You’re the only eyewitness to a murder. The killer has vanished. In trying to track her down, the police come to you for help. At least according to detective novels, this is when a police artist sits down with you. Based on your descriptions of various facial features—her nose, her lips, her moustache, her chin—the artist comes up with a likeness.
How successful will this effort be? (All right, scratch the moustache). That depends on your ability to recall, separately and clearly, the features of someone you probably saw only briefly. Not easy. Think of the face of a good friend you meet every day. So what’s his nose like? Is it hooked? Straight? Flat? Wide? What about his chin, jaw, eyes?
I’m guessing that you would find it hard, as I do, to accurately describe each of these facial details. And this is someone you see daily. Imagine the difficulty, then, of deciding whether someone you saw fleetingly, likely in a moment of great stress, has thin eyebrows or ones like angry caterpillars.
On the other hand, you can easily visualize your friend’s whole face.
Likewise, even a brief glimpse has probably left in your mind a reasonable mugshot of the murderer. If you were offered several photographs, one of which is hers, you would probably pick out the right one.
Which raises the question: How is it you can recognize a whole face easily, but find it hard to recall individual features on that same face?
Because recognition is far easier than recall.
Several years ago, a professor at New Mexico State University, Victor Johnston, developed a system called Faceprints that underlined this point. Johnston used it to understand our ideas of beauty, but he also used it to reconstruct faces—just as the police asks you, eyewitness to murder, to do.
It went like this. Faceprints shows you 20 different faces on a screen. You rate each one on how close it is to your memory of the killer. Then Faceprints uses several techniques to vary each of five facial features—mouth, eyes, nose, hair and chin. (Why not the moustache, you ask? Well, there’s a reason). These variations generate more faces, each of which you rate again. This process continues until you think you’ve got close enough to the killer’s image.
Faceprints never asks you to choose between, say, different noses—because specific recall like that is difficult. Instead, it relies on your memory of a whole face. Exactly as you recognize your friends.
And it works. In one experiment, Johnston’s students watched a simulated crime and, at varying times afterward, used Faceprints to reconstruct the simulated criminal’s face. One witness produced, three days later, a computer-generated visage that was nearly indistinguishable from the actual simulated (yes) criminal’s face.
This is fascinating stuff that, if I think about it, is even familiar.
For example, if you ask me to name the Bollywood films whose posters I saw while on my morning waddle today, I can only remember one or two (go ahead, you try it). But as I pass each poster tomorrow, I will know without question that I have seen it before.
Yet, if we so easily recognize posters and faces, we must have some memory of them tucked away somewhere in those grey cells. So why can’t we pick out specifics about them? John McCrone has a nice metaphor for this in his book The Ape That Spoke. Trying to remember things, it’s “as if our memory banks were a darkened room full of objects with us blundering around with a penlight, hoping to bump into what we are looking for”.
But when we approach the original experience—pass a film poster, produce a Faceprints image that’s similar to the murderer’s face—it’s “as if all the lights in the room have suddenly been turned on”.
The memory comes flooding back, and you are certain it’s the right one.
That’s the power of recognition.
In fact, in trying to recall specifics, we often use this power of recognition without even thinking about it. This morning, someone stopped me to ask exactly where a particular bank was. I knew it was somewhere nearby, but I could not immediately say which building or street. Out of the blue I remembered some policemen I had waddled past, about a kilometre away. And there in my mental image of the scene, on the building behind the cops: the bank’s name. Clear as if I was staring at it.
For me, one simple lesson from all this is that our minds are more than just static networks of images and associations. They actually work because they are so. Find a hook into the network, and you can pull up whole baskets of memories. Even ones you can’t quite recall just now.
Though for the life of me I can’t explain why “moustache” has me thinking of…well, never mind.
Once a computer scientist, Dilip D’Souza now lives in Mumbai and writes for his dinners. A Matter of Numbers will explore the joy of mathematics, with occasional forays into other sciences. Comments are welcome at firstname.lastname@example.org
To read Dilip D’Souza’s previous columns, go to www.livemint.com/dilipdsouza