Skip to content
 

What makes a face attractive?

Susan sent me this link and asked for my thoughts about some related question which, unfortunately, I’ve forgotten. That’s what happens when you wait over a month to answer an email. Anyway, the website is cute, much cuter than ours. We clearly have a lot of work to do.

The work looks interesting. I wonder about time trends. It’s my impression that characters in old TV shows were often pretty ugly (for example, consider the guy in Mr. Ed), but now they all seem pretty attractive. But maybe some of that is technology–cameras are better so they don’t have to slap on all the greasepaint or whatever.

10 Comments

  1. greg says:

    I really like the face-averaging stuff, but I'm very skeptical of the attractiveness claims. In psychology they tend to be remembered as "symmetric faces are more attractive", but I doubt it's that simple.

    The "Attractive Faces Are Only Average" paper compares composites to the averages of ratings for individual faces and finds differences corresponding to less than a point on a 5-point scale.

    This implies that several of the individual faces were rated higher than the composite face. And anecdotally, stereotypically attractive people have striking assymetries (Angelina Joli for example). And rarely are the composite faces strikingly attractive. So it doesn't seem clear to me that evolution favors mating with people with symmetric faces.

    On the other hand, it's quite possible that faces are represented as averages of faces (for example, the concept of eigenfaces as used in computer vision) and people seem to have slight preferences for things that are easier to process. This makes me think that babies and adults seem to like symmetrical faces for processing reasons rather than attractiveness ones.

  2. Richard D. Morey says:

    There's a fairly mundane explanation for the finding that "averaging" faces gives you more attractive pictures. When you average, all the blemishes in the skin get averaged out, making the face look more attractive.

    It seems to me that averaging pixels is the wrong way to go about creating an "average" face. Instead, one could measure the distances between various features and average those, or something of the sort.

  3. greg says:

    Richard — They do do that as well, but they tend to morph actual faces so that the average faces don't look too computer-generated. There's a literature of measuring the faces of known attractive people. Denzel Washington is apparently very symmetrical as is Elizabeth Taylor.

    Also there's a fun demonstration site here that lets you make average faces with your own photos if you want: http://www.faceresearch.org/demos/average

  4. greg says:

    Richard — They do do that as well, but they tend to morph actual faces so that the average faces don't look too computer-generated. There's a literature of measuring the faces of known attractive people. Denzel Washington is apparently very symmetrical as is Elizabeth Taylor.

    Also there's a fun demonstration site here that lets you make average faces with your own photos if you want: http://www.faceresearch.org/demos/average

  5. Keith O'Rourke says:

    My wife is an artist and she once told me the less lines you require to draw a face the more attractive it will be.

    Seems related ….

  6. c. says:

    Why do the hairstyles change (albeit subtly) in the face-averaging process? That seems to be letting im a confounding factor…

  7. Thom says:

    I think that averageness probably is part of attractiveness, but (as well as other factors) I've always been skeptical about linear models. Average faces should be less distinctive … but extremely attractive faces will be more distinctive so I think there is some tension at the extremes.

    It is also worth pointing out that average faces aren't necessarily easier to process. It depends on the task … it is easier to tell something is a face if it is average/proto-typical, but harder to tell whose face (because discrimination is easier between dissimilar items). Arguably the more important (and more difficult) task is within-category discrimination.

    One link to statistics is that the early work on 'averageness' and attractiveness is due to Galton (who made composite faces from pictures that looked more attractive than the individual contributors; his materials suffered from the problem that blemishes were smoothed out by the process).

  8. me says:

    why is everyone tip toeing around the plain truth? Are you all afraid to just mention that a broad nose , big lips and a fat face- as symmetric as it may be- is still UGLY!Universally! I'm sure the babies will cry if you show a face with those specific features.

  9. S Stonehill says:

    @ Morley: Angelina Jolie's face is likely extremely symmetrical, despite her large lips. The symmetry isn't measured top to bottom, rather side to side – the left matching the right. This principle applies to the entire body.

  10. S Stonehill says:

    @ greg: Sorry, my last comment was directed towards your statement. Also, you said "So it doesn't seem clear to me that evolution favors mating with people with symmetric faces." I believe there is a strong evolutionary component. Though mostly in extreme cases, facial asymmetry often indicates genetic abnormalities.