Face Perception

A well-told tale teaches us that we only get one chance to make a first impression. From personal relationships to casual encounters, we base our judgements on information gathered spontaneously from the appearance of those who stand in front of us. The focal point of this automatic process seems to be the face, deemed “the window of the soul” (Zebrowitz, 1997). Following a Gibsonian approach to object perception (Gibson, 1979), Zebrowitz & Montepare have stated in 2008 that the qualities conveyed in facial structure can accurately reveal social interaction opportunities, and guide our responses even when they should be driven by more unbiased data. For example, asymmetrical faces are perceived as less intelligent than more symmetrical ones (Rhodes, Zebrowitz et al., 2001), and men identify women with masculine-looking features as less bright, healthy and sociable (Cunningham, 1986).

Furthermore, Secord et al. (1954) investigated whether perceived physiognomic and perceived personality characteristics are related, laying the groundwork for the modern approach to the subject. In their study, 146 college students rated 23 perceived physiognomic and 35 perceived personality traits of 24 individuals' photographs. The results showed that 93.9% of the predicted correlations between physiognomic and personality traits were in the expected direction, in addition to a wide correspondence between the similarity clusters of both judgements; for example, faces perceived as distinguished, intelligent and determined had thin lips and wrinkles around their eyes. This findings were coherent with physiognomic stereotypes, with a plausible psychological origin.

Facial appearance can therefore be held accountable for a number of behavioral responses. The next question concerns whether the forming of an impression has to be a long and thoughtful process or if it can happen in a time frame so short that it eludes the conscious mind. In 2006, Willis and Todorov created a series of five experiments to explore the minimal conditions necessary for the formation of an impression based on facial appearance, each focused on a different trait (attractiveness, likeability, competence, trustworthiness and aggressiveness), concluding that a 100 ms exposition to the stimulus is sufficient to gather an extensive amount of information in order to create the judgement. A supplementary exposition doesn’t affect the initial judgment, and it can give more confidence in the decision. However, further exposition can make the impression more negative, driven by a pronounced effect of the positivity bias in a minimal information condition (Sears, 1983).

The short time needed and the accuracy of the impressions suggests that information gathering from the face is a System 1 process: fast, intuitive, and unreflective (Todorov et al., 2005). Further research has shown that a minimal exposure of 39 ms is enough to create trait inference from a threatening face, with a significantly above chance performance, whether a 26 ms exposition is not enough to gather a useful amount of information from the stimulus.

Influence of impressions

Our impressions of others influence our behavioral response in a number of settings, even if it is based on a very limited batch of information and is created in a very short time frame; our evaluations can therefore predict important social outcomes such as profit, rank and results of political elections. Three examples will be discussed in the next paragraphs regarding the importance of impressions in three different scenarios.

Research has shown that the inferences from female Chief Executive Officers’s faces can accurately predict their success (Rule & Ambady, 2009). In this experiment, the faces of the top 1-25 and the bottom 1-25 female CEOs from the U.S. Fortune 500 list were rated by 90 participants on the basis of competence, dominance, likeabilty, facial maturity and trustworthiness on a 7-point Likert-type scale. In the second part of the study, the financial performance of the companies guided by the CEOs were obtained and compared with the trait evaluations. The results indicated that scores of competence and leadership highly correlated with company profits [r(14)=.52, p=.04; r(14)=.60, p=.01], thus confirming that leader who were perceived as more successful actually led successful companies.

Mazur et al. have inquired in their 1984 study whether the physical appearance of male cadets could affect social mobility in the military ranks, based on prior research that showed that males are perceived as either dominant or submissive individuals, focusing on particular characteristics such as facial features, tallness, and an athletic physique (usually associated with a dominant personality). Using the yearbook of the West Point Class of 1950 as a reference, they were able to obtain the facial portraits of the graduating cadets, close approximations of their height and athletic prowess, and the military ranks while at the academy. The faces were shown to 20-40 judges that rated them on a scale from 1 (very submissive) to 7 (very dominant), and a substantial correlation between the facial appearance of the cadets and their final military rank while at West Point was found, proving that dominant-looking men advance to higher ranks in the military hierarchy than submissive-looking ones.

The study was then replicated in 1996 by Mueller and Mazur, because the facial dominance of the cadets was found to be unrelated to their ranks in later career in the 1984 study. The missing information regarding the cadets' final ranks from the 1980 register were integrated with a questionnaire sent to the men in question, investigating new variables such as graduation from a war college. Final results showed that the cadets' facial dominance was again the most important variable and, while still unrelated to their ranks at the middle of their career, was related to promotions in late career, 20 or more years after the portraits were taken.

In addition to the previous examples, inferences of competence have been shown to predict the results of political elections. Eventualmente esperimento 2005 Todorov Rapid judgments of competence based solely on the facial appearance of candidates can predict the outcomes of gubernatorial elections (Ballew & Todorov, 2007). To test this hypothesis, in the first experiment the faces of the winner and the runner-up of 89 election campaign were presented to the participants, who had to decide who was more competent in three different conditions: a 100 ms exposure, 250 ms and unlimited time exposure. In addition, participants were asked to express a binary choice (choosing only which candidate was more competent), a judgement of competence on a 9-point scale, and whether or not they recognized the canditate. If the final answer was affirmative, the trial was excluded to ensure impartiality. Results showed that fast, simple and binary judgements of competence in a time frame as fast as 100 ms could accurately predict the outcome of the elections, and that additional exposure did not improve the predictions.

In the second experiment, participants were explicitly asked to deliberate and make a good judgment; this addition dramatically increased their response times and reduced the predictive accuracy of judgments, adding noise to automatic trait judgments and thus reducing the accuracy of prediction.

In a final experiment, data collected before the elections in 2006 regarding competence judgments effectively predicted 68.6% of the gubernatorial races and 72.4% of the Senate races when a Republican and a Democratic candidate were compared, further suggesting that a rapid, unreflective judgment of competence inferred from the face of a candidate can affect voting decisions and predict the election outcomes.

The concept

“We look at a person and immediately a certain impression of his character forms itself in us. A glance, a few spoken words are sufficient to tell us a story about a highly complex matter. We know that such impressions form with remarkable rapidity and with great ease. Subsequent observation may enrich or upset our first view, but we can no more prevent its rapid growth than we can avoid perceiving a given visual object or hearing a melody. We also know that this process, though often imperfect, is also at times extraordinarily.” (S.E. Asch, 1946). Solomon Asch, one of social psychology’s founding fathers, gives us an early but accurate definition of the process that creates an impression of an individual, a rapid but complex inference that guides us towards a future behavioral response.

The face is therefore our first medium to obtain information about a person’s emotional and mental state (Todorov & Oosterhof, 2011), with the consequent inclination of “reading too much into the human face”. This suggests that the social perception of faces can be modeled to further comprehend which differences in the structure of the face lead to inferences based on physical appearance. For this matter, a data-based approach was chosen to create a statistical model of face representation, utilized to extract the subtle alterations of facial characteristics that cause the changes in social perceptions. The authors used the “space face” model (Blanz, & Vetter, 1999) implemented in Facegen (www.facegen.com), highlighting 50 dimension for the shape of the face and 50 dimension for the face reflectance (brightness, color, and texture variations on the surface map of the face), represented by the correspondent vector. Five versions of every face were used for every dimension, modeled to increase or decrease its perceived value. These social dimensions can reveal the facial cues that lead to specific social judgments. In the case of trustworthiness, when its vectorial value is increased the face seems to express more positive emotions.

Further relevant information regarding the importance of trustworthiness judgements comes from the 2013 study by Todorov et al., broadening the prior research by creating 7 databases each containing 25 identities related to a single social judgement: attractiveness, competence, dominance, extroversion, likability, threat, and trustworthiness, selected because spontaneously used by people to describe unfamiliar faces. Each identity was manipulated to take on 7 different dimension values, in a range from +3 to -3 with 1 SD varying between every interval for a total of 175 stimuli in each database. The models account for at least 75% of the variance of all judgments, fitting remarkably well for each social judgement. In the case of trustworthiness, increasing the value of the SD interval results in a higher perception of its presence, so that the faces with a value of +3 SD are perceived as more reliable than the others.

Thanks to the use of validated faces to measure the variable of trustworthiness, it has been possible to run a number of studies to investigate how its value influences our life. As it will be demonstrated in later paragraphs, perceived trustworthiness from other individuals’ faces has numerous consequences on our social behavior, in fictional sceneries and in reality.

The implications of the judgment of trustworthiness

The ability to cooperate with other individuals has served as a strong foundation for our social behavior, and it’s a necessary value to sustain the growth of our society. In the stipulation of a contract, whether social or formal, the variable that can effectively predict and guide the drafting is the perception of trust. In the business world, corporate men prefer to “look a man in the eye” before the final decision (further stressing the importance of the face in social judgements as previously noted). When this is not possible, our perception of trust must come from precedents and, most importantly, reputation, as the behavior in precedent interactions is widely considered to be the best predictor of the level of reliability (King-Casas et al., 2005). To investigate the concept of trust, the appropriate approach should start with an initial evaluation based on the available information, gradually updating it with the additional details provided by subsequent interactions.