Interpretation of appearance: the effect of facial features on first impressions and personality

Abstract

Appearance is known to influence social interactions, which in turn could potentially influence personality development. In this study we focus on discovering the relationship between self-reported personality traits, first impressions and facial characteristics. The results reveal that several personality traits can be read above chance from a face, and that facial features influence first impressions. Despite the former, our prediction model fails to reliably infer personality traits from either facial features or first impressions. First impressions, however, could be inferred more reliably from facial features. We have generated artificial, extreme faces visualising the characteristics having an effect on first impressions for several traits. Conclusively, we find a relationship between first impressions, some personality traits and facial features and consolidate that people on average assess a given face in a highly similar manner.

Figure 1. Example of two facial features, PC2 and PC13, and their interaction.

The faces visualise how two principal components, PC, extracted by an Appearance Model, interact with each other. The coordinate system shows the change in a face when a principal component is moved two standard deviations in either the positive or the negative direction. The face in the middle shows the mean for all factors. E.g. the face in the upper right shows PC2 and PC13 at +2 standard deviations. It is seen that PC13 explains the shape of the mouth and PC2 the face width.

Figure 2. Network graph of all significant correlations between Ratings.

The network depicts the relationship between the individual Ratings as the correlation coefficient, r, between scores. A dashed line depicts negative and a solid line positive correlations and the thickness of the line indicates the strength of the relationship with r as the edge label. Relationships significant for both genders are black, for men blue and for women magenta. Three clusters can be seen in the network with Trustworthy, Responsible, Friendly and Intelligent in the first, Extraverted, Adventurous, Emotionally Stable, Attractive and Physically Healthy in the second and Temperamental, Dominating and Masculine in the third. We named the clusters trustworthiness-friendliness, attractiveness-health-extraversion and dominance-masculinity.

Figure 3. Correlations between Ratings and self-reported personality traits visualised by heat maps.

Heat map A shows the correlations for women and heat map B the correlations for men. The personality traits are on the x-axis and the Ratings on the y-axis and a positive correlation is indicated with purple and a negative with green, where darker colours stand for bigger effect sizes. Only significant correlations with abs(r) ≥.20 and p<.01 are shown. Calculating the average of the correlations between personality traits and Ratings given by individual judges resulted in a drop in effect size; therefore the correlations in these heat maps should not be seen as significant on the individual level. Abbreviations for the Ratings are: Trustw.  =  Trustworthy, Adv.  =  Adventurous, Temp.  =  Temperamental, Healthy  =  Physically Healthy, Ext.  =  Extraverted, Dom.  =  Dominating, Att.  =  Attractive, Masc.  =  Masculine, Em. Stab.  =  Emotionally Stable, Resp.  =  Responsible and Int.  =  Intelligent.

Figure 4. Prediction of Ratings from facial features.

The plot shows the average correlation coefficient and standard deviation between observed and predicted scores for each Rating and each gender. A linear regression model was built in a 20-fold cross-validation with a varying number of the most correlated facial components as predictors, chosen based on the training set. Standard deviations are gathered by running the calculations thirty times with different folds for each run. The Ratings are in the plot ordered based on performance for the male faces. The size of the points indicates the Cronbach's α for that trait and it is seen that larger α-values correlate positively with prediction performance. Abbreviations for the Ratings are: Trustw.  =  Trustworthy, Adv.  =  Adventurous, Temp.  =  Temperamental, Healthy  =  Physically Healthy, Ext.  =  Extraverted, Dom.  =  Dominating, Att.  =  Attractive, Masc.  =  Masculine, Em. Stab.  =  Emotionally Stable, Resp.  =  Responsible and Int.  =  Intelligent.

Figure 5. Extreme faces for the Ratings.

For each face pair the left extreme face is predicted as being judged very low for a given trait and the right face as very high. Each face is based on the β-coefficients from the best linear regression model for that given Rating and gender. We generated the faces by multiplying each β-coefficient to either +4 standard deviations or -4 standard deviations of the matching facial component. A: Male extremes for Adventurous. B: Male extremes for Friendly. C: Male extremes for Dominating. D: Female extremes for Adventurous. E) Female extremes for Trustworthy. F: Female extremes for Dominating.

Figure 6. Validation of extreme faces.

In the validation we presented four faces to 116 persons and asked them to choose which one they found to represent a given trait the most. The left plot shows results for the male extremes and the right results for the female extremes. The length of each section in each bar indicates the percentage of times the given face was chosen. The dotted line indicates the percentage representing a random selection of the extreme face. In all cases except one the extreme face was chosen more often than random. For the male faces we found the extremes to be chosen significantly over random. For the female faces this was only found for the Friendly and Adventurous extremes. The colours are from .