Why we see faces when they're not really there (with poll!)

i-6038f51ebea515e43aab4cd4fc7a184f-smokeface.jpgThis photo of the World Trade Center burning on September 11 attracted a lot of attention for a curious pattern in the smoke. Was it the face of God? Satan?

Of course it was just a random pattern in the smoke that briefly seemed to look like a face, but many people were not convinced by appeals to rationality. Why is it that people seem to see faces so many places that there aren't actually faces? The New York Times has an excellent article describing how and when we see faces.

As we've reported here before, even if you view a photo or painting at an extreme oblique angle, you can still see a face:

i-d22378c3220a9ded7d8505297746d101-jimfat.jpgi-ccf5ddd2fd467d302a213e3fa7bae776-norathin.jpg

The Times article describes the research of Pawan Sinha, who has been working for some time developing computer models for recognizing faces:

As the computer amassed the information, it was able to discover relationships that were of great significance to almost all faces, but very few nonfaces. "These turn out to be very simple relationships, things like the eyes are always darker than the forehead, and the mouth is darker than the cheeks," Dr. Sinha said. "If you put together about 12 of these relationships, you get a template that you can use to locate a face."

When he presented human subjects with blurry face images, containing only 12 by 14 pixels' worth of visual information, they performed similarly well, recognizing 75 percent of the face images accurately. This suggests that like the computer, the human brain processes faces holistically, like coherent landscapes, rather than one feature at a time.

Hmmm... 12 by 14 pixels. Let's see if we can reproduce that result here. Take a look at these pictures:

i-36491118462baccd4d2ee14464482f77-tinypix.gif

Which one is a face?

After we've gotten some results, I'll reveal the correct answer in the comments.

Tags

More like this

Decades ago, the tendency to see faces that weren't really there was taken to be some kind of malfunction of the mind or brain.

Back then, nobody wondered whether there was any survival value to this.

If you extend the study to include animal faces, then it all starts to make sense, since we humans are meat sources for large animals of the fangy-clawy persuasion. The ability to see predators lurking in the shadows or nearly concealed in foliage has huge survival value, so great a value that a sizeable false positive rate is an acceptable cost.

Incidentally, you should now be able to make sense of those vertical black stripes on the faces of cheetahs. They are not there to cut glare, as they effect would be less than a hundredth of a decibel. They conceal the eyes in tall grass, making the eyes hard for prey species to see as a pair of eyes watching them.

Wow!

Our readers have impressive accuracy. Yes, the picture on the right is the human face. We might have to try a Casual Friday where we go for even lower resolution.

Bonus question: What notable figure is represented in the picture on the right?

Even harder bonus question: What is portrayed in the picture on the left?

This was a lot easier when I took my glasses off - the picture on the right suddenly looked much more like a face. I think the hard square edges make it more difficult to see, and the blurriness that comes with removing my glasses makes it more natural-looking.

The picture on the left sort of looks like a monkey or ape with its face tilted.

cephyn:

Wrong on the face. I'll give you partial credit for "flower," but there's more to the picture on the left (which is, I assume what you meant when you said "pic on right").

Jennifer:

That works, as does stepping back from the monitor (but both are sort of like "cheating"). When the pictures are displayed at their original, tiny resolution, they are *much* easier. I'll post them once I've given folks a bit more time to answer the bonus question.

Hmmm...could the pic on the right be Hillary Clinton?

OMG I did what you did, Jennifer and that totally works! AND it looks like hilary to me now too.

the LEFT one does look like a monkey face too, i can see that, but i still see flower more.

I'm using Vista and when the pictures on this page are displayed in the little preview screen in the taskbar that Vista gives you (at the bottom of your screen), it's dead easy to see that the face on the right is a face ... The left-side image is still tricky, but might be a flower? Sorry for cheating - it wasn't deliberate (I chose right-hand image as face before noticing this)! What does the Vista preview do to the images to make them easier to decode? Is it the same thing as when people take off their glasses?
Christian

sorry, just realised flower has already been guessed - using my Vista Hack there seems to be like a square thing round the middle of the flower

They both looked like they could be faces to me when I first looked, and I plumped for left. However, after reading Christian's comment, I resized to 60% (using Opera, so images resize nicely with the page) and the face on the right was very, very obvious.

I thought about Hilary even before having read the comment. I'd say it's her.

...or to be correct and precise: Hillary Rodham Clinton

The picture on the left looks like the monkey face you've used in the other test (so I voted "both" I'm afraid).

I'd guess the good ol' Mona Lisa for the face.

is the face Greta? no idea about the pic on the left.

Very simply, seeing faces that don't really exist is either the result of pariedolia or simulcra.

By David Group (not verified) on 16 Feb 2007 #permalink

I hate to break the news to Pawan Sinha, but this is nothing new. Bell Labs did exactly the same study back in the 1960s. The research was for the "video-phone" which everyone expected to be a reality back then, in just a few years. The study was to see just how degraded an image could be and still be recognizable. I recall as a child having a pamphlet put out by the phone company about this, with a 10x10 pixilated image that was easily recognizable as Abe Lincoln.

To follow up on Matt's comment about the Bell Labs study: in the early 1970s there was an article in a major news publication that included the pixelated Lincoln image. It was so long ago that I don't remember the publication but I believe it was something like Time or Newsweek.