Subscribe to Computing Intelligence

Tuesday, January 20, 2009

Top-down Processing in Visual Perception Part II: Faces

I started this series of posts a couple months ago with Part I on the definition and role of top-down processing. When I originally wrote that post, I had meant to expound upon the topic in a more timely fashion, but I clearly became distracted and forgot about it. If there are other topics which you think I have been neglecting of late, please do not hesitate to leave a comment and I will endeavour to correct such lapses.

As I was saying in the first part of this series, optical illusions are a great way to get one thinking about how one's perceptual system works. In one particular vein of optical illusions are those that 'jump' between interpretations, the most basic being that of the Necker Cube mentioned in the previous post. I think it is particularly revealing about our visual system, however, that when one surveys a large number of optical illusions of that nature, the vast majority are devoted, in at least one of their interpretations, to faces. One of the classic examples of this is shown in figure 1, in which both the back and side of a young lady's face are visible along with the direct side profile of an old woman.

Figure 1: Old lady and young woman

There is good reason for our focus on facial perception, as it is our primary method for recognizing other individuals in social interaction. The ability to differentiate between individuals is an exceedingly important aspect of social intelligence, as there would be, for example, no way without it to differentiate between cheaters and trustworthy members of a tribe. The supreme prevalence of our nuanced ability to analyse faces, however, is often discounted by people. Interestingly, there is a condition known as prosopagnosia in which sufferers lack the ability to distinguish individual faces. There is no problem with the person's sight, but rather faces look as indistinguishable from each other as any other body part (for example, if you could only see peoples' torsos, it would be quite difficult to correctly identify others. There would of course be certain indications like weight and muscle tone, but telling the difference between a pair of scrawny teen boys or flabby middle-aged business men would be awfully difficult). The fact that something can be so selectively lost is rather indicative of quite specialised neuronal processing involved in the identification and distinguishment of faces (although it may be that there are other cognitive impairments that are less obvious). Our predisposition to seeing faces in ambiguous images or in anthropomorphising objects most readily with the appearance of a face (figure 2, 3, , and 4) indicates just how greatly our brain tries to match incoming sensory data with the expecation of seeing a face.

Figure 2

Figure 3

Figure 4

Of course, there are other forms of anthropomorphism available, but the appearance of a face resounds more greatly within us and accentuates the illusion of life. The next post in this series will look at another area of top-down processing as well as some of the ramifications.

Continue reading in Part III: Artificial Edges.


cornucrapia said...

Maybe I have some form of that disorder, for the life of me I can't see the old woman

Mozglubov said...

The old lady's eye is the young lady's ear, while the young lady's nose is a bump (wart?) on the old lady's nose. The young lady's neck is the old lady's chin, and the necklace around the young lady's neck is the old lady's mouth.

Hopefully that helps. If you still cannot see it, I'll see if I can find anther one that might be easier to spot.