Why Do Optical Illusions Happen?

An optical illusion isn't your eyes going wrong. It's your eyes and brain working as designed and producing a perception that disagrees with what a ruler or a light meter would say. That gap, between the world as it is and the world as you experience it, is where every illusion on this site lives.

Perception isn't measurement

It's tempting to think of vision as an internal photograph, a faithful copy of the world. It isn't. The visual system evolved to help animals find food, avoid danger and move around, quickly. Speed and usefulness mattered more than precision. A system that took an extra half-second to produce a perfect image of a charging predator would lose out to one that produced a rough but good-enough image straight away.

That trade-off runs through the whole system, from the retina upwards. Once you see perception as a fast, useful approximation, illusions stop looking like glitches and start looking like the approximation showing through. The site's tests make the point concrete: on the Müller-Lyer, Ebbinghaus and Ponzo pages you can measure, in percent, how far your own judgement is pulled.

Three broad causes

Vision scientists group the causes of illusions into a few overlapping kinds, and many illusions involve more than one (Eagleman, 2001).

Early processing. Neurons in the retina and early visual cortex respond to contrast, edges and change rather than absolute values. That's why the café wall lines look tilted: tiny distortions in how each tile edge is registered add up along the line before any interpretation happens. Illusions of this kind survive knowing how they work, because the distortion is introduced too early for knowledge to reach it.

Inference. The retina is flat, so the brain has to reconstruct depth and size from clues such as converging lines, relative size and overlap. Richard Gregory described perception as a kind of hypothesis-testing that uses past knowledge to make sense of incomplete data (Gregory, 1997). The Ponzo illusion exploits this: converging lines signal distance, and the brain enlarges whatever seems further away. The hollow face illusion shows how strong such inferences can be: the brain's expectation that faces bulge outwards overrides direct evidence that the mask is hollow.

Context. The brain rarely judges brightness, colour or size on its own; it judges them against their surroundings, because relative judgements are more reliable under changing light. In the checker shadow illusion, the visual system answers the question "what shade must this square be, given the shadow on it?" rather than reporting the light reaching your eye. Dale Purves and colleagues go further and argue that everything we see reflects what similar images have most often turned out to be in past experience, which is why the same patch can look so different in different surroundings (Purves et al., 2001).

Ambiguity is a category of its own

Some illusions involve no faulty inference at all, only an image that genuinely fits two interpretations. The Necker cube matches two different 3D cubes exactly, and the brain, which insists on committing to one, switches between them. Figure-ground reversals such as Rubin's vase work the same way and are explained by the Gestalt principles the brain uses to decide what counts as an object. The polls on the spinning dancer and Necker cube pages track which interpretation visitors see first.

A bargain, not a design flaw

None of these mechanisms is a mistake. The contrast processing that tilts the café wall also makes edges stand out in real scenes. The depth inference fooled by the Ponzo illusion also lets you see that a car far down the road is distant rather than tiny. Illusions are the price of a bargain: some accuracy traded for speed and reliability across all the conditions real eyes face. The route signals take from the eye to the cortex is covered in how the brain processes visual information, and the modern theory of why the brain leans so heavily on assumptions is in predictive processing and the brain's shortcuts.

Sources and further reading

1. Gregory, R. L. (1997). Knowledge in perception and illusion. Philosophical Transactions of the Royal Society B, 352(1358), 1121–1127. Find this reference on Google Scholar
2. Eagleman, D. M. (2001). Visual illusions and neurobiology. Nature Reviews Neuroscience, 2(12), 920–926. Find this reference on Google Scholar
3. Purves, D., Lotto, R. B., Williams, S. M., Nundy, S., & Yang, Z. (2001). Why we see things the way we do: Evidence for a wholly empirical strategy of vision. Philosophical Transactions of the Royal Society B, 356(1407), 285–297. Find this reference on Google Scholar