Facial attractiveness has been shown to be closely related to positive personality traits, creating a halo effect. The facial appearance heuristic (e.g., Vrij, 2004 ) leads humans to judge others whose faces are attractive as honest. Consequently, perceptions of deception in daily life may be easily biased by appearance. Past research has indeed suggested that facial appearance, including attractiveness, affects the detection of deception (e.g., Aune, Levine, Ching, & Yoshimoto, 1993 ; Bond, Berry, & Omar, 1994 ; Bull, 2004 ; Bull & Rumsey, 1988 ; Masip, Garrido, & Herrero, 2004 ; Porter, Campbell, Stapleton, & Burt, 2002 ; Rowatt, Cunningham, & Druen, 1999 ; Zebrowitz, Voinescu, & Collins, 1996 ). For example, Porter et al. (2002) and Bull and Vine (2003) examined the effect of attractiveness of male and female models on judgments of lying/truth telling. Attractive models who were telling the truth were judged as more truthful. However, the attractive and unattractive models presented in the stimulus videos were different persons; thus deceptiveness judgments may have been affected by differences in the models. By contrast, Aune et al. (1993) created attractive and unattractive model stimuli by changing the appearance of a single female by altering her clothing, makeup, and hairstyle. The unattractive model was perceived as more deceptive than the unattractive model, but only by male participants. However, the model’s speech was prepared by the researchers and practiced by her for four hours before recording; thus the model did not speak spontaneously, and because her performance was manipulated to conform to stereotypical liar behavior, she may have behaved unnaturally.
The purpose of the present study was to address the limitations of previous research in examining whether or not more attractive people are perceived as less deceptive. We compared perceived deceptiveness between attractive and unattractive models by using the same three female models in both high and low attractiveness conditions. Each model’s facial attractiveness was manipulated using makeup. To make the models’ speech and behavior appear more natural, we instructed each model to choose her own words and manner of speaking in the stimulus videos. Each model prepared her own responses to a situation similar to the “video-dating service” scenario of Aune et al. (1993) , in which a single female introduced herself to potential boyfriends. The models repeated the same responses in Japanese for both high attractiveness and low attractiveness conditions. We predicted that the more attractive females would be judged as less deceptive. In the design of this study, we controlled for sex to reduce the variance when conducting ANOVAs and used personality traits to get insights into future research.
2.1. Stimulus Construction
Four females (age range 19 - 24), all native speakers of Japanese, volunteered to serve as models. Two were undergraduate students, one was a graduate student, and one was a teacher who had graduated from university a year earlier. Each model was asked to arrange her own makeup and hairstyle in order to present herself as attractively and unattractively as possible. Each model was presented with 15 questions in Japanese (translated into English in Table 2) and was asked to answer each question either truthfully or untruthfully, as she wished, and to respond as if she were introducing herself to potential dates or boyfriends. The models were recorded responding to the questions while wearing first the unattractive makeup and then the attractive makeup. The videos of one of the four models were excluded because she had a general difficulty in performing responses. Thus videos from three models, Models A, B, and C, were included in the experimental stimuli. Figure 1 shows the three models in attractive and unattractive makeup (models are not identified by letter in the figure for reasons of privacy).
Two series of video stimuli were created, one that included only video clips of the models in the attractive makeup (high attractiveness condition), and the other that included models in the unattractive makeup (low attractiveness condition). A total of 45 stimuli (15 stimuli for each model) were arranged in random order and presented in each condition.
In order to estimate appropriate sample size, we conducted a power analysis. Our pilot research indicated that manipulation of attractiveness may sometimes fail because young female models without makeup or with “unattractive” makeup may sometimes be regarded as attractive solely because of their youth. Thus we predicted only a moderate effect size for attractiveness (d = 0.5, Cohen’s index; Cohen, 1992 ). Assuming alpha = 0.05 and (1 − beta) = 0.80, we determined that 64 participants were required for each condition. A total of 133 Japanese university students volunteered to participate (54 males and 79 females); each was randomly assigned to one of the two conditions in a between-participants design (high attractiveness: 28 males and 44 females; low attractiveness: 26 males and 35 females).
Figure 1. Three models in two attractiveness conditions: high attractiveness (left column) and low attractiveness (right column).
The experiment was administered to groups of participants in a classroom setting. Participants responded to a series of items printed on a questionnaire after watching each video clip. After all stimuli were presented, participants were asked to provide demographic information (sex, age, university faculty, department, and year) and to respond to several personality scales. The experiment lasted approximately 30 minutes.
The Perceived Deceptiveness Scale consisted of three items: “Deceptive”, “Straightforward”, and “Trying to show oneself in the best light”. Participants indicated their impression of the model in the video clip by responding to each item using a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree).
The Perceived Attractiveness Scale comprised five bipolar adjective sets: “Like-Dislike”, “Bad impression-Good impression”, “Adorable-Not adorable”, “Unattractive face-Attractive face”, and “Beautiful-Not beautiful”. Participants indicated their impression of the model in the video clip by responding to each adjective set using a five-point scale.
Participants also responded to a Yes or No question asking whether they were acquainted with the model in the video clip. The data of participants who either answered Yes or failed to respond were excluded from analyses of that model; 2 participants were excluded for Model A, 3 for Model B, and 2 for Model C.
After responding to the 45 video stimuli, participants were administered the Japanese version of the Ten Item Personality Inventory for Big Five Personality (TIPI-J; Oshio et al., 2012 ), the General Trust Scale ( Yamagishi & Yamagishi, 1994 ), the Tendency to fall in love with looks and Tendency to fall in love with personality scales ( Ochi, 2015 ), and the Visual Analogue Scale (VAS). The horizontal VAS was used, which was anchored on one end of the scale with “Put a high value on looks” and on the other end with “Put a high value on personality”. Except for the VAS, participants indicated their agreement with each item on a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree).
2.5. Human Research Ethics
Ethical approval by the research ethics board of Bunkyo Gakuin University was obtained prior to the experiment.
3. Results and Discussion
3.1. Scale Reliability
Cronbach’s alphas for the three items of the Perceived Deceptiveness Scale ranged from 0.44 to 0.80. We removed the item with the lowest alpha, “Trying to show oneself in the best light”, for all stimuli. Alphas for the remaining items improved to around 0.80. A deceptiveness score was then calculated by summing the “Deceptive” and “Straightforward” scores, with ratings for the item “Straightforward” reverse-scored. Higher scores thus indicated higher deceptiveness.
Cronbach’s alphas for the Perceived Attractiveness Scale ranged from 0.70 to 0.80 for all models. An attractiveness score was calculated by summing the ratings of all five items, with ratings for “Bad impression-Good impression” and “Unattractive face-Attractive face” reverse-scored. Higher scores thus indicated higher attractiveness.
Cronbach’s alphas for the Ten Item Personality Inventory were low for all five subscales, ranging from 0.04 to 0.52. Cronbach’s alphas for the General Trust Scale, Tendency to fall in love with looks scale, and Tendency to fall in love with personality scale were 0.82, 0.82, and 0.75, respectively.
3.2. Attractiveness Manipulation Check
As indicated in Table 1, each model in the high attractiveness condition was rated as more attractive than her counterpart in the low attractiveness condition. This difference was significant for Model A, t(129) = 5.32, p < 0.01, and Model B, t(128) = 3.96, p < 0.01, but not for Model C, t(129) = 0.98, ns. This may be attributable to Model C’s frequent blinking in the videos, which participants may have found distracting.
3.3. Deceptiveness Analysis
For each model’s reply to each question, we conducted a 2 (condition: high vs. low attractiveness) × 2 (sex of participant) ANOVA on deceptiveness ratings. The results are presented in Table 2. Although there were relatively few significant differences in deceptiveness ratings, for all of the differences in condition that were significant (Model A: Questions 2 and 10; Model B: Question 13) and marginally significant (Model A: Questions 4, 13, 14; Model B: Question 3), deceptiveness was lower in the high attractiveness condition than in the low attractiveness condition. In general, deceptiveness ratings tended to support the hypothesis that more attractive people would be perceived as less deceptive, although the present effect was not strong.
3.4. Effects of Participants’ Sex and Personality Traits
As indicated in Table 2, sex of participants had little influence on responses to Models A and B, with only one significant effect for each model. However, six of the questions produced significant or marginally significant sex effects for Model
Table 1. Means (standard deviations) of attractiveness and results of t tests.
Table 2. Means (standard deviations) of deceptiveness ratings and results of ANOVAs.
Note. blue: p < 0.01; green: p < 0.05; yellow: p < 0.10.
C, all with males producing higher deception ratings than females. This difference may be attributable to more frequent blinking by Model C, as noted above.
Correlations between deceptiveness and the various traits identified in the personality measures were generally low, with the highest correlation coefficient around 0.30 in absolute value. Personality traits of participants thus appeared to be unrelated to perception of deceptiveness when attractiveness factors were implicated.
4. General Discussion
Unlike previous studies of the effects of attractiveness on deceptiveness perception, the present study employed attractive and unattractive female models who were actually the same person. Further, each model was able to speak in the stimulus video clips using her own words, rather than memorizing and rehearsing texts written by the experimenter. We obtained experimental evidence that more attractive females are perceived as less deceptive, in keeping with previous results by Porter et al. (2002) and Bull and Vine (2003) .
Some limitations of our method should be noted. Although our models could produce their own speech during the recordings, slight differences in speech content and ways of speaking between conditions may have arisen, because it is difficult to say the same thing twice in identical ways. In order to make the speech content more equivalent across conditions, professional actors who can speak the same content twice in the same way without reading should be employed.
Our models did all of their makeup by themselves, and for this reason, the makeover in the low attractive condition was rather extreme. Future studies should employ a professional makeup artist who understands the objectives of the research. However, the present technique of manipulating a female adolescent’s facial attractiveness is nevertheless applicable to future research.
Female attractiveness has many aspects, with facial attractiveness only one of them. Further, the application of cosmetics may have less effect than commonly assumed (e.g., Jones & Kramer, 2015 ). Future research should consider female attractiveness more broadly.
Studies which examine effects of males’ facial attractiveness on females’ perceptions of deceptiveness are clearly needed. In the present research, the stimulus persons were limited to females because we had no convenient way to manipulate a male’s attractiveness using widely available cosmetic products. As past research has revealed that females are better lie detectors than males ( Vrij, 2008 ), females’ perception of deception in a between-sex context may well differ from that in the within-sex context of the present study.
We believe that this study provides basic information on the relationship between attractiveness and deceptiveness. The results advance research on the effects of facial attractiveness of females on perceived deceptiveness, but may also potentially contribute to the understanding of heterosexual relationships, especially adolescent males’ perception of females and possible biases in this perception. In this way, this and similar studies may promote better relationships between males and females. It is suggested that theoretical explanations for the findings of this study should be explored in future studies (e.g., Maestripieri, Henry, & Nickels, 2017 ).
This work was supported by JSPS KAKENHI Grant Number JP26380850.
Jun’ichiro Murai designed the experiment, conducted the experiment, analyzed the data, and wrote the paper. Izuru Nose designed the experiment and checked the draft of the paper. Yuuta Takiguchi created the experimental stimuli.
This study was presented at the 2nd International Conference on Deceptive Behavior (Decepticon 2017: Truth, Trust, and Tech), Stanford University, Palo Alto, CA, USA, August 21-23, 2017.