I have often wondered about how we recognize faces. We are so very good at recognition and so bad at describing faces in words. One person will say big nose, freckles, oval shape – and this is correct – but it is of no help to someone else in forming an image of the face. The way faces are recognized rarely rises to consciousness and therefore to useful for verbal description.

The current theory is that we build up by experience an ‘average face’. We then compare faces we encounter to this average face. It can be thought of as a mathematical ‘space’, a multi-dimensional face-space. The distance from the average corresponds to the amount of difference the face is from the average and the direction from the average face corresponds to the way/s the face differs from the average. So there is a center with arrows going out various distances in various directions to each known face.

This makes certain things clearer.

First, this is probably the reason that a person who has encountered only a few people from another racial group, has difficulty identifying people of that group. People are left stammering that all Chinese people look alike, knowing how stupid this sounds. They do not have a good average face for the unfamiliar group and therefore have difficulty establishing the differences between any face and the average.

Second, we use the average face as a sign post for beauty. The closer a face is to our average, the more attractive it is.

Third, the closer a face is to the average, the faster we recognize that it is a face. But knowing that a object is a face, the farther it is from average, the faster it is recognized as a particular face. These results depend on the density of faces in the face-space: high near the average and low far from the average.

Fourth, caricatures are recognized as the target although they are actually very, very different from the target face. But a good caricature is on the near exact direction from the average but just a much farther distance away then the target face. It is like following our imaginary arrow to the particular faces and then carrying on in the same direction for some distance.

Fifth, another interesting effect is archetypes. We can think of face-space having places (other than the average) of unusual high density of encountered faces surrounded by low density areas. Some particular face near the center of such a clump could come to stand for this type of face.

Sixth, this explains why we can sometimes hardly notice such prominent changes as new glasses, loss of a mustache etc. These are just not dimensions/directions in the face-space so they are not used to recognize faces.