sciphile.org

credit

Fig. 1: Schematic of nerve connections in the retina. The back of the retina is at the top of the diagram. Adapted from work made available by Peter Hartmann and Marc Gabriel Schmid at Wikimedia Commons under the CC By-SA 3.0 license.

credit

Fig. 2: Representative photoreceptor connections to neural bipolar cells, showing the field of photoreceptors that stimulate the bipolar neuron (blue) and the field that inhibits the bipolar neuron (red).

credit

Fig. 3: Television test pattern illusion. Although each rectangle is a uniform shade, lateral inhibition makes it appear that each is graded light to dark from left to right.

credit

Fig. 4: Color lateral inhibition. Although the magenta rectangle in the middle is a uniform color, it appears bluer and darker at the top edge, and redder and brighter at the bottom edge.

credit

Fig. 5: Craik-O'Brien-Cornsweet illusion. The outside edges of the rectangles are exactly the same shade.

credit

Fig. 6: Hermann grid. Dark spots appear at the intersections of the white lines due to lateral inhibition. The effect is more pronounced in peripheral vision.
 

credit

Fig. 7: Lightness contrast illusion by Edward H. Adelson. The squares marked A and B in the image are the same shade, as you can verify by using a mask to cover the rest of the image.

credit

Fig. 8: Woman's Tear by Henri Matisse. Our brains have evolved to emphasize edge processing and can easily recognize objects from simple line drawings.

neural connections

The human retina contains around 125 million photoreceptors at the back of the retina (about 7 million are cones for color vision and the remaining are rods for low light sensitivity), which are connected to a little over one million neural cells at the front. These neural cells, called ganglions, extend out of the eye to form the optic nerve, which carries visual signals to the brain.

Figure 1 shows a schematic of the retinal layers. Light enters the retina from the bottom in the diagram and passes through the (relatively transparent) layers of nerve cells to be absorbed by the photosensitive layer of rods and cones at the top. When light is absorbed by a rod or cone, it triggers an electrical signal to be transmitted out the front end of the photoreceptor. These signals are transferred to the ganglion nerve cells through bipolar cells (yellow in the diagram). On average about 100 photoreceptors are connected to each ganglion cell, but the number varies greatly over the surface of the retina. In the central region, which provides the most acute vision, there may be as few as five photoreceptors connected to each ganglion, while in the periphery there may be thousands. When the ganglion receives a signal above a certain threshold, it triggers an electrical signal to be sent to the brain. Since peripheral ganglions collect signals over a larger region of the retina (with thousands of photoreceptors), peripheral vision is more sensitive to low light levels. Astronomers learn that viewing a dim star in a telescope is best done by looking to the side of the star.

light perception

Have you ever wondered how it is that you see so naturally over a large range of light conditions? The difference in light intensity between noonday sun and candlelight can easily be a factor of a million. Nonetheless, our eyes and brain adjust so that we can see quite comfortably in both those extremes, and in fact our perceived brightness does not vary all that much (certainly not by a factor of a million). The reason for this perceived lightness constancy, derives from the nature of neural connections in the retina.

Many of the photoreceptors are connected directly through the bipolar cells to the ganglion cells. However, there are also lateral connections across photoreceptors made by the green horizontal cells in the diagram, and there are connections from neighboring bipolar cells to ganglion cells made by the grey amacrine cells in the diagram.

As one would expect, signals sent to the ganglion directly through the bipolar cells will excite the ganglion and cause it to fire. However, it is commonly believed most of the signals that pass from neighboring regions through these lateral connections will inhibit the ganglion from firing (figure 2). Thus, when the overall light intensity increases, the ganglion is more excited by the retinal region that is directly connected, but it is also more inhibited by the neighboring regions. The net result is that the firing rate of the ganglion does not change all that much. In this way, our perceived lightness does not vary significantly over a wide range of overall brightness.

The lateral inhibition from neighboring regions also has another important effect on visual perception: it enhances our perception of edges. If we focus on a light edge next to a dark border, the light edge appears even brighter than normal because the neighboring inhibition is not as great as it is for a light region sourounded by a light border. Similarly the dark region next to a light border is even darker than normal. Contrast is enhanced at edges where there is a change in brightness.

This effect may be part of the reason that the human brain has evolved to process edge information so effectively. As figure 8 demonstrates, we can identify a face from a few simple lines in a drawing, which look nothing like the continuously shaded image from a real face.

lightness illusions

Figure 3 shows a pattern reminiscent of television test patterns from the days before color television. (Click on the thumbnail to see a larger version.) In this pattern each rectangle is actually a uniform shade of grey, but the left edge of each rectangle (near the next darker rectangle) appears lighter, and the right edge of each rectangle (near the next lighter rectangle) appears darker. The contrast at the edges is enhanced due to lateral inhibition.

Figure 4 shows the same effect for color vision. The magenta bar in the center is a uniform color, but near the upper red region, the red sensitive cones in our eye are more inhibited and that side of the magenta bar looks less red and more blue (and darker). Near the lower blue region, the blue sensitive cones are more inhibited and that side of the magenta bar looks less blue and more red (and lighter). Check out our lesson on Color Perception for more details in color vision.

Figure 5 shows a famous illusion called the Craik1-O'Brien2-Cornsweet3 illusion. The right rectangle becomes gradually darker from left to right, while the left rectangle becomes gradually lighter from right to left. The left and right outside edges of the image are actually the same shade. However, because our brains have evolved to pay attention to contrasting edges and not pay much attention to gradual changes in shade, we perceive the right rectangle to be a more or less uniform light shade, and we perceive the left rectangle to be a more or less uniform dark shade. Note: some computer screens change shade and color depending on the angle of view; this illusion may be better printed on paper (see the attached file at bottom). Be sure also to visit Michael Bach's interactive demo and Edward Adelson's animation of the Craik-O'Brien-Cornsweet effect for even more dramatic versions.

Figure 6 shows an illusion known as the Hermann Grid, discovered by Ludimar Hermann in 1870.4 The image is a uniform black background with a field of white crossing lines superimposed. Since the intersections are surrounded by brighter regions than the centers of the lines, the intersections are subject to greater lateral inhibition, and they appear darker. The effect is stronger for peripheral vision, where the ganglion cells collect signals over larger regions, than for central vision. Thus, when you stare directly at an intersection, the dark spot may disappear.

Some of the best lightness illusions take advantage of processes in the brain, as well as lateral inhibition in the retina. A region of medium shading can be interpreted either as a highly reflective surface under dim illumination or as a low reflectance surface under bright illumination. Fooling your brain about this choice can lead to some very powerful illusions such as the one shown in figure 7, by Edward Adelson at MIT. Be sure to check out other wonderful illusions based on Adelson's work at the link in further resources below.

further resources

The article "Why We See What We Do"5 at http://www.americanscientist.org/issues/pub/why-we-see-what-we-do/1 takes the discussion of lightness contrast well beyond the simple explanation in this lesson to include some effects of visual processing in the brain that can enhance or override the effects of processes in the retina. Figure 4 in that article demonstrates a particularly effective version of the Craik-O'Brien-Cornsweet illusion.

A number of fabulous lightness illusions based on work by Edward H. Adelson at MIT can be found at http://web.mit.edu/persci/gaz/. My personal favorite is the Kaffka Ring.

• 1. Craik, K.J.W. The nature of psychology. (Sherwood SL, ed). Cambridge, UK: Cambridge UP, 1966.
• 2. O’Brien, V. "Contrast by contour-enhancement." Am J Psych 72 1959:299–300.
• 3. Cornsweet, T.N. Visual perception. New York: Academic, 1970.
• 4. Hermann, L. "Eine Erscheinung simultanen Contrastes." Pflügers Archiv für die gesamte Physiologie 3 1870: 13–15.
• 5. Purves, D., Lotto, R.B., and Nundy, S. American Scientist 90(3) May-June 2002: 236-243.