Ponzo Illusion

Two bars on a railway line

Two horizontal bars of exactly the same length lie across a pair of lines that converge towards the top, like railway tracks running into the distance. The bar near the top, where the lines are closer together, looks longer than the bar lower down.

Distance cues and size constancy

Converging lines are one of the strongest depth cues there is. Parallel edges such as railway tracks, corridors and roads appear to converge as they recede, and the visual system has learned the pattern so well that it applies it automatically, even to a simple diagram.

Once the lines are read as distance, size constancy kicks in. That's the visual system's habit of keeping objects at their true size as their distance changes: a person walking away from you doesn't seem to shrink, because the brain scales them up to compensate. The top bar is read as further away, yet it produces the same image on your retina as the "nearer" bar, so the brain concludes it must be physically longer. Scott Murray and colleagues found that this scaling shows up as early as the primary visual cortex: an object that looked bigger because it seemed further away activated a larger area of V1, even though its image on the retina was unchanged (Murray, Boyaci & Kersten, 2006).

The Müller-Lyer illusion also distorts length, but it depends on the shapes at each end of a line; the Ponzo illusion is a more direct demonstration of size constancy driven by a depth gradient across the whole scene.

How the test above works

The test runs eight trials in random order. In three, you adjust the bottom bar to match the top one; in three more, you adjust the top bar, so a habit of always setting the adjustable bar too long or too short cancels out instead of masquerading as the illusion. The last two trials take the converging lines away. They measure your ordinary eyeballing error, which gives the other six trials something to be compared against: if the rails shift you 12% while bare bars only put you 3% out, the difference is the illusion.

The adjustable bar starts at a random length each time, and both bars are shifted sideways by different amounts, so you can't simply check whether their ends line up vertically. The bars have square ends rather than rounded ones, so the length you see is exactly the length being measured.

Your Ponzo score and your Ebbinghaus score

When Samuel Schwarzkopf, Chen Song and Geraint Rees measured how strongly people experienced the Ponzo illusion and the Ebbinghaus illusion, they found that both were stronger in people whose primary visual cortex (V1) has a smaller surface area. Yet the two scores didn't predict each other: someone strongly fooled by one could be barely affected by the other (Schwarzkopf, Song & Rees, 2011). That suggests the two illusions, which look so similar on paper, lean on different parts of the visual system's size machinery. The Ebbinghaus page has the same kind of test, so you can compare your own two scores.

The same trick, everywhere

The Ponzo effect is common outside the lab. It's thought to contribute to the moon illusion, where the moon looks bigger on the horizon than high in the sky. Photographers and film-makers use it deliberately: placing someone near the vanishing point of a corridor or a road makes them look larger or more imposing than the same person without those lines, and forced perspective takes the same idea into real three-dimensional space.

Ponzo's lines

The Italian psychologist Mario Ponzo described the illusion in the early 1910s; it's usually dated to 1911, though his paper appeared in 1912 (Ponzo, 1912). He was interested in why the moon looks bigger near the horizon. Herschel Leibowitz and colleagues later showed that the illusion grows stronger when the converging lines are part of a real photographed scene with plenty of depth cues, which supports the size-constancy explanation (Leibowitz et al., 1969).