What is it that makes a face look beautiful? What are the differences between very attractive and less appealing faces? For every historical period and every human culture, people have always had their own ideal of beauty. But this ideal has never been constant and is still subject to changes. In our research project we adopted an empirical approach and created prototypes for unattractive and attractive faces for each sex by using the morphing technique. For example, the prototype for an unattractive face (“unsexy face”) was created by blending together four faces that had previously been rated as very unattractive. The “sexy face” was created by blending together four of the most attractive faces, respectively (see report).
In order to find out the characteristic differences between attractive and unattractive faces, we presented pairs of one “sexy” and one “unsexy” image for both sexes to test subjects. The task was to report which facial features were perceived to be different between the two faces. For the results see the list below.
Characteristic features of the female “sexy face” in comparison to the “unsexy face”:
- Suntanned skin
- Narrower facial shape
- Less fat
- Fuller lips
- Slightly bigger distance of eyes
- Darker, narrower eye brows
- More, longer and darker lashes
- Higher cheek bones
- Narrower nose
- No eye rings
- Thinner lids
Characteristics of the male “Sexy face” in the comparison to the “unsexy face”:
- Browner skin
- Narrower facial shape
- Less fat
- Fuller and more symmetrical lips
- Darker eye brows
- More and darker lashes
- Upper half of the face broader in relation to the lower
- Higher cheek bones
- Prominent lower jaw
- More prominent chin
- No receding brows
- Thinner lids
- No wrinkles between nose and corner of the mouth
In our research project we examined the influence of facial symmetry on attractiveness. According to evolutionary theory faces are supposed to be more attractive the more symmetrical they are. In order to test this hypothesis we produced symmetrically optimised versions over a range of different faces (of low, average and high attractiveness). Each of these symmetrically optimised faces was presented to test subjects together with the corresponding original face. The task was to select the face that was perceived as being more attractive.
There are several ways of producing symmetrical faces: The most common method is creating so-called “chimeric faces”. Using image processing software, one half of the facial image is duplicated, mirrored along a vertical axis and finally added to the remaining half of the original face. The resulting, perfectly symmetrical face consists either of two left or two right halves of the original face. But there is one problem: Because faces are not perfectly symmetrical, it does make a difference whether you use the left or the right half as a starting point. Another problem is that, by using this method, birthmarks, pimples or irregular hair structures are doubled, too, so that the overall resulting symmetrical face looks quite odd.
Left: Symmetrical face using the left half of the original face. Middle: Original face. Right: Symmetrical face using the right half of the original face.
A clearly advanced and better way to produce individual symmetrical faces is the morphing technique. Morphed images are made from two images of different faces by averaging face shape and then blending red, green and blue intencity (RGB colour) across comparable pixels. All symmetrical faces we used in our research project were generated by blending together the original face with a duplicate that has been mirrored along a vertical axis.
In contrary to the Chimaerengesicht, the resulting face does not show a distinct dividing line along the centre of the face, and the question which half to take becomes obsolete.
Example: If a face has a broad left and a narrow right lower jaw, the mirror image method produces a face with either a broad or a narrow lower jaw on both sides. Avoiding these unpleasant effects, the morphing method automatically calculates the average breadth of the left and right half of the lower jaw.
In addition, asymmetries like a high-standing or slant eye are levelled out by this method. For our project we used a modified morphing procedure that symmetrized only the face proportions. Skin and hair remained unaltered, which made the resulting face look more natural and life-like.
Left: Symmetrically optimised face calculated by using the conventional morphing technique; bad skin is ameliorated, hair gets blurred, every single feature is reflected, facial hairs and pimples included.
The results from our experiment regarding ‘symmetry’ show that facial symmetry affects the perceived attractiveness. However, the effect is rather small and by far not as influential as it has been reported in the media. To sum up our findings: Very asymmetric faces are judged rather unattractive, but very unattractive faces are not necessarily asymmetric. And vice versa : very symmetrical faces need not necessarily be judged attractive and very attractive faces often show deviations from perfect symmetry (see report!). Based on our results, symmetry only seems to be a rather weak indicator for attractiveness. Often it is even difficult to distinguish between the original and the perfectly symmetrical version, because irregularities in shape are rather insignificant. Therefore, the strong influence of symmetry that has been reported in the scientific literature over and over again is questionable.
A remarkable result of our research project is that faces which have been rated as highly attractive do not exist in reality. This became particularly obvious when test subjects (independently of their sex!) favoured women with facial shapes of about 14 year old girls. There is no such woman existing in reality! They are artificial products – results of modern computer technology.
Same applies to the morphed average faces: Faces with such a smooth, pure skin, without any irregularities do not and cannot exist. But it is this kind of perfection that obviously overwhelmed our test subjects. Taking everything together it can be said that the most attractive face does not exist in reality – they are computed according to certain principles by machines.
Having these results in mind it is also not surprising that a model agency from Munich chose 88% artificial faces (14 out of 16 selected faces) for potentially being interesting as a model for the category “beauty”. Only two natural male faces could keep up with the computer generated ones, within the group of female faces no natural faces have been selected! We also asked test subjects to indicate the most attractive faces found the same pattern: 81% (13 out of 16) of the selected faces had been generated by the computer.
A selection of faces which have been judged by the model agency as being suitable as a model. All six faces were produced by the computer and don not exist in reality.
Natural faces cannot keep up with their artificial “competitors”. This becomes even clearer when you look at the labels of the rating scale that has been used for the evaluation: Just 3% (!) of the natural faces were rated as “rather attractive” – the judgements “quite attractive” and “very attractive” were never applied to natural faces. However, within the group of morphed faces 30% of all female and 23% of all male faces were perceived as being at least “rather attractive”.
This woman was found most attractive by our test subjects. Also staff of the model agency selected it as being suitable for a model career. But this person does not exist in reality – she was computed by blending together the eight most attractive original female faces. Their skin is absolutely perfect and actually looks rather artificial. But it is this kind of perfection that obviously attracts test subjects.
All this shows that we judge people and ourselves on a totally unrealistic basis. We compare ourselves with the most beautiful faces of the world which seem to be omnipresent in the media. They are integral parts in movies, in music video-clips, in commercials, they are on the title pages of magazines, on posters and so on. But the most absurd thing is that these “natural” faces the way they are depicted do not exist in reality either. Most of them are at least partially “artificial”. Their digital images are increasingly optimized by modern image processing software. By doing so faces are generated with attributes that are unreachable for even the most famous super models.
Even Claudia Schiffer does not look that perfect in reality. Her face has been electronically finished as different magazine editors admit: The skin gets its perfection by using various filters – thus pimples, large pores and small wrinkles disappear. Larger skin irregularities and wrinkles (e.g. under the eye) are retouched manually. All colors can be changed, for instance: the teeth get this bright white colour, the colour of the skin can be made browner and the blue of the eyes gets intensified. The eyes are worked on with sharp design filters – thus they appear more lively and more interesting. The manipulation of the white part of the eyes is particularly evident. It is strongly brightend (see her right eye which lies slightly in the shade and should normally look light-grey). Thus the eyes appear to be larger, younger and more beautiful. –> zoom in
Being surrounded by so much perfected beauty, it is not surprising that so many people are frustrated by their own appearance or that of their partners. This feeling of inadequacy is utilized by a whole branch of industry. More and more people seek to find help at beauty surgeons and many secretly hope to benefit from genetic engineering in order to provide at least their descendants with a perfectly styled body. So they keep on dreaming about an idealized form of beauty that has become an integral part of the media and which does not exist in reality.
Do attractive people have any advantages? Are they treated better than less attractive? Is it important to look good on an application photo? According to our investigations the answer to these questions is yes. We could show that people are perceived more positively the more attractive they are.
In order to examine this hypothesis we presented a range of different faces (unattractive, average attractive and attractive ones) to test subjects and asked to judge the person shown with respect to the following personality items: