The length of the gaze affects human preferences, new study shows

PASADENA, Calif.—Beauty may be in the eye of the beholder, but a new psychophysical study from the California Institute of Technology suggests that the length of the beholding is important, too.

In an article appearing in the December 2003 issue of the journal Nature Neuroscience, Caltech biology professor Shinsuke Shimojo and his colleagues report that human test subjects asked to choose between two faces will spend increasingly more time gazing at the face that they will eventually choose as the one more attractive. Also, test subjects will typically choose the face that has been preferentially shown for a longer time by the experimenter. In addition, the results show that the effect of gaze duration on preference also holds true for choices between abstract geometric figures.

The findings show that human preferences may be more fundamentally tied to "feedback" between the very act of gazing and the internal, cognitive prototype of attractiveness than was formerly assumed. Earlier work by other researchers has relied on the "attractiveness template," which assumes that an individual's ideal conception of beauty has somehow been imprinted on his or her brain due to early exposures to other people's faces, such as the mother.

In fact, Shimojo says, the new results come from experiments especially designed to minimize the influence of earlier biases and existing preferences. Even when images of faces have been computer-processed to eliminate possible biases due to ethnic origins and even such trivial factors as hairstyles, the results still show strongly that the gaze is subconsciously oriented toward the eventual choice. This holds true even more strongly when a test subject is asked to choose between two abstract geometric figures, suggesting that the slightly lower tendency to fix the gaze on the eventual choice of two faces is influenced by existing selection biases that cannot be totally controlled.

The findings in Nature Neuroscience comprise two experiments. The first was the choice of the more attractive face, in which all the test subjects were asked to rate the faces from 1 (very unattractive) to 7 (very attractive). The average rating for each face was then calculated so that faces in pairs could be matched in different ways.

In the "face-attractiveness-easy task" the faces were paired according to gender, race, and neutrality of facial expressions, but comprised a choice of a "very unattractive" face with a "very attractive" face. Five test subjects were then shown 19 face pairs and were asked to choose the face they preferred. A video camera recorded the movements of their eyes as they directed their attention from one face on the screen to the other.

The results showed that the likelihood of gaze of the test subjects started from chance (50 percent) but rose above 70 percent of their time gazing at the face till they chose that face.

Even more striking was the difference in gaze devoted to the "face-attractiveness-difficult task," in which 30 pairs of faces were matched according to the closeness in which they had been ranked for attractiveness. In this experiment, the test subjects spent up to 83 percent of their time gazing at the face they would choose immediately before their decision response, suggesting that the gaze is even more important when there is little difference in the features of stimuli themselves.

The test subjects were also asked to choose the least attractive face, as well as the rounder face, and the results also showed that the length of the gaze was an important indicator of the eventual choice. In addition, the subjects were asked to choose between abstract geometric shapes, and the length of gaze also correlated highly with the eventual choice.

The second experiment is "gaze manipulation," in which the faces are not shown simultaneously, but in sequences of varying duration on the two sides of the computer screen. In other words, one face was shown for a longer time (900 milliseconds) than the other face (300 milliseconds), and as a control, the faces were also shown to other subjects in the center of the screen in an alternating sequence.

The results show that the face shown for a longer time tends to be chosen at chance level (50 percent) with only two repetitions of the sequence, but about 59 percent of the time with 12 repetitions. This suggests that the duration of the gaze can influence the choice. However, this manipulation did not work in the control experiment without gaze shift, as mentioned above, indicating that it is not mere exposure time, but rather active gaze shift, that made the differences.

In sum, the results indicate that active orienting by gaze shift is wired into the brain and that humans use it all the time, albeit subconsciously, Shimojo says. One example is our preference for good eye contact with people whom we are engaging in conversation.

"If I look directly into your eyes, then glance at your ears, you can immediately tell that I've broken eye contact, even if we're some distance apart," Shimojo explains. "This shows that there are subtle clues to what's in the mind."

In addition to Shimojo, the other authors are Claudiu Simion, a graduate student in biology at Caltech; Christian Scheier, a former postdoctoral researcher in Shimojo's lab; and Eiko Shimojo, of the School of Human Studies/Psychology at Bunkyo Gakuin University in Japan. Shinsuke Shimojo and Claudiu Simion contributed equally to the work.