Monday, January 13, 2020

Humans judge faces in incomplete photographs as physically more attractive

Humans judge faces in incomplete photographs as physically more attractive. Diana Orghian & César A. Hidalgo. Scientific Reports volume 10, Article number: 110 (Jan 2020).

Abstract: Attractive people are perceived to be healthier, wealthier, and more sociable. Yet, people often judge the attractiveness of others based on incomplete and inaccurate facial information. Here, we test the hypothesis that people fill in the missing information with positive inferences when judging others’ facial beauty. To test this hypothesis, we conducted seven experiments where participants judged the attractiveness of human faces in complete and incomplete photographs. Our data shows that—relative to complete photographs—participants judge faces in incomplete photographs as physically more attractive. This positivity bias is replicated for different types of incompleteness; is mostly specific to aesthetic judgments; is stronger for male participants; is specific to human faces when compared to pets, flowers, and landscapes; seems to involve a holistic processing; and is stronger for atypical faces. These findings contribute to our understanding of how people perceive and make inferences about others’ beauty.


We often judge others based on their physical appearance. Such judgments are driven by inferential mechanisms that help us fill in missing information. Here, we showed that (i) the inferential mechanism that we use to judge the physical appearance of human faces is positively biased, (ii) the bias is more pronounced in male participants, (iii) is specific to aesthetic judgments, but generalizes to other dimensions when the bias is strong enough, (iv) seems to be specific to human faces when compared to dog faces, landscapes, and flowers, and (v) is driven by the use of a holistic representation of what is a typical/average face. We also ruled out similarity to the self, positive expectations, and mood differences as explanatory mechanisms for the effect.
Presented with an incomplete human faces and instructed to judge their attractiveness, participants resort to what they know about faces (structure and features) and their representation of a prototypical face to generate new holistic representations. An inferential process that stems from matching the type of stimuli – i.e., human faces – with a prototype already existent in their memories. While incomplete human faces lead to an overall positive bias effect, stimuli such landscapes, pets, and flowers showed not positivity bias, which is likely due to the absence of a clear prototypical representations of these stimuli in people’s memories. Although our experiments suggest that typicality may have a role in the attractiveness positivity effect, further and more direct evidence is necessary to prove the robustness of this relationship. If typicality does play a relevant role, is also important to better understand how is this prototypical representation created and what are exactly the past experiences that shape it.
While the hypothesis that people fill in the missing pieces with positive inferences was never explicitly raised and tested, Saegusa and Watanabe stumbled on similar findings while investigating other phenomena. In their research on how information from individual facial parts contributes to the judgements of whole-face attractiveness over time, they found that attractiveness was higher for independent facial parts (e.g., eye, mouth) than for whole-faces36. Another study found that, on average, back-view photographs were rated as more attractive than front-view photographs37. The back-view condition can be seen as an extreme case of our incomplete treatments, in which the only information provided about the person is the shape of the head and the hair type, color, cut, and length. On a similar note, Miyazaki and Kawahara38 in an attempt to look into how the use of sanitary-masks by the Japanese women affects people’s perception of their beauty and health, found that certain types of occlusions also lead to higher perceived attractiveness, but only for originally unattractive faces judgements. Finally, Lu and collaborators39 manipulated the amount of information and attractiveness of cartoon characters (computer generated, gouache, and stick-figures), with the purpose of studying gender difference in attractiveness judgements. However, no significant differences were found between attractiveness judgements of the three types of cartoons. Overall, these findings support our hypothesis: when perceiving incomplete faces people fill in the missing information with positive details. Also, noteworthy, but in a domain different from that of facial perception, the work by Norton and colleagues8 showed that people perceive others’ personalities more favorably when they are provided with fewer personality traits as opposed to many.
Being positively biased about the attractiveness of strangers might have been a mechanism evolutionarily selected, as it might have facilitated social and reproductive events. However, the impact of this bias might only apply to impressions and interactions in first encounters. It is known that first impressions get diluted as we get to know and acquire more information about a person7. Thus, an interesting question for future research is the influence of the positivity bias on subsequent interactions with the target-person.
Whether the effect is unique to human faces also requires further research. More homogeneous categories than the ones we used need to be submitted to the same analysis to reach a more robust conclusion regarding the specificity of the positivity bias effect.
The contribution of face symmetry should also be studies in more detail. A meta-analysis performed by Rhodes in 200640 tells us that symmetric faces are perceived as more attractive when they result from blending the original and mirror-reversed images, but they are not when they are “chimeras” (pure mirror-reversed with no blending). Pure mirror-reversed photographs lead to less attractive exemplars due to enlargement or reduction of the mid-line features41. In our second experiment, we used chimeras because we wanted to understand if one half of the face is used to infer the missing half, but it would be interesting to test whether using a blended symmetric face (and thus more naturally looking) would lead to a similar conclusion.
One limitation of our work is that all experiments were performed online with Mechanical Turk participants. While there is research showing that data from online experiments is comparable to data from lab-based experiments42,43, these conclusions need to be replicated in the laboratory and in contexts where the implications of the research might be directly relevant (e.g., social media, recruitment, fashion industry, entertainment, advertisement, and marketing).

No comments:

Post a Comment