[MM] Header Banner Horizontal Desktop 2026
1
Media MICE
Hermann grid with grey spots appearing at the intersections of black squares and white corridors

The optics behind the classics

Grey spots that aren’t there, straight lines that lean, and two equal lines that refuse to measure up. Vision scientists explain what classic illusions reveal about the processing between retina and cortex, and why it matters for the patient in your chair.

Look at a Hermann grid and ghostly grey spots appear where none exist. Stare at a Café Wall pattern and perfectly parallel lines start to lean. Add arrowheads to two identical lines in the Müller-Lyer illusion and one suddenly looks longer than the other. 

We know these perceptions are wrong. Our visual system, evidently, didn’t get the memo.

Even the term “optical illusion” doesn’t quite hold up, according to vision scientist Prof. Michael Bach (Germany).

“‘Optical illusions’ are a misnomer—that would only apply to the rainbow and to floaters.,” he said. “‘Perceptual illusion’ would be better, and I frequently prefer ‘visual phenomenon’ because there is no error in perception.”

What we perceive, he explained, is far more than what actually reaches the brain via the eyes. “Much is actually ‘invented’ by the visual system, allowing for rapid assessment in typical situations,” he explained. “But this inventiveness runs astray in atypical ones, which leads to what we call ‘optical illusions’.”

READ MORE: The Summer Blind Spot

Not so retinal after all

The Hermann grid has long been held up as the classic demonstration of retinal processing. Black squares are separated by white corridors, and in peripheral vision, grey spots appear at the intersections. Those spots vanish the moment you look directly at one.

For decades, the standard explanation centered on lateral inhibition and the center-surround receptive fields of retinal ganglion cells. These cells respond to the difference between the center of their receptive field and its surrounding, rather than to absolute luminance. That organization enhances local contrast and helps the visual system find boundaries.

“Center-surround organization is a dominant way to organize visual information because it intrinsically highlights anything that is different from the background,” explained cognitive neuroscientist Prof. Hinze Hogendoorn (Australia). 

Combining signals from these cells also allows neurons further along the visual pathway to respond to increasingly complex features, including orientation.

Applied to the Hermann grid, the logic seems straightforward. A receptive field centered on an intersection has more white in its inhibitory surround than one centered on a corridor. It should, therefore, be more inhibited, producing a darker spot.

Elegant? Yes, but incomplete.

“For a while, the Hermann grid was thought to result from retinal processes alone, but slight modifications to the image that should not have affected the putative retinal causes were found to eliminate the illusion,” said Prof. Hogendoom. 

“The modern interpretation is due to the extraction of more complex features in the cortex,” he added. “It seems to be that the brain is detecting overlapping bars, but is somehow making incorrect assumptions about how those bars behave at their intersections, causing the strange scintillating dots at the intersections when viewed peripherally.”

Wavy lines, for example, can abolish the effect. Vision scientist Prof. Juno Kim (Australia) noted that even rotating the grid can modify it. Such findings are hard to reconcile with a purely retinal center-surround account.

READ MORE: Contrast Sensitivity Outperforms Visual Acuity in GA, Study Finds

Off the straight and narrow

Now consider a brick wall. In the Café Wall illusion, rows of alternating dark and light tiles are offset horizontally from one another and separated by thin “mortar” lines. The mortar lines are perfectly parallel, yet they appear to converge, diverge or tilt.

Here, there isn’t even a tempting retinal explanation. Prof. Bach noted that retinal receptive fields are broadly circular, whereas cortical receptive fields can be elongated and orientation-selective. He attributes the Café Wall illusion primarily to cortical processing.

The key ingredients are contrast boundaries, orientation and the spatial arrangement of neighboring elements. Computational models suggest that interactions among orientation-sensitive receptive fields can produce neural response patterns that match the perceived tilt.¹

Prof. Kim explained that when consecutive rows of dark tiles are displaced in the same direction, the mortar appears to tilt relative to the direction of the tiles. When the pattern then zigzags the other way, the perceived orientation changes. “This depends on local and global processing of orientation information in the brain,” he said.

In other words, the visual system doesn’t analyze each line in isolation. Neurons respond to edges, orientations and spatial relationships, and these local signals are integrated into a global percept. Prof. Hogendoorn groups the Café Wall with other geometric illusions, in which the orientation and context of contrast edges bias perceived geometry. 

The mortar hasn’t changed. Its neighbors have.

The visual system adds a dimension

The Müller-Lyer illusion takes things a step further. Two lines of identical length are capped with fins pointing either inward or outward, and one reliably appears longer. Retinal center-surround processing can’t account for this, and neither can simple edge detection.

Both Prof. Bach and Prof. Kim favor an explanation based on perspective and higher-level visual processing. In a 2005 paper, vision scientist Richard Gregory argued that the effect reflects the visual system’s interpretation of three-dimensional space.² 

Prof. Bach still prefers that account to the so-called “cultural byproduct” explanation. Prof. Kim agreed that the effect “really depends on 3D perspective,” drawing on the same processes used to recover shape and magnitude from a visual scene.

That points to a problem the visual system must solve every waking moment: the world is three-dimensional, but the retinal images from which we reconstruct it are not.

“We only understand 3D world properties from 2D image structure that reaches our eye,” said Prof. Kim. Low-level processes extract features. Higher-level processes must then work out what those features are likely to represent in the real world.

Because those retinal signals are inherently incomplete, Prof. Hogendoorn added, the brain has to make assumptions about the most likely cause of any given pattern. Most of the time, those assumptions work spectacularly well. The Müller-Lyer figure simply supplies contextual cues that push the calculation towards an answer that disagrees with the ruler.

Retina, cortex and everything in between

It’s tempting to file each illusion neatly under one anatomical level: the Hermann grid under retina, the Café Wall under V1, Müller-Lyer under higher cognition. As the Hermann grid has already shown, reality is less tidy.

Prof. Kim suggests a more useful framework: think of perception as involving low-level retinal, mid-level cortical and higher-level cognitive processing.

Information is already being transformed within the retina before it travels via the optic nerve and thalamus to the primary visual cortex. Cortical neurons then extract and combine increasingly complex features. Later stages interpret those features according to context, prior knowledge and likely real-world causes.

Nor should the retina be dismissed as a mere camera sensor. Cognitive neuroscientists may underestimate how much computation happens there, Prof. Hogendoorn cautioned, particularly for evolutionarily fundamental properties such as position and motion.

Why clinicians should care

For optometrists, all of this matters well beyond visual curiosities on a screen. 

Patients don’t experience retinal or neurological disease as optical coherence tomography scans, visual-field plots or contrast-sensitivity curves. They experience changes in perception: missing areas, distortion, altered contrast, movement or visual phenomena that are hard to put into words. 

What a patient sees is the product of processing throughout the visual system, not simply an image projected onto an intact retina.

Prof. Bach pointed to negative scotomata and Charles Bonnet syndrome as particularly interesting clinical examples. Charles Bonnet syndrome is a striking demonstration that when visual input becomes incomplete, the perceptual system can generate experiences that go beyond what the eyes are delivering.

Prof. Hogendoorn is more cautious about using illusions as models of pathology. He sees them primarily as “non-invasive (and fun) ways to investigate and gain insight in healthy vision,” because they expose the shortcuts a normally functioning visual system takes to solve difficult perceptual problems. Understanding those shortcuts may be where their real clinical value lies.

Seeing, in short, is anything but passive. Before we’re consciously aware of anything, the visual system has enhanced contrast, found boundaries, extracted orientation, integrated context and inferred three-dimensional structure. Illusions are simply the moments we catch it in the act. 

As Prof. Bach put it, they’re “not an error in perception.” Instead, they offer a window onto the unconscious processing that lets us act quickly, and usually correctly, on incomplete information.

READ MORE: The Precision Paradox

Editor’s Note: This content is intended exclusively for healthcare professionals and eye care practitioners. It is not intended for the general public. Products or therapies discussed may not be registered or approved in all jurisdictions, including Singapore. A version of this article was first published in COOKIE Issue 25. 

References 

  1. Todorović D. Extension of a computational model of a class of orientation illusions. Vision Res. 2024;223:108459.
  2. Gregory RL. The Medawar Lecture 2001 knowledge for vision: vision for knowledge. Philos Trans R Soc Lond B Biol Sci. 2005;360(1458):1231-1251.

About The Contributors

Prof. Michael Bach
Prof. Michael Bach

Vision scientist at the Eye Center, University Medical Center in Freiburg, Germany.  

bach@uni-freiburg.de 

index_eis
Prof. Hinze Hogendoorn

Professor of cognitive neuroscience at Queensland University of Technology in Australia.   

hinze.hogendoorn@qut.edu.au 

images
Prof. Juno Kim

Vision scientist at the School of Optometry and Vision Science,  University of New South Wales, Australia. 

juno.kim@unsw.edu.au