How the Müller-Lyer Illusion Tricks Human Perception

The Müller-Lyer illusion is one of the most studied optical illusions in psychology: two lines of identical length appear dramatically different because of the angled “fins” or arrowheads attached to their ends. When the fins point outward (like the ends of an open book), the line looks longer. When they point inward (like arrowheads), the line looks shorter. Despite more than a century of research, no single theory fully explains why the brain makes this mistake, and the illusion continues to reveal surprising things about how human perception works, from its roots in genetics to the way it shifts across cultures, ages, and even clinical conditions.

What the Illusion Looks Like and How It Works

The classic version of the illusion uses two horizontal line segments of equal length. One segment has fins angling outward at each end, and the other has fins angling inward. Most people perceive the outward-fin line as roughly 10 to 30 percent longer than the inward-fin line, depending on the angle of the fins and the specific version used. The effect is immediate and involuntary: even after you measure the lines and confirm they are the same length, the illusion persists. Knowing the answer does not make the wrong perception go away.

The illusion is surprisingly robust against simplification. You might expect that removing half the fins would cut the effect in half, but that is not what happens. When researchers tested “single-wing” versions of the figure, eliminating one fin component from each end, the illusion was reduced but still substantial. Both the outward-fin and inward-fin forms shrank by a similar amount, with the largest distortion concentrated near the angles rather than in the middle of the line.¹ Separately, studies of “fractional” figures, which have line junctions at only one end of the shaft, found that the illusion persisted at nearly full strength, with only a slight reduction.² In other words, the visual system does not need the full symmetrical arrowhead pattern to be deceived. Even a partial cue is enough to distort perceived length.

Competing Theories for Why It Happens

The most widely known explanation is Richard Gregory’s “misapplied size constancy” theory. The idea is that your brain interprets the outward fins as depth cues resembling an interior corner of a room, which recedes from you, and the inward fins as an exterior corner of a building, which juts toward you. Because objects that are farther away but produce the same retinal image must be physically larger, your brain unconsciously inflates the perceived length of the “receding” line. It is an elegant story, and it shows up in virtually every introductory psychology textbook.

The trouble is that direct tests of the theory’s key assumption have not gone well. When researchers had participants judge the relative distance of actual three-dimensional model corners presented in a viewing box, the interior corner was perceived as nearer, not farther, than the exterior corner. That is the opposite of what Gregory’s theory requires.³ Another line of evidence comes from experiments using stereoscopic displays to place the flat illusion figure at different apparent depths. When the figure was made to look closer or farther away, the illusion persisted unchanged alongside large changes in overall perceived size driven by normal depth perception. If the depth-based theory were correct, adding real depth cues that conflict with the “implied” depth should weaken or reverse the illusion, but it did not.⁴

These findings do not necessarily kill the depth-based explanation entirely, but they suggest that something more fundamental is going on. Many researchers now think the illusion originates earlier in visual processing, before the brain even gets around to interpreting depth.

What Early Visual Processing Has to Do With It

A competing family of theories locates the distortion at the level of low-level spatial filtering, the initial stages where your visual system breaks down an image into components of different scales and orientations. In one study, participants were adapted to gratings of different spatial frequencies before viewing the illusion. Adaptation to low-spatial-frequency vertical gratings reduced the illusion, while other combinations had no effect. This is consistent with the idea that the illusion arises from how the visual system processes broad-scale spatial structure in the image.⁵

One complication is that the illusion survives even in images where the low spatial frequencies have been filtered out. This seems to undercut a purely low-frequency explanation. However, modeling work has shown that a “second-order” visual mechanism, one that extracts the overall envelope or outline of a texture pattern rather than its fine detail, can reproduce the distortion even in band-pass or high-pass filtered images. The physical distortions in the computed envelope consistently match the direction of the perceived illusion.⁶ Computational models of the visual cortex tell a similar story: when simulated images of Müller-Lyer figures are processed through a hierarchical model of visual neurons, a statistical bias toward the illusion exists at the raw input stage. As the signal passes through successive layers of complex cells (neurons that pool information across positions), the bias is reduced in most cases.⁷ In other words, the illusion may be partly “baked in” to the statistics of how the visual system samples the world, and the cortex actually works to correct it rather than create it.

What Brain Imaging Reveals

If the illusion involves both early processing and higher-level interpretation, you would expect to see activity in multiple brain regions, and that is exactly what neuroimaging studies find. Using fMRI, researchers identified two areas whose activity tracked the strength of the illusion: the lateral occipital cortex, which is part of the ventral visual stream involved in recognizing objects and their features, and the right superior parietal cortex, which belongs to the dorsal stream involved in spatial awareness and action planning. Modeling the information flow between these areas showed that the best-fitting account involved bidirectional connections between the ventral and dorsal streams, with the strength of the illusion modulating those connections. The lateral occipital cortex appeared to be more directly involved in the size distortion itself, while the parietal region reflected higher-level processing, including task-related monitoring.⁸

The fact that both streams are involved matters for a long-running debate in vision science about whether perception and action rely on separate visual representations. The Müller-Lyer illusion has been a key testing ground for this idea.

Does the Illusion Fool Your Hands Too?

An influential theory in vision science proposes that your conscious perception of the world (“vision for perception”) and the visual guidance of your movements (“vision for action”) rely on partly separate brain pathways. If true, your eyes might be fooled by the Müller-Lyer illusion, but your hand, when reaching to grasp something, might not be. Early experiments supported this: when people reached to grasp Müller-Lyer figures, their maximum grip aperture seemed immune to the illusion even though their perceptual judgments were clearly distorted.⁹

But a quantitative review of many such studies paints a more complicated picture. Across the literature, grasping was indeed less affected than perception, but almost all grasping studies still showed effects larger than zero, and there was substantial overlap between the two distributions. The apparent immunity of grasping turned out to depend heavily on whether participants could see their hand during the movement. When online visual feedback was available, the hand could correct itself mid-reach, shrinking the measurable illusion effect. When feedback was removed, grasping was fooled in much the same way as perception. The review concluded that the difference between grasping and perception in the Müller-Lyer illusion is better explained by real-time motor correction than by truly independent visual representations.¹⁰

Eye Movements Tell Their Own Story

Your eyes are also misled by the illusion, but not purely because of what you perceive. When people make a quick eye movement (a saccade) to the end of a Müller-Lyer shaft, the saccade lands in a biased position consistent with the illusion: it overshoots for outward fins and undershoots for inward fins. Researchers initially assumed this was just the saccade following the illusory percept, but more careful experiments showed otherwise. When a non-Müller-Lyer setup was used that largely eliminated the perceptual distortion, the saccade bias did not shrink proportionally. Part of the saccadic error turned out to reflect a “center of gravity” effect, where the eye is pulled toward the spatial average of nearby visual elements, regardless of the perceptual illusion. Furthermore, trial-by-trial analyses found no correlation between the size of someone’s perceptual error and the size of their saccadic error, suggesting the two biases arise from largely independent sources of processing noise.¹¹

How Age Changes Susceptibility

The strength of the Müller-Lyer illusion is not fixed across the lifespan. In children, the illusion tends to weaken with age: a study of children aged 4 to 17 found a clear negative correlation between age and illusion intensity, in both normally sighted children and those with visual impairment. The use of low-vision aids was also associated with lower illusion intensity among visually impaired children.¹²

In adulthood, the trend reverses. A comparison of young adults and normally sighted elderly participants found that the elderly group showed roughly twice the illusion intensity as the younger group (about 8% versus 4%). Regression analysis confirmed that age was the sole significant predictor of increased illusion strength; visual acuity, even when it differed slightly between groups, did not contribute.¹³ So the illusion shrinks as the visual system matures during childhood, then grows again as it ages, and this late-life increase is not simply because older adults see less sharply.

Cultural and Environmental Variation

One of the most famous findings in cross-cultural psychology is that people living in different built environments show different susceptibility to the Müller-Lyer illusion. The “carpentered world” hypothesis suggests that people who grow up surrounded by rectilinear architecture, with rooms, buildings, and furniture full of right angles, internalize those geometry cues more deeply and are more susceptible to illusions that exploit them. Studies in the mid-twentieth century reported that Western, urbanized populations were more vulnerable to the illusion than people living in more open or circular-dwelling environments.

However, when researchers tested children in Zambia matched for education but living in either “carpentered” or “uncarpentered” environments, they found traditional differences in susceptibility even with modified versions of the Müller-Lyer illusion that removed the perspective cues the carpentered-world theory depends on.¹⁴ This suggests that while environmental experience can shift how strongly the illusion affects people, the difference is not solely driven by familiarity with rectangular architecture. Other factors, including education, testing methods, and the general visual ecology of a person’s environment, likely play a role.

You Do Not Even Need Eyes to Be Fooled

One of the more striking demonstrations of the illusion’s depth is that it works through touch alone. When raised Müller-Lyer figures were presented to participants who explored them with their fingers, the illusion appeared reliably: the inward-fin stimuli were judged shorter than the outward-fin stimuli and shorter than plain lines. More telling, the effect was equally strong in blindfolded sighted individuals, people who lost their sight later in life, and people who were born blind. Visual status had no significant effect on illusion strength.¹⁵ This rules out visual imagery as the source of the haptic version and points to something more general: the distortion seems rooted in how spatial extent is processed, whether the input arrives through eyes or fingertips.

Animals Fall for It Too

If the illusion were purely a product of human cultural learning or high-level reasoning, you would not expect to see it in other species. But capuchin monkeys, tested on a touchscreen task where they had to choose the longer of two lines, consistently chose the line with inward-pointing arrowheads as shorter, showing the same directional bias as humans. A point-of-subjective-equality analysis confirmed that the arrowheads and their orientation influenced the monkeys’ perceived line length.¹⁶ The illusion has also been demonstrated in other primate species under saline control conditions, adding to evidence that the underlying mechanism predates human-specific cognition.¹⁷ Whatever produces this illusion, it is deeply embedded in primate visual processing, not a quirk of language, culture, or abstract reasoning.

Genetics of Susceptibility

If the illusion reflects something structural about the visual system, you might expect it to run in families. It does. A family study testing the Müller-Lyer and Ebbinghaus illusions in over 200 parent-offspring groups and 300 sibling pairs found significant familial resemblance in illusion susceptibility, consistent with a heritable component.¹⁸ More precise estimates come from a twin study using ACE genetic modeling, which decomposed the variance in illusion strength into genetic and environmental components. The heritability of the overall Müller-Lyer effect was estimated at about 61%, with a 95% confidence interval spanning roughly 36% to 77%. The underestimation and overestimation portions of the illusion each showed similar heritability, around 50 to 56%.¹⁹ In practical terms, more than half the variation in how strongly you experience the illusion can be attributed to genetic factors. The remaining variation comes from environmental influences and measurement noise, but genes are the single biggest contributor.

What the Illusion Reveals About Schizophrenia and Autism

The Müller-Lyer illusion has drawn clinical interest because susceptibility to it shifts in certain neuropsychiatric conditions. A systematic review of visual illusion processing in schizophrenia found concordant evidence of abnormal responses, with the Müller-Lyer illusion among the most consistently affected.²⁰ Interestingly, the pattern is not a simple reduction. Some evidence suggests that heightened susceptibility may be a marker of vulnerability detectable in early stages of the illness, which then changes as the condition progresses.²¹ In a translational study comparing chronic schizophrenia patients to healthy controls on the Brentano form of the illusion, patients showed higher susceptibility.²²

In autism, the picture is more contested. An early and influential study found that children with autism were less susceptible to several visual illusions, including the Müller-Lyer, which was interpreted as evidence for “weak central coherence,” a tendency to process local details at the expense of global context.²³ But later work using different measurement methods complicated this picture. When autistic children were tested with a method-of-adjustment task (where they actively manipulated one line to match the other), they actually showed heightened susceptibility to the Müller-Lyer illusion. The researchers suggested this discrepancy might reflect differences in how the children set their decision criteria rather than a fundamental perceptual difference.²⁴ How you measure the illusion, in other words, can determine whether you find reduced or increased susceptibility in the same population, a cautionary lesson that applies across much of illusion research.

Why No Single Theory Wins

After more than 130 years, the Müller-Lyer illusion still lacks a consensus explanation, and this is itself informative. The depth-interpretation theory cannot fully account for the evidence against it. The low-level spatial filtering account explains a lot but must be supplemented by second-order models to handle filtered images. The center-of-gravity account explains part of the saccadic bias but not all of the perceptual distortion. The haptic version rules out any theory that depends entirely on visual processing. The animal data rule out theories that depend on human-specific cognition or cultural experience. And the genetic data indicate that a large share of individual variation is biologically determined, not learned.

What seems increasingly clear is that the illusion is not a single phenomenon with a single cause. It likely reflects overlapping contributions from low-level spatial statistics, mid-level grouping and segmentation processes, and higher-order contextual integration, each adding its own small distortion in the same direction. This convergence of multiple small biases is probably why the illusion is so persistent, so resistant to knowledge, and so widespread across species, modalities, and cultures. No single layer of the visual system is wholly responsible, which means disabling any one layer does not make the illusion disappear.