ADVERTISEMENT
See Inside August/September 2008

Shading Illusions: How 2-D Becomes 3-D in the Mind

Light cues influence the brain's processing of depth



Mitsuaki Iwago/Minden Pictures

THE VISUAL IMAGE is inherently ambiguous: an image of a person on the retina would be the same size for a dwarf seen from up close or a giant viewed from a dis­tance. Perception is partly a matter of using certain assumptions about the world to resolve such ambiguities. We can use illusions to uncover what the brain’s hidden rules and assumptions are. In this column, we consider illusions of shading.

In illustration a, the disks are ambiguous; you can see either the top row as ­con­vex spheres or “eggs,” lit from the left, and the bottom row as cavities—or vice ver­sa. This observation reveals that the vi­sual centers in the brain have a built-in sup­position that a single light source illuminates the entire image, which makes sense given that we evolved on a planet with one sun. By consciously shif­ting the light source from left to right, you can make the eggs and cavities switch places.

In illustration b, the image is even more compelling. Here the disks that are light on the top always look like eggs, and the ones that are light on the bottom are cavities. So we have uncovered another premise used by the visual system: it expects light to shine from above. You can verify this by turning the page upside down. All the eggs and cavities instantly switch places.

Amazingly, the brain’s assumption that light shines from above the head is preserved even when you rotate your head 180 degrees. Ask a friend to hold this page right side up for you. Then bend down and look between your legs at the page behind you. You will find that, again, the switch occurs, as if the sun is stuck to your head and shining upward from the floor. Signals from your body’s center of balance—the vestibu­lar system—guided by the positions of little stones in your ears called otoliths, travel to your visual centers to correct your picture of the world (so that the world continues to look upright) but do not correct for the location of the sun.

From this experiment we learn that despite the impression of seamless unity, vision is actually mediated by multiple parallel information-processing modules in the brain. Some of the modules connect to the vestibular system; however, the one that handles shape from shading does not. The reason might be that correcting an image for placement in so-called world-centered coordinates would be too computationally expensive and take too much time. Our ancestors generally kept their heads upright, so the brain could get away with this shortcut (or simplifying assumption). That is, our progenitors were able to raise babies to maturity often enough that no selection pressure acted to produce vestibular correction.

If you look at illustration c, you find that you can almost instantly mentally group all the eggs and segregate them from the cavities. As visual scientists discovered decades ago, only certain elementary features that are extracted early during visual processing “pop out” conspicuously and can be grouped in this manner. For example, your brain can discern a set of red dots in a background of green ones but cannot group smiles scat­tered among a backdrop of frowns. Color is thus a primitive feature that is extracted early, whereas a smile is not.

(It makes survival sense to be able to piece together fragments of similar color. A lion hidden behind a screen of green leaves is visible merely as gold fragments, but the visual brain assembles the pieces into a single, gold, lion-shaped form and warns: “Get out of here!” On the other hand, objects are not made up of smiles.)

The fact that you can group the eggs in c implies that shading information, like color, is extracted early in visual processing. This prediction was verified in recent years by recording activity in the neurons of monkeys and by conducting brain-imaging experiments in humans. Certain cells in the visual cortex fire when the observer sees eggs; others respond only to cavities. In illustration d, where the circles have the same luminance polarities as in c, you cannot perceive the grouping; this fact suggests the importance of perceived depth as a cue that is extracted early in visual processing.

Share this Article:

Comments

You must sign in or register as a ScientificAmerican.com member to submit a comment.
Scientific American Holiday Sale

Give a Gift &
Get a Gift - Free!

Give a 1 year subscription as low as $9.99

Subscribe Now! >

X

Email this Article

X