mach noticed in 1865 that a smooth ramp of brightness, with no actual step anywhere in it, appears to have a thin bright band where the ramp meets a flat light region and a thin dark band where it meets a flat dark region. Centre-surround filtering explains this too: at the point where the gradient's slope changes abruptly (the ramp meeting a flat plateau), the surround inhibition briefly overshoots, producing an edge-enhancing spike that has no physical counterpart in the actual light distribution. Mach bands are, in effect, edge detection with a visible side effect.

The checker-shadow illusion: context beats local luminance

Adelson's checkerboard takes the same principle further by adding a full scene interpretation on top of local contrast. Square A sits on a light tile in full light; square B sits on a dark tile inside a cast shadow. Measured with a light meter, A and B reflect the identical amount of light. But the visual system does not report raw luminance — it estimates reflectance, the physical property that stays constant regardless of lighting, and to do that it discounts the shadow using scene cues (the shadow's soft edge, the cylinder casting it, the checkerboard pattern's regularity). That discounting pushes B's apparent lightness up to match its true reflectance, and the two squares end up looking clearly different even though no low-level receptive field alone can account for the full effect — it requires mid-level grouping and shadow inference layered on top of contrast coding.

Why the brain does this at all

Reporting contrast instead of absolute luminance is not a design flaw; it is what makes vision work under wildly varying illumination. Absolute light levels swing across many orders of magnitude between an overcast room and full sun, but the ratio of light reflected by a white object to a nearby grey object stays roughly constant regardless of how bright the illumination is. By encoding relative differences, the visual system achieves rough lightness constancy — objects keep looking approximately the same colour across lighting conditions — at the cost of exactly the illusions demonstrated here, where two isolated patches with identical luminance are placed in surroundings the visual system was never designed to see in isolation.

Frequently asked questions

Try it live

Everything above runs in your browser — open Colour Contrast Illusions and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.