From The Oxford English Dictionary (2014): physiognomy is (1a) “a person’s facial features or expression (originally frequently considered as indicative of the mind and character),” and (2) “the supposed art of predicting the future from the features of the face.”
Chinese beliefs in physiognomy have been dated back to the Zhou Dynasty (1046-256 BCE) (representative writings: Classic of Rites (Fang, 2008)), well- developed when in the period of two Han dynasties (206 BC-9 AD) (representative writings: Miscellaneous on Face Features (Xie & Dong, 2008), Discourse Balance (Wang, Han Dynasty)), and had a prosperous period in Song Dynasty (960-1279) (representative writings: Shen Xiang Quan Bian (Chen, Song Dynasty) and Ma Yi Shen Xiang (Mayi, 1997). The ensuing Tai Qing Shen Jian (Wang, 2009) and Liu Zhuang Shen Xiang (Yuan, Ming Dynasty)) in Ming Dynasty (1368-1644), Bing Jian (Zeng, Qing Dynasty) in Qing Dynasty (1636-1912) are all widespread in China nowadays. However, over 3000 years of development, few scholars have explored the relationship between facial features and personality traits scientifically (Xu, 2008).
The history in western countries of belief and practice of physiognomy can be traced back to ancient Greece, when Aristotle published Physiognomica (see, e.g., Brandt, 1980; Zebrowitz, 1997). In the late eighteenth century and the early nineteenth century, Johann Caspar Lavater’s (1797) resoundingly successful book was the historic apex of physiognomy (see, e.g., Hegel, 1931/1967, pp. 337-372) and was welcomed by the general public and scholars of this neglected area (see, Shookman, 1993). Emerging from phrenology, physiognomy originally studied protuberances on the skull in the late nineteenth century (Ran & Yaacov, 2000). By the end of the 19th century, phrenology was in decline and had been eliminated by science. Early research (see, e.g., Anderson, 1921; Cook, 1939; Thornton, 1943; Secord, 1965; Secord et al., 1953; Secord, Dukes, & Bevan, 1954; Strich & Secord, 1956; Albright et al., 1997; Berry, 1990, 1991; Zebrowitz, 1997; Berry & McArthur, 1985; McArthur & Apatow, 1984) seemed to suggest an acceptable reliability of physiognomic inferences by experiments, which means independent judges reading similar personality from a face. Conversely, early research also showed poor validity and repeatability of physiognomy inferences (see, e.g., Cohen, 1973; Alley, 1988).
Early scientific studies on physiognomy found very little (if any) evidence for a relationship between facial features and personality traits (Cleeton & Knight, 1924). Whereas early studies examined relationships between isolated facial features and personality (e.g., eye size and impulsiveness, Cleeton & Knight, 1924), similar results were found throughout most of the 1900s (Alley, 1988). Many studies have now shown that perceptions of personality drawn from single face images, short videotapes, or very brief personal encounters align with the target’s self-reported personality (Ambady, Hallahan, & Rosenthal, 1995; Berry, 1990; Bond Jr., Berry, & Omar, 1994; Borkenau & Liebler, 1992a, 1992b, 1993a, 1993b). Roy (2003) even considered physiognomy as an art and pseudoscience. As technology develops, accurate personality judgments can even be made from computer-averaged faces. Penton-Voak, Pound, Little, and Perrett (2006). Further studies have found that facial dimensions are associated with dominance and aggression (facial width-to-height ratio; Carré et al., 2009).
In this paper, physiognomy only refers to identifying personality traits, not includes the part of telling future. Facial features in physiognomy have different categories, such as five main face parts (eyes, eyebrow, ear, mouth, nose), up, middle, lower part of your face, and 12 functional parts of your face (Xu, 2011). Different facial features (e.g. face shape) can reveal certain meanings in Physiognomy. This paper only uses face shape, one of the facial features, as the object to study whether people with different face shapes will have different personality traits. Meanwhile, this study also investigated how face shape influences people’s judgement of their corresponding personality traits.
2. Research Method
The whole study consisted of two stages. The first stage was to collect experimental materials, and the second stage was to carry out experiments.
2.1. Category of Face Shapes
This study synthesized many books about physiognomy (e.g. Ma Yi Shen Xiang) and finally employed 8 face shapes to categorize, which are oval, round, rectangular, square, triangular, diamond, inverted triangle and heart face shape. This face shape category is extensively applied in practice and seen in many books, and also has clear distinction among each shape (Figure 1).
2.2. Typical Faces and Tested Personality Traits
After receiving informed consent, over 4000 new university students from a same university participated in the first stage of the study, which was to complete sixteen personality factor questionnaires (16PF in short) and submit their color electronic ID photo. The measurement of 16PF was conducted on a class
Figure 1. Face shapes categories.
basis and on computer-based software in the computer room. Electronic photos employed the standard identification photo and have been taken in a photo studio. Finally, 3816 photos corresponding to their 16PF scores have been selected based on the criterion of not wearing glasses, no hair covering the face, no makeup, no obvious scar, no obvious mole or birthmark, showing the whole face. Ten assistants attended the training to learn the method of identifying eight different typical faces from all qualified photos, and picked 30 face photos for each type of face shape. Comparing all results of assistants’ selections, one photo was selected as typical face if it has been selected by at least seven assistants and above. The above process was repeated until eight photos have been chosen for each face shape for this experiment. Each photo of the typical face has a code (see Figure 2), such as R1 (round face No.1), S5 (square face No.5), and I7 (invert-triangular No.7).
By searching typical faces’ student’s ID number, the corresponding score of 16PF has all been found and matched. The reason to select university students’ faces as typical faces is because their faces features are obviously well-developed and the skeletal structure has been basically finalized, which can exclude the problem of wrinkles among elder people and baby faces among children or teenagers. Besides this consideration in choosing typical faces, the reason for using 16PF scores from university students is also because the personalities of university students are basically mature and stable.
2.3. Artificial Faces
Although the selected typical faces have the most typical features respectively for eight types of face shape, other facial features still remained on typical faces, such as eye size and nose height. To eliminate the influence of all other facial features, the researcher created artificial faces for the experiment to control the experimental variables.
Figure 2. Typical face examples.
The process of creating artificial faces started from original digital human face photos, which were found from searching “ID photo” as keywords in a search engine, such as Google. The criteria for selecting these ID photos are: non-famous people, non-high click photos, average face, without glasses, showing full face, and no copyright. Eight photos (male and female equally) have been randomly chosen from up-to-standard original digital ID photos, and downloaded in a high- quality JPG file format. Then in Photoshop software, each of these eight ID photos was reshaped into eight different face shapes (8 ID photos × 8 face shapes = 64 pieces in total) by deleting the hair and ears, leaving remaining the facial features unchanged. As is seen from the samples in Figure 3, the only variable is face shape, while other facial features were made uniform.
2.4. Participants in the Second Stage of the Study
In order to get a big sample, the experiment in the second stage applied class- based group testing. To prevent participants who would potentially know the person in the typical face, all participants were recruited from another university to assess the personality traits of the typical faces and artificial faces. Based on the principle of convenience sampling and their curriculum schedule, 949 university students (18 to 25 years old, M = 21.63, SD = 1.821) participated in the study (male = 405, 42.7%; female = 544, 57.3%).
2.5. Experiment Process
The experimental process was conducted in a computer laboratory with E-prime program in every computer. In each experimental trial, the participants from the same class entered the laboratory with an appointment and every participant received an evaluation manual. The experimenter started to read out the experimental instruction: “This is a psychological experiment about personality traits judgment. Please open the evaluation manual in your hand and score on the correspondence page when you evaluate each face later. There is no right or wrong about personalities here. You will score in sixteen dimensions for each face. Please read carefully the instruction on the first page in the manual and make sure you understand the description of every personality dimension. You are welcome to ask questions if you don’t understand.” If participants clearly
Figure 3. Sample images of artificial faces.
understood, the experimenter would say “please judge their personalities according to the face being shown. Each face has sixteen dimensions for you to judge and please circle the corresponding number in the middle column of the form. You only have two minutes to rate each face, so use your intuition.” After the experiment finished, the participant is welcome to leave after handing in the manual (Figure 4).
Based on the experimental design, there would be 128 faces (8 face shapes × 8 typical faces + 8 artificial faces × 8 face shapes) that need to be judged, which will take 256 minutes to complete. To shorten the experiment time, every trial and every participant was regularly arranged to judge 16 faces (8 artificial faces and 8 typical faces between guidance words and ending black screen) by using the method of permutation and combination, which will take 32 minutes in total.
2.6. Evaluation Manual
The researcher created an evaluation manual for personality traits judgment. Every face with a specific code, which had a same code on the manual, was displayed on the computer screen. In addition, every page was composed of a personality profile table (including sixteen personalities’ traits descriptions) adopting from 16 Personality Factors Questionnaire (Russell & Karol, 2002). The first page required demographic information from participants, such as gender and age. The cover page showed the instructions for the experiment.
3.1. The Personality Profile of Eight Face Shapes from Typical Faces
From the above figure, it is apparent that the personality profiles of the different face shapes were generally similar and even overlapped on some factors. Most of the face shapes had one or more personality traits that significantly differed from other face shapes. Based on the results from post-hoc test of variance analysis, these significant differences have been circled in Figure 5. The inverted triangular face and rectangular faces were not significantly different from any other face shape in any of the 16 personality traits. However, on the whole, no one face shape was significantly different from all others in any one personality trait.
Figure 4. Experimental process.
Figure 5. The personality profile of eight face shapes from typical faces.
3.2. The Assessed Personality Profile of Eight Face Shapes from Artificial Faces
Compared to the result 3.1, the personality traits of eight face shapes are significantly different from each other in the participants’ eyes. Each face shape was assessed to have certain personality traits that differed from all other face shapes, which have been circled in Figure 6.
3.3. The Difference between the Tested and Assessed Personality of Each Face Shape from Typical Faces
For each face shape from typical faces, the researcher compared 16 tested personality traits with assessed personality traits, and listed all the significantly different traits for each face shape in the table above. Some face shapes showed significant differences between perceived personality (assessed personality) and real personality (tested personality), such as the inverted triangular face and the square face. For other face shapes, people’s judgments of personality traits closely matched the tested results, such as with the triangle face (Table 1).
Figure 6. The personality profile of eight face shapes from artificial faces.
Table 1. The difference between the tested and assessed personality of each face shape.
4.1. Whether Traditional Physiognomy Is Reliable
Result 3.1 showed that different face shapes were not associated with any particular personality traits. The Spearman rank correlation also showed that there is only a slight correlation between face shape and factor B (intelligence) (r = −0.247, P = 0.049). In other words, the traditional physiognomy theory does not seem to hold water. Traditional physiognomy often says: people with round faces are gentle and smooth, easy to get along with, that those with oval faces have good compliance, while those with square faces have strong willpower and an extroverted personality, and those with inverted triangular faces lack mobility and so on (Livia, 1986), which cannot be seen in the results based on experimental data in this study.
In Figure 5, the personality profiles of various face shapes are similar. This phenomenon may be because the participants got the 16PF questionnaire score by completing questions about how to evaluate their own behavior and attitude. Most people tend to bring out the better side of themselves to fit society’s expectations. Therefore, many participants got a similar score of the 16PF questionnaire. In a later experiment, the participants were asked to rate personality traits by looking at other people’s faces. This setup allowed participants in later experiments to freely evaluate others’ personalities without considering social expectations. This is the limitation of this experiment. Also, it’s inevitable for this error if the experiment employing a traditional psychological measurement.
4.2. Judging Personality from Faces in Modern Days
As result 3.2 reveals to us, people already stereotype each face shape. The reason is not only coming from the effect of traditional physiognomy theory, but also comes from everyone’s daily life experience, as well as the associations of each face shape’s appearance. For instance, people normally associate square faces, also called “Guo Zi Lian” in China, with responsible, strong and upright personality traits. The study also found that the high score traits and low score traits combination for each face shape, which is meanwhile stereotype the face shape coming from people (Table 2).
Table 2. High score traits and low score traits combination for each face shape.
5. Implication and Limitations
This is an era of looking at faces. Even if we can’t find the evidence that directly relates to face shape and personality, however, we found that people still make judgments on others’ personality according to their face shapes, which is consistent with those of previous study (Todorov, Said, & Verosky, 2011). As the Chinese saying goes, we “judge people with their appearance”. Whether or not we are correct to do so (as showed in the result that at least 21.9% of the time it is not correct), this prediction already makes the first impression for us when we see others at the first sight. This finding can be well applied in daily life, for instance, using makeup techniques such as highlighter and shadow powder to change your face shape according to different occasions and other people’s preference, will enable us to easily get trust and build a relationship. Especially in the scenarios of blind dates or interviews, we will have more advantages by doing in this way.
This study has some limitations. First, the sample size isn’t big enough to represent the population of a national norm. Second, it is not enough to exclude the influence of other facial features on the experimental results through the subjective evaluation of 10 laboratory assistants. Third, the typical faces and artificial faces in the study are all Asian faces, which cannot well cover the features of face shape from other ethnicities. Therefore, follow-up research needs to improve in these areas.
 Albright, L., Malloy, T. E., Dong, Q., Kenny, D., Fang, X., Winquist, L., & Yu, D. (1997). Cross-Cultural Consensus in Personality Judgments. Journal of Personality and Social Psychology, 72, 558-569. https://doi.org/10.1037/0022-35220.127.116.118
 Ambady, N., Hallahan, M., & Rosenthal, R. (1995). On Judging and Being Judged Accurately in Zeroacquaintance Situations. Journal of Personality and Social Psychology, 69, 518-529. https://doi.org/10.1037/0022-3518.104.22.1688
 Berry, D. S. (1991). Accuracy in Social Perception: Contributions of Facial and Vocal Information. Journal of Personality and Social Psychology, 6, 298-307. https://doi.org/10.1037/0022-3522.214.171.1248
 Borkenau, P., & Liebler, A. (1992a). The Cross-Modal Consistency of Personality: Inferring Strangers Traits from Visual or Acoustic Information. Journal of Research in Personality, 26, 183-204. https://doi.org/10.1016/0092-6566(92)90053-7
 Borkenau, P., & Liebler, A. (1992b). Trait Inferences: Sources of Validity at Zero Acquaintance. Journal of Personality and Social Psychology, 62, 645-657. https://doi.org/10.1037/0022-35126.96.36.1995
 Borkenau, P., & Liebler, A. (1993a). Consensus and Self-Other Agreement for Trait Inferences from Minimal Information. Journal of Personality, 61, 477-496. https://doi.org/10.1111/j.1467-6494.1993.tb00779.x
 Borkenau, P., & Liebler, A. (1993b). Convergence of Stranger Ratings of Personality and Intelligence with Self-Ratings, Partner Ratings, and Measured Intelligence. Journal of Personality and Social Psychology, 65, 546-553. https://doi.org/10.1037/0022-35188.8.131.526
 Carré, J. M., McCormick, C. M., & Mondloch, C. J. (2009). Facial Structure Is a Reliable Cue of Aggressive Behavior. Psychological Science, 20, 1194-1198. https://doi.org/10.1111/j.1467-9280.2009.02423.x
 Penton-Voak, I. S., Pound, N., Little, A. C., & Perrett, D. I. (2006). Personality Judgments from Natural and Composite Facial Images: More Evidence for a “Kernel of Truth” in Social Perception. Social Cognition, 24, 607-640. https://doi.org/10.1521/soco.2006.24.5.607
 Roy, P. (2003). “Marginalized Practices”, The Cambridge History of Science: Eighteenth- Century Science. The Cambridge History of Science, (4 Illustrated ed.). Cambridge University Press, 495-497.
 Secord, P. (1965). Facial Features and Inference Processes in Interpersonal Perception. In R. Taguiri & L. Petrullo (Eds.), Person Perception and Interpersonal Behavior (pp. 300-315). Stanford, California: Stanford University Press.
 Secord, P. F., & Bevan, W. (1956). Personalities in Faces: III. A Cross Cultural Comparison of Impressions of Physiognomy and Personality in Faces. Journal of Social Psychology, 43, 283-288. https://doi.org/10.1080/00224545.1956.9919224
 Thornton, G. R. (1943). The Effect upon Judgments of Personality Traits of Varying a Single Factor in a Photograph. Journal of Social Psychology, 18, 127-148. https://doi.org/10.1080/00224545.1943.9921704
 Todorov, A., Said, C. P., & Verosky, S. C. (2011). Personality Impressions from Facial Appearance. In A. Calder, J. V. Haxby, M. Johnson, & G. Rhodes (Eds.), Handbook of Face Perception (pp. 631-652). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780199559053.013.0032