Optical illusions: why your clever brain gets fooled

Picture a ring of colorful patches that slowly turns while you look at it, even though nothing on the page is moving. Or two circles of exactly the same size, where one looks bigger just because of the circles around it. Optical illusions feel like tiny magic tricks played on our eyes.

Here is the fun part. When scientists dig into these tricks, they keep finding that much of the action happens in the brain, not only in the eyes. And every illusion turns out to be a little window into how seeing works.

Seeing is a kind of guessing

Hermann von Helmholtz described what we see as “unconscious inferences.” The brain quietly works out what is out there from two things: the signals coming in from the senses, and knowledge picked up in the past.

Richard Gregory, a psychologist at the University of Bristol, took that idea and ran with it. In 1980 he compared perceptions to hypotheses in science. A radio telescope collects signals and processes them using certain assumptions to turn them into useful data. In the same way, he wrote, the brain processes its nerve signals to give us what we see. In 1997 he added a lovely twist. Our perceptions are like the predictions of science, he wrote, except that we experience them as out there in space and accept them as our most immediate reality.

So why study illusions? Gregory argued that systematic mistakes in perception are important clues. They show where the signal channels have limits, and they hint at the rules the brain uses to make its guesses. Illusions, he wrote, can reveal the brain’s knowledge of objects and its working rules for vision. When a guess goes wrong in the same way for many people, we get to peek at the rule behind it.

A gray that changes with its neighbors

Start with something simple. In 1993, Edward Adelson pointed out that how bright a gray patch looks depends on what surrounds it. A medium-gray patch looks darker on a bright background and brighter on a dark one.

Simple models explain that with low-level mechanisms, the early steps of visual processing. But Adelson made a new set of brightness illusions that those models could not explain. In them, the brightness people saw depended strongly on how the whole picture was organized. Small changes that should hardly matter to low-level mechanisms made the illusion much stronger or weaker. Adelson suggested that more complex processing, later in the visual system, was at work. The brain was not just measuring gray. It was making sense of a scene.

The mask that won’t stay hollow

Now for a spooky-cute one. Look at the hollow inside of a face mask, and the face can seem to pop out toward you like a normal face. This is called the hollow-face illusion.

In 1993, Harold Hill and Vicki Bruce measured how strong it is in a clever way. People walked away from a mask, and the researchers noted the distance at which the illusion appeared. Then people walked toward the mask, and the researchers noted the distance at which it disappeared. In every experiment, the direction of walking made a difference. Whatever people saw first tended to stick.

The illusion was stronger for an upright face and with light from above, which fits a brain that prefers upright faces and top lighting. But upside-down masks also fooled people to some extent. That suggested a more general preference for shapes that bulge outward. Seeing with both eyes helped break the spell when people were close to the mask. Each eye gets a slightly different view, and that difference carries depth information that can overrule the brain’s favorite guess.

Same lines, different worlds

If seeing uses knowledge from the past, could people who grow up surrounded by different things see illusions differently? In 1963, Marshall Segall, Donald Campbell and Melville Herskovits reported data from 15 societies. They found substantial differences between societies in how strongly people were fooled by geometric optical illusions, the kind made from simple lines and shapes.

The researchers read the pattern as a sign of different “habits of perceptual inference,” tied to cultural and ecological features of the visual world around people. In other words, the brain’s guessing rules may be partly tuned by what it usually gets to see.

Your brain’s own size map

The primary visual cortex, or V1, is one of the earliest stops for visual signals in the cortex. It holds a kind of map of what falls on the eye.

Two objects can cast exactly the same size image on the eye, yet look very different in size if they seem to be at different distances. In 2006, Scott Murray, Huseyin Boyaci and Daniel Kersten used fMRI to see what V1 does with such size illusions. An object that seemed farther away, and so looked bigger, lit up a larger area of V1 than an object of the same image size that seemed closer and smaller. So the image on the eye and the sense of depth in a scene are combined very early in the visual system. The illusion is already there, written into the brain’s map.

Then comes a surprise about us as individuals. V1 is not the same size in everyone. In 2011, D. Samuel Schwarzkopf, Chen Song and Geraint Rees noted earlier reports that its surface area varies over a threefold range between people. They tested 30 people on two illusions. One was the Ebbinghaus illusion, where two identical circles look different in size because of the circles around them. The other was a version of the Ponzo illusion, where two identical checkered circles look different because of the three-dimensional scene behind them.

The result: the smaller someone’s V1, the stronger both illusions were for them. Each person’s results stayed very steady when they were tested again, so the differences were not a fluke of one test. The link held for V1, but not for the neighboring visual areas V2 and V3.

The researchers were careful about what this means. Their study took a snapshot, so they could not say whether the link forms during development or later in life. They also noted that the Ebbinghaus illusion apparently differs between cultures. That raises the possibility, they wrote, that some of those differences could partly reflect differences in brain structure.

Edges that aren’t there

Some drawings make people see an edge where no line has been drawn at all. These are called illusory contours. Your brain simply fills in the border.

In 1984, von der Heydt, Peterhans and Baumgartner showed such drawings to alert monkeys and recorded single cells in the visual cortex. Cells in a visual area called area 18 responded as if the edges were made of real lines. When the drawings were changed in ways that weaken the edge for human viewers, the cells responded less, too. Cells in area 17 apparently could not “see” these edges. So filling in a missing border looks like real work done by real neurons.

Pictures that seem to move

The rotating snakes illusion by Akiyoshi Kitaoka is a completely still image that seems to turn. A 2018 paper notes that it produces a strong sense of motion in humans, and that other studies have found it works on cats and even fish.

In 2005, Bevil Conway, Kitaoka, Margaret Livingstone and colleagues set out to find how a still image of this kind can move. First, color did not matter. The illusion still worked when yellow was swapped for light gray and blue for dark gray. What mattered was how light or dark the neighboring patches were. Motion was seen running from black to dark gray to white to light gray and back to black. Each pair of neighboring patches, shown alone, produced motion in the same direction as the full illusion.

Then the team recorded single neurons in the visual cortex of macaque monkeys. Neurons that normally prefer movement in one direction responded to the still patch pairs, and in the same direction as the illusory motion. It was the first time single neurons had been shown to give directional responses to still displays. The team’s explanation is that early motion detectors in the brain respond to light and dark patches with slightly different timing, and they read that timing difference as movement.

There is a newer twist. A leading idea in brain science, called predictive coding, says the brain is always predicting what it will see next and uses its mistakes to improve its predictions. In 2018, Eiji Watanabe, Kitaoka and colleagues tested a computer network built on this idea, called PredNet. They trained it on videos from cameras mounted on caps worn by people spending their day at Disney World in Florida. The network learned to predict the motion of a spinning propeller in videos it had never seen.

Then they showed it the rotating snakes. The network had never seen the illusion, and the picture wasn’t moving. Yet its predictions contained rotation in the same direction people see. For control pictures in which people see no motion, it found none. The researchers wrote that visual prediction is extremely accurate, but it occasionally makes “mistakes,” such as illusions. Their result supports the idea that prediction is one of the roots of motion illusions.

Pictures that flip

Some pictures can be seen in two ways, and our view keeps flipping between them. In a 1999 review, David Leopold and Nikos Logothetis pointed out that these flips behave a lot like actions. They start on their own, can often be brought about on purpose, and are swayed by attention and mood. They get easier with practice, and they can be impaired by damage to brain areas outside the visual system.

So Leopold and Logothetis offered an idea they themselves called speculative: the flips may be set off by higher brain areas that help plan and coordinate behavior, not only by a tug of war inside the visual system.

So is your brain bad at seeing?

Not at all. Gregory pointed out that vision leans heavily on knowledge from the past. That raises a puzzle: how do we see the present without mixing it up with the past? His suggestion was that the present is signalled by live input streaming in from the senses.

Remember what the PredNet team wrote: visual prediction is extremely accurate, with only occasional mistakes. Illusions are those mistakes, the moments when a clever rule meets a picture made to fool it. A gray square changes with its neighbors, a hollow mask turns into a face, a still drawing starts to spin. Each one shows a rule your brain uses every day, and it usually serves you very well.

#optical illusions #perception #vision #visual cortex #everyday brain

Sources

1. Knowledge in perception and illusion. Gregory RL. Philosophical Transactions of the Royal Society B 352, 1121–1127 (1997)
2. Perceptions as hypotheses. Gregory RL. Philosophical Transactions of the Royal Society B 290, 181–197 (1980)
3. Perceptual organization and the judgment of brightness. Adelson EH. Science 262, 2042–2044 (1993)
4. Independent effects of lighting, orientation, and stereopsis on the hollow-face illusion. Hill H, Bruce V. Perception 22, 887–897 (1993)
5. Cultural differences in the perception of geometric illusions. Segall MH, Campbell DT, Herskovits MJ. Science 139, 769–771 (1963)
6. The representation of perceived angular size in human primary visual cortex. Murray SO, Boyaci H, Kersten D. Nature Neuroscience 9, 429–434 (2006)
7. The surface area of human V1 predicts the subjective experience of object size. Schwarzkopf DS, Song C, Rees G. Nature Neuroscience 14, 28–30 (2011)
8. Illusory contours and cortical neuron responses. von der Heydt R, Peterhans E, Baumgartner G. Science 224, 1260–1262 (1984)
9. Neural basis for a powerful static motion illusion. Conway BR, Kitaoka A, Yazdanbakhsh A, Pack CC, Livingstone MS. Journal of Neuroscience 25, 5651–5656 (2005)
10. Illusory motion reproduced by deep neural networks trained for prediction. Watanabe E, Kitaoka A, Sakamoto K, Yasugi M, Tanaka K. Frontiers in Psychology 9, 345 (2018)
11. Multistable phenomena: changing views in perception. Leopold DA, Logothetis NK. Trends in Cognitive Sciences 3, 254–264 (1999)