Four Pac-Men. You See a Square. There is No Square.
Which patch is lighter?
You are looking at the Kanizsa square, a close cousin of the more famous Kanizsa triangle (also invented by Gaetano Kanizsa in the 1950s). Four pac-man shapes sit at the four corners of an invisible square, each with its mouth opening toward the centre. You perceive a vivid white square, complete with straight edges connecting pac-man mouth to pac-man mouth, and a faint brightness boost filling the interior. Nothing has been drawn: just four black pac-men. Everything else is your brain’s contribution.
What you are about to learn. What the Kanizsa square actually is, how it differs subtly from the triangle, the statistical “best explanation” account of illusory contours, what changes when you rotate the inducers, why the brightness illusion is present but mild, and why the square version is a favourite of perceptual-learning researchers.
What the Illusion Looks Like
Take four black discs. Cut a pac-man wedge out of each. Arrange them at the corners of a square, each pac-man’s mouth facing the centre.
You see a bright white square floating above the four pac-men, its edges crisp and straight, its corners exactly where the four mouths open. The square’s interior appears slightly brighter than the surrounding paper · a subtle glow. Measure any part of it with a colour picker: all white is identical.
The minimal recipe. Four inducers arranged at the corners of a virtual square, each oriented to suggest a corner of the hidden shape. The square illusion is weaker than the triangle · the four-corner geometry is slightly less aggressive than the three-corner triangle, because inferring a triangle from three inducers is more statistically overdetermined than a square from four. Still, the effect is clear.
Why It Works: Best-Explanation Inference
Like the Kanizsa triangle, the square is a demonstration of illusory contour completion. But the square version lets us see the mechanism more cleanly: the inducers provide less information per inducer (a single corner, versus the triangle’s corner that constrains more of the overall geometry), so your visual system has to do more guessing.
Four inducers, each a corner. Each pac-man suggests a right-angle corner of some foreground shape. Nothing yet fixes the shape’s type.
Your brain pools the corners into a hypothesis. Four right-angle corners arranged in a rectangle pattern → the simplest explanation is a rectangle (or square if the sides are equal). Your visual system accepts this hypothesis and renders the square.
The interior brightness follows. Once the square is hypothesised to be a foreground surface, your cortex assigns it a slightly brighter luminance than the background · a standard signature of foreground objects in natural scenes, which tend to reflect more light than what is occluded behind them.
Illusory contours are Bayesian inference. Your visual system holds a prior distribution over possible scene explanations and selects the most-likely one given the inducer cues. The cost of not inferring a foreground square when there is one (missing it, walking into it, failing to segment the scene) is higher than the cost of inferring a square that is not there (seeing a slight glow and a few phantom edges). Your brain is doing cost-weighted guessing, and the Kanizsa is that guessing made visible.
What Rotation Does
If you rotate each pac-man independently so that its mouth no longer points at the centre of the virtual square, the illusion collapses. Without aligned inducers, there is no coherent foreground-shape hypothesis, and your visual system parses the scene as four disconnected pac-men.
The alignment test. Mentally rotate one pac-man by 90 degrees so its mouth faces outward. The illusory square on that side immediately fragments · you still see some ghost contour along three of the four sides, but not the fourth. Your visual system tolerates some inducer noise, but a single misaligned inducer is enough to break the hypothesis for that edge. This shows how tightly the illusion depends on geometric alignment between inducers.
Triangle vs. Square: The Informativeness Difference
Why is the triangle illusion stronger than the square? Three points, three inducers, three corners · in a triangle, each inducer carries a third of the shape’s information. In a square, each inducer carries a quarter. The triangle’s inducers are therefore more informative per inducer, and the triangle hypothesis is more strongly supported.
The dose-response curve. Illusory-contour strength depends on the “support ratio” · the fraction of the total illusory contour that is actually induced by inducer edges. A Kanizsa triangle with fat pac-men (where the mouth takes up more of each disc) has higher support, stronger illusion. Skinny pac-men (small mouths) have lower support, weaker illusion. The square typically has slightly lower support per inducer than the triangle, which is one reason its effect is a touch milder.
The Illusory Brightness
Kanizsa figures produce not just illusory edges but also an illusory brightness boost in the interior region. The interior is perceived as slightly brighter than the surrounding paper · typically 3 to 8 percent. This brightness is generated by the same closure mechanism: once the visual system has decided the interior is a foreground surface, it paints the interior with the brightness expected of a foreground.
Common misconception: “there are two separate illusions here · edges and brightness.” They are the same illusion, running in two different perceptual channels. The closure mechanism that renders illusory edges also biases the interior brightness. Both follow from the same foreground-surface hypothesis. A clean demonstration: if you block any inducer with your finger, both the illusory edges and the brightness boost fade together. They are the output of a single inference, not two.
A Harder Variant
Below is a Kanizsa square at difficulty 3 · with cleaner inducers and slightly more aggressive geometry. The square is still entirely invisible in physical ink.
Which patch is lighter?
Tilt the figure. Rotate the whole figure by 45 degrees so the square stands on a vertex. The illusion is still present but slightly weaker · your visual system has a mild bias toward horizontal-and-vertical edges (the “oblique effect”) and struggles a little more with diagonal orientations. This is a reminder that your cortex has its own preferred axes, and illusions don’t float free of those axes.
Where Kanizsa Squares Appear in the World
- Modernist logo design. Rectangular negative-space logos (the IBM 8-bar logo, for example) rely on Kanizsa-style closure. Your visual system completes the letters even when only interrupted strokes are drawn.
- Architecture. Facades with regular rectangular reveals (recessed windows, balconies, cornices) create Kanizsa-square effects at the intermediate viewing distance where the reveals blur into inducer corners. The building reads as having extra planes and shapes that are not physically there.
- Packaging. Rectangular brand-name plates on product boxes are often indicated by corner marks only, with no drawn outline. Consumers see a clean rectangle, produced by Kanizsa-style completion.
- Interface design. Modern flat UI design often uses corner inducers to define clickable regions without actually drawing a border. Your browser’s developer tools probably show no border on some of the “buttons” you are clicking · the border exists only in your perception.
- Military camouflage and counter-camouflage. Disrupting inducer alignment (random mottled patterns) breaks Kanizsa closure and hides shapes. Conversely, adding inducer cues can make hidden shapes suddenly pop out. This is the cognitive principle behind both effective camouflage and effective counter-camouflage.
Test Yourself on 50 More Illusions
The Kanizsa square is one of more than 50 classical illusions on PlayMemorize. Each round draws a deterministic SVG scene and asks one grounded question: which is larger, which is brighter, which is actually parallel. The reveal overlay shows the true geometry plus a one-line “why it works” caption.
- Keep playing Kanizsa Square → · the standalone game, pinned to this one figure with fresh seeds each round
- Play Illusions → · spot the tricks across size, colour, orientation, and impossible figures
- Play Spatial → · train mental rotation and area estimation
- Play Matrix → · abstract pattern reasoning under time pressure
The takeaway. The Kanizsa square is more Bayesian evidence from your cortex. Four pac-men at rectangular corners → probably a square · so your visual system renders one, complete with edges and a brightness boost. The inducers are necessary. The hypothesis is automatic. The rendering is vivid. You cannot will the square away by knowing it is not drawn, because the inference is running in V2 of your visual cortex, below the reach of conscious volition. It is one of the simplest demonstrations that perception is a hypothesis, not a transcription · and once you see it working, you see it working everywhere in your visual experience.
Illusions
Your eyes lie - the math knows the truth. Spot equal lengths, identical greys, and truly parallel lines across 57 classic optical illusions
Jetzt spielen - kostenlosKein Konto nötig. Funktioniert auf jedem Gerät.
