Selection and Perception

A remarkable discovery about visual perception, and about sensory systems more generally, is that there are deep and elegant principles that explain, sometimes with mathematical precision, a diverse array of apparently unrelated phenomena. Two such principles are the principle of generic views and the principle of satisficing utility.

When we view a painting, such as M.C. Escher's Relativity, why do we see the straight lines in the painting as straight lines in three dimensions? After all there are, in principle, an infinite number of other three-dimensional interpretations. One could, for instance, interpret a straight line in the painting as a circle in three-dimensions that is seen edge on. Or one could interpret it as sinusoidal wiggle, again seen edge on.

Although these other interpretations are logically possible, they are visually implausible. If in fact one were viewing a circle, and the image at one's eye happened to be a straight line, then if one moved ever so slightly the line at the eye would change to an ellipse. If one were viewing a wiggle, and the image was again a straight line, then a slight move would reveal the wiggle. A small change in viewpoint would make a qualitative change in the image one sees. From a generic view, a circle or wiggle do not appear like a straight line.

However, if in fact one were viewing a straight line in three dimensions, then the image at the eye would remain a straight line from almost every viewpoint, i.e., from a generic view. (The only exception would be viewpoints in which the line happens to look like a point.)

Human vision prefers interpretations that imply a generic view. This principle explains many visual phenomena. For instance, when the endpoint of one line appears to touch another line, we assume that they touch in three dimensions. If they did not, then we would be looking at the lines from a non-generic viewpoint in which the endpoint just happened to coincide with the image of the other line. Our commitment to this principle is so strong that when it leads us astray, as it does when we see impossible waterways wending their way through Escher's Waterfall, we are unable to see other interpretations that violate the principle but would avoid the impossibilities.

Why do certain dragonflies prefer pools of oil to pools of water, and pay the ultimate price for their choice? Why do certain male beetles try to mate with beer bottles, and forsake available females? Why do our eyes detect squirrels more easily than handguns? Why is a face with limbal rings around its irises more attractive than the same face without rings? Each is a consequence of the principle of satisficing utility.

Dragonflies need to lay eggs in water. Their visual systems evolved to detect horizontal polarization of light as a way to find bodies of water suitable for oviposition. In the environment in which they evolved, this simple trick worked well. But another species has altered their environment. H. sapiens learned to consume oil, and on occasion leaves slicks or pools of oil, which happen to polarize light more strongly than water. Horvath and Zeil find that the oil is a supernormal stimulus to the dragonfly, tempting it more strongly than water, and leading to the demise of all who follow temptation. Kriska and colleagues report that Mayflies face a similar problem with certain asphalt roads that strongly polarize light, leading them to lay eggs where they are doomed to die. Oil slicks and asphalt roads are ecological traps for dragonflies and mayflies. Their visual systems evolved a satisficing solution for finding water which fell far short of guaranteeing a genuine find but which, in a world before oil slicks and asphalt roads, was useful enough to let them survive.

A certain jewel beetle in the Australian desert has wing casings that are dimpled, glossy and brown. The males fly around searching for eligible females, a strategy that has served the species well for thousands of years. That is, until H. sapiens started dumping empty beer bottles in the desert that are bumpy, glossy, and just the right shade of brown. Gwynne and Rentz discovered that male beetles find the bottles far more attractive than real females, and swarm the bottles attempting to mate. Their visual systems evolved a satisficing solution for finding females which fell far short of guaranteeing a genuine find but which, in a world before beer bottles, was useful enough to let them survive.

H. sapiens evolved to detect and monitor objects that could be dangerous. In the environment in which H. sapiens evolved, animate objects were a primary source of potential danger. Today inanimate objects such as guns and cars are also dangerous. But psychophysical studies by New, Cosmides and Tooby show that H. sapiens is faster to detect a squirrel than to detect a gun or car. Our visual systems evolved a satisficing solution for finding dangerous objects which fell short of guaranteeing a genuine find but which, in a world before guns and cars, was useful enough to let us survive.

H. sapiens evolved to be more attracted to conspecifics that are more reproductively fit. In the environment in which H. sapiens evolved, a larger and more pronounced limbal ring around the iris was a good probabilistic indicator of youth and health, and therefore of reproductive fitness. Peshek and colleagues found that human observers find a face with limbal rings around its irises more attractive than the same face without rings. (Pronounced limbal rings can be seen in the famous National Geographic photograph of an Afghan girl.) We have evolved a satisficing solution for finding reproductively fit conspecifics, a solution which worked well in the environment in which it evolved. Now, however, contact lenses are sold which endow the wearer with artificially larger and more pronounced limbal rings, but which nevertheless trick our visual systems and make the wearer look more attractive.

The principle of generic views and the principle of satisficing utility each explain a diverse array of apparently unrelated perceptual phenomena. Perhaps as our understanding of sensory systems advances we will find that the principle of generic views is itself a consequence of the principle of satisficing utility.