Animal vision has not evolved as one might think. In contrast to the invention of photography and film—which began with the first black-and-white daguerreotypes in 1839, then added color in 1861, and finally motion in 1891—motion perception in animals appears to have evolved long before color vision. Indeed, as vision researcher Gordon Walls declared in 1942, perceiving motion is one of the most ancient and primitive forms of vision.
Even the humble housefly, which can only distinguish four to six different colors, is remarkably good at seeing motion. Try to swat a fly with your hand, and it will be gone long before you even get close. (The best way is to clap your hands above it so that it flies up between your hands. Wear gloves.) Oddly, however, while a fly is quick to register these fast movements, it cannot recognize slow movement at all. Move your hand very, very slowly toward a fly, and you can tap its back before it knows that you are there.
As good as animals are at detecting motion, they can also be fooled. Seeing the errors that a system makes can help us to understand how that system works normally.
Much of the early research on motion perception was performed on insects,1 but similar results have been found for a huge range of species, from fishes to birds to mammals. Frogs, which eat insects, respond to small, rapidly moving prey, as well as to overall dimming or darkening that likely signals an approaching predator, but they often ignore stationary objects, perhaps because they cannot see them.2
Mammals are likewise tuned in to motion. Although many people believe that it is the bright red color of the matador’s cape that enrages the bull, the popular TV program Mythbusters found that the color made no difference; it was the motion of the cape’s fabric that mattered. Red, blue, and white capes got equal, half-hearted attacks when they were motionless, but waving the capes elicited an aggressive charge. In fact, most mammals, including domestic and big cats, deer, cattle, and dogs, see a limited range of colors. Apes may have evolved color vision in order to find the ripe fruit among green leaves (see “The Rainbow Connection,” The Scientist, October 2014), but lions eat other mammals, most of which have evolved to match their surroundings, rendering color vision useless in finding prey. When a gazelle runs away, however, it becomes a strong stimulus for the lion’s keen motion vision. It’s no wonder that young deer will often freeze when they sense danger. Correspondingly, prey animals would find color vision of little use, but they are extremely good at seeing the motion of an approaching predator.
But as good as animals are at detecting motion, they can also be fooled. I study visual illusions of motion because seeing the errors that a system makes can help us to understand how that system works normally. Visual perception goes far beyond our retinal images, which provide only partial sensory information. We use our knowledge and expectations of the world to fill in the gaps, for instance, when an object is partly hidden. Ambiguous illusions that can be interpreted in two different ways, but not both ways at the same time, can also shed light on how we perceive the world around us.
Illusions of movement
Visual movement can be thought of as a change in brightness, or luminance, over space and time. A white spot that glides across a black screen shows real movement. If the same spot jumps back and forth between two positions, or makes a series of intermittent forward jumps, the brain can still perceive movement. Small, fast jumps give the smoothest impression of movement, but even large, slow jumps give a strong impression that the spot is, in fact, moving across the screen.
Why does the visual system treat this jumping dot as a single object in motion, instead of seeing one spot disappear while an unrelated spot appears nearby at the same instant? First, the brain usually treats “suspicious coincidences” as being more than coincidences: it is more likely that this is a single spot in motion rather than two separate events. Second, the visual system is tolerant of brief gaps in stimuli, filling in those gaps when necessary. This perception of apparent motion is, of course, the basis of the entire movie and TV industries, as viewers see a smooth motion picture when in reality they are simply watching a series of stationary stills.
We can pose a riddle to the visual system by presenting two apparent motions in opposite directions simultaneously. For example, an image of a white horse and an image of a black horse suddenly exchange positions. But you do not see each horse independently changing in color. Rather, you see the horses jumping from one location to the other. The coincidence is too great, and, instead of two independent events, the visual system economically infers a single event: the jumping horse.
But which horse jumps? The answer depends on the context. On a dark background, the white horse appears to jump back and forth; on a light background, the black horse appears to move. In other words, the horse with the higher contrast wins. This is because the strength of a motion signal in the brain of the observer is equal to the product of the contrast of each horse against the background color, a measure called motion energy.3 Interestingly, if contrast is held constant, the color of the horses makes no difference because color has little or no input into the motion pathways of the brain.4
Contrast can also explain why a black or white object on a background of the opposite color seems to move faster than a gray object on a gray background, and why cars appear to move more slowly in the fog. Indeed, we tend to judge motion not in absolute terms, but relative to the background: the perceived strength and speed of motion depend on the contrast of the moving object against its surroundings.5 In fact, a driver partly judges his own speed by the rate at which landmarks such as trees flash past him. In the fog, the trees appear slowed down, so he underestimates the speed of all cars, including his own, with potentially disastrous consequences.6
Combining movement and changes in contrast results in an even more complex outcome. Suppose that a black spot on a medium-gray background makes a small jump to the right—a total distance much smaller than the diameter of the spot itself—and, at the same time, instantaneously changes to white. Instead of seeing a slight motion to the right, one sees something quite unexpected: the spot appears to move to the left, toward the starting position and opposite to the physical displacement.
MOTION IN CONTEXT: A yellow bug and a red bug both fly around in perfect clockwise circles of the same size, though the red bug moves much more rapidly. When a background is added that also circles clockwise, the yellow bug’s orbit, which syncs up with the motion of the background, seems to shrink to about half the size of the red bug's orbit.
This effect, known as reverse phi, is particularly strong in peripheral vision: if someone fixes his gaze on a small stationary cross and observes the moving spot out of the corner of his eye, the backwards leap will be even more pronounced.7,8 Once again, this phenomenon is consistent with the idea that perceived motion depends on motion energy, or the product of the contrasts of moving objects.3 If the spot makes a long series of jumps to the right, changing between black and white on each jump, one still sees steady motion to the left, but after a while the observer will recognize that, paradoxically, the spot is now farther to the right, demonstrating that position and motion are signaled independently.
Why we are fooled
The phenomena described above are “low-level” illusions that are probably based on “bottom-up” sensory signals from brain cells in the visual system that are specialized to detect motion. Normally, sensory information agrees. If a cat is partly hidden behind a tree, for example, all the cues of color, shadows, and texture tell the same story—that the hidden part of the cat exists out of view behind the tree. The brain acts like a judge, confirming the same story as told by independent witnesses. The brain also strengthens this verdict with “top-down” information based upon prior learning: if the cat’s whiskers stick out on one side of the tree, and its tail on the other, the brain automatically “fills in” that there is a continuous cat partly hidden by the tree, not two unrelated cat bits. This interpolation process, called visual amodal completion, starts from a representation of the visible features of the stimulus in early visual cortex, probably an area called V1, and ends with a completed representation of the stimulus in the inferior temporal cortex.9 Jay Hegdé of the University of Minnesota and colleagues even found two regions in the object-processing pathways of the brain that actually responded more strongly to partly hidden objects than to complete ones.10
Visual object recognition thus involves two stages: a bottom-up inputting of perceptual information, and a top-down memory stage in which perceptual information is matched with an object’s stored representation. Tomoya Taminato of Tohoku University School of Medicine in Japan and colleagues last year presented volunteers with blurry pictures that gradually became sharper. Observers responded once when they could guess the identity of the object in the image, representing the perception stage, and a second time when they were certain of the identity, the memory stage. Their results attributed the perception stage to the right medial occipitotemporal region of the brain, and the memory stage to the posterior part of the rostral medial frontal cortex.11
If a cat’s whiskers stick out on one side of the tree, and its tail on the other, the brain automatically “fills in” that there is a continuous cat partly hidden by the tree, not two unrelated cat bits.
Visualizing motion is similarly subject to both bottom-up and top-down processes. Reverse phi, in which an object that changes contrast as it travels is viewed as moving in the reverse direction, is a bottom-up illusion that happens early in the brain’s visual processing pathway. Researchers have tracked the origin of this illusion to V1 cells, which in awake monkeys respond to the reverse phi illusion in the same way they respond to backwards-moving objects.12 Meanwhile, top-down processes predict what objects these signals probably represent, based upon memory and previous learning. Object parsing, for example, is a process that guides perception by deciding what objects are likely to be present based upon prior knowledge of the world.13
Consider the closing blades of a pair of scissors. The intersection itself is not an object; only the blades are. This distinction is not lost on the visual system. Observers make 10 times the tracking errors—their eyes deviating from the target—when they attempt to follow a sliding rather than a rigid intersection.14 Although you can sense the movement of a sliding intersection, you do not interpret it as an object.
This phenomenon stems from the fact that smooth eye movements require a smoothly moving target. Move your thumb from side to side in front of you and ask a friend to follow your thumb with his eyes. Watch his eyes and you will see them move smoothly from side to side. Now hold up both your thumbs a yard apart and ask him to move his eyes smoothly from one stationary thumb to the other. He cannot do it! You will see his eyes moving in a series of jerky eye movements called saccades. This shows that a moving object is necessary to drive smooth-pursuit eye movements.
Visual signals flow forward from the visual cortex at the back of the brain, then travel along the ventral stream for the decision about what objects are present, and also up along the dorsal stream to the medial temporal area, which analyzes motion. Finally, the nerve signals travel forward to the frontal eye fields that control eye movements. A sliding intersection is not parsed as a real object, and it cannot support smooth eye movements.
The visual system can also flip between local and global motions, but it cannot see both at once. The brain considers incompatible interpretations—Are there many small groups, or a few large groups?—and adopts them in alternation, but never both at the same time. The shape and spacing of spots on a screen, the duration and position of your fixations, and other factors can all influence which percept you see.
LOCAL OR GLOBAL: At first, viewers see pairs of spots, each pair rotating about their common center. But if you watch for a while, you will suddenly see it reorganize into two larger squares on top, or eight interdigitating octagons on the bottom. The visual system can alternate between either percept, but it cannot see both at once.
Motion can shift an object’s perceived position. If an image of an upright cross flashes briefly on a textured wheel that is rotating clockwise, the cross itself will appear to be tilted clockwise, and it sometimes even looks distorted. Notably, only the motion of the background that occurs after the flash can drag the cross along: motion beforehand has no effect.15
In sum, illusions teach us that perception goes far beyond the information picked up by our senses. Perception is an indirect, interpretive top-down process that is not driven simply by stimulus patterns, but is instead a dynamic, active search for the best interpretation of the available sensory data.
Stuart Anstis is a professor of psychology at the University of California, San Diego, and a visiting fellow at Pembroke College in Oxford, U.K. Working with international collaborators, he has published some 170 articles on visual perception.
All videos courtesy of Stuart Anstis.
- W. Reichardt, “Autocorrelation, a principle for the sensory evaluation of sensory information by the central nervous system,” In: Sensory Communication, W.A. Rosenblith, ed., Cambridge, MA: The MIT Press, 1961, pp. 303-18.
- J.Y. Lettvin et al., “What the frog’s eye tells the frog’s brain,” Proc IRE, 47:1940-51, 1959.
- E.H. Adelson, J.R. Bergen, “Spatiotemporal energy models for the perception of motion,” J Opt Soc Am A, 2:284-99, 1985.
- P. Cavanagh, S.M. Anstis, “The contribution of color to motion in normal and color-deficient observers,” Vision Res, 31:2109-48, 1991.
- S. Anstis, C. Casco, “Induced movement: The flying bluebottle illusion,” J Vision, 6:1087-92, 2006.
- M. Green et al., Forensic Vision with Application to Highway Safety, Tucson, AZ: Lawyers & Judges Publishing Co, 2008.
- S.M. Anstis, “Phi movement as a subtraction process,” Vision Res, 10:1411-30, 1970.
- S.M. Anstis, B.J. Rogers, “Illusory reversal of visual depth and movement during changes of contrast,” Vision Res, 15:957-61, 1975.
- S. Weigelt et al., “Separate cortical stages in amodal completion revealed by functional magnetic resonance adaptation,” BMC Neurosci, 8:70, 2007.
- J. Hegdé et al., “Preferential responses to occluded objects in the human visual cortex,” J Vision, 8:16.1-16.16, 2008.
- T. Taminato et al., “Neuronal substrates characterizing two stages in visual object recognition,” Neurosci Res, 89:61-68, 2014.
- M. Livingstone, B. Conway, “Responses of V1 neurons to reverse phi stimuli,” J Vision, 2:127, 2002.
- S. Anstis, “Imperceptible intersections: The chopstick illusion,” in AI and the Eye, A. Blake and T. Troscianko, eds., London: John Wiley and Sons, 1990, pp.105-17.
- S. Anstis, H. Ito, “Eyes pursue moving objects, not retinal motion signals,” Perception, 39:1408-11, 2010.
- P. Cavanagh, S. Anstis, “The flash grab effect,” Vision Res, 91:8-20, 2013.
Clarification (September 4, 2015): The original version of this story stated that most mammals are color-blind. There are different kinds of color-blindness, and many mammals do see some colors. This point has been clarified in the text. The Scientist regrets any confusion.