Vision scientists are obsessed with illusions.
This isn't because illusions shatter the sense that we have direct access to the physical properties of the external world. And it isn't because illusions give us the feeling — itself a deception — that for one brief moment we've transcended appearance to understand things as they truly are.
At least, these aren't the only reasons vision scientists are obsessed with illusions. They're also obsessed with illusions because they can teach us about the mundane, nonillusory percepts that help us navigate everyday life so effectively. Hermann von Helmholtz, the noted 19th century physician and scientist, had it right:
"It is just those cases that are not in accordance with reality which are particularly instructive for discovering the laws of the processes by which normal perception originates."
That's because there are lots of ways to get things (mostly) right, but fewer ways to get things wrong in just the right way so as to produce a particular illusion.
Consider color printing. When your printer gets things right, it's hard to know which inks make up its basic palette. It's when things go wrong — a misaligned edge, for example, or a half-empty cartridge — that you might spot the underlying components of cyan, yellow and magenta. Similarly for vision: Illusions can reveal the underlying components of veridical perception.
Ben Backus, a professor at SUNY's College of Optometry, is a vision scientist. And he's no exception to the rule: He, too, is obsessed with illusions. He's published scientific papers about them, and without having ever set foot in his office, I'll wager he has more than one prominently displayed. (I asked and the number is closer to 20!) Last week he helped me explain a color illusion. This week he's back to explain a motion illusion that appeared at the end of last week's post: the Rotating Snakes illusion.
About 10 years ago, Backus and Ipek Oruc, then a postdoc in his lab and now a professor at the University of British Columbia, became fascinated with the Rotating Snakes illusion. The pair conducted a series of experiments that culminated in a paper published in the Journal of Vision, which explains how basic aspects of visual perception give rise to the anomalous perception of motion.
Like a detective, the vision scientist knows how to recognize clues, especially when they're staring him in the face. Says Backus:
"The first thing to notice about Rotating Snakes is that the motion grinds to a halt if you stare at just one part of the image. On the other hand, it keeps going if you keep looking around. So eye movements are important. However, the illusory motion is not actually caused by the motion of the image across your retina. Instead, what matters is that the image be at different positions on the retina from time to time. In fact, a briefly flashed image of Rotating Snakes appears to rotate like crazy, even if there's no time to move your eye across the image."
This is a first clue to what's driving the illusion of motion. Here's another:
"Notice that any given disk within the image always rotates in the same direction. Kitaoka, who created this version of the illusion, has included both clockwise and counter-clockwise rotating disks, but if you pay attention, you will notice that each disk in the image is faithful to its own direction of rotation. The illusory motion is always in the same direction as this progression of colors: black, blue, white, yellow."
And that's a second clue. So the illusion has something to do with the visual processing that occurs when the image first hits the retina at a particular location, sending signals to the brain's visual cortex, and also with the progression of colors (black, blue, white, yellow), which determines the direction of rotation. These clues tell us something about the visual processes that give rise to the illusion, but what?
There are (at least) three important facts about the visual system at work in this illusion. Backus explains:
"The first fact is that neurons fire very rapidly whenever the image on the retina changes, then quickly slow down. You might think this would cause the image to fade, but that's not what happens. The fast initial firing rates allow visual perception to be fast. Then neurons go into 'maintenance mode' until the image changes again. This decrease in firing rate has been called 'adaptation' in the sense that neurons 'adapt (stop responding) over time, but it's better to think of it as an efficient coding scheme that saves energy by not making neurons fire more than necessary.
One implication of this fact is that when Rotating Snakes first projects onto a particular location on the retina, there's a lot of neural activity. That activity is highest for the regions with the highest contrast (black and white), and lower for regions with moderate contrast (blue and yellow). The activity for the high contrast regions also dies down (i.e., adapts) at a faster rate than the activity for the moderate contrast regions. The resulting differences in firing and adaptation turn out to be crucial:
"The difference in the rate at which neurons adapt in the black vs. blue regions, and in the white vs. yellow regions, causes a shift in the location of peak neural activity from black towards blue, and from white towards yellow. There could still be more activity at black and white than there is at blue and yellow, but the ratio changes, and this causes movement in the location of the center of mass of the neural activity, which is detected by motion mechanisms."
This shift in the ratio of activity across the image generates the perception of motion, even though the sharp edges of the colored sections are decidedly static:
"The second fact is that the visual system decomposes images into separate representations at a variety of spatial scales, from coarse to fine. Motion is measured separately within each of these representations before they are recombined. In Rotating Snakes, we see motion even though the fine details (the crisp edges) aren't moving, because motion detectors at intermediate spatial scales get activated."
In other words, it's predominantly sensors at a coarser scale that generate the illusion of motion, and they do so despite the fact that motion sensors at a finer scale are mostly keeping mum.
A final fact about the visual system helps explain why we seem to perceive rotating disks, and not individual, shifting patches:
"The third critical fact is that large scale 'global' motions, like disk rotations, have their own separate detectors at a secondary stage of processing within the visual system. These detectors look for larger patterns of motion in the image, and they are very sensitive. Thus, a small amount of illusory motion at many different places in the disk can cause the entire disk to rotate — provided the motion is consistent throughout the disk. There's hardly any illusory rotation in a disk if some of it is colored black-blue-white-yellow in the clockwise direction, and some of it is colored in the opposite, counter-clockwise direction."
And that, in a nutshell, is how some cleverly arranged blobs of color can generate a compelling (though illusory) sense of motion.
But you'd be wrong to conclude that the visual system is sloppy or easily fooled. In fact, the aspects of the visual system that give rise to the illusion are pretty darned smart.
Adaptation, for example, is a way to save energy with a minimal loss in signal. And the ability to detect coherent motion from isolated patches makes us good at spotting moving objects under difficult conditions — imagine, for instance, a stealthy predator at dusk, approaching under the fragmented cover of leaves. True to Helmholtz's word, the illusion reveals "the processes by which normal perception originates," and how it's typically so veridical.
Finally, Rotating Snakes promises to teach us even more. Even for vision scientists, some mysteries remain. For example, Backus notes that for most people, a grayscale version of Rotating Snakes still rotates, but less than the color version. Nobody knows why for sure, but Backus speculates that neuronal firing rates for colored patches might not fall as quickly as they do for grayscale patches.
A second mystery is that not everyone sees illusory motion when they look at the illusion. Backus notes:
"These people have vision that is more accurate than most people: they see the thing correctly! There is no great difference in how they see motion — like everyone else they decompose images into different spatial scales and detect motion separately at each scale. Perhaps the rates at which their neurons adapt to new images keeps all the rates proportional to one another, so there is no change in the average location of the peak firing rate."
So vision scientists still have their work cut out for them. And in the meantime, Backus provides an exercise for readers:
"If you see motion in Rotating Snakes, you may be able to notice illusory motion in other places you hadn't noticed it before. Good places to look are brick walls and venetian blinds. These objects have repeating patterns that can resemble the black-dark-white-light sequence in Rotating Snakes."
To be sure, the rotating snakes (and the creeping walls and blinds) are all in your mind. But that doesn't make them any less compelling!
You can keep up with more of what Tania Lombrozo is thinking on Twitter: @TaniaLombrozo