Movement is a neglected component of many studies of facial perception, yet biological movement can influence social perceptions by conveying information about identity, sex, and disposition. We aimed to quantify the importance of facial movement in perceptions of attractiveness, trustworthiness, and dominance, by isolating static and dynamic cues using facial motion-tracking. 42 speed-daters were filmed talking with an opposite-sex partner during a speed-dating interaction. Motion-tracking was used to isolate the movement information and create point-light faces, which were rated along with the original videos and facial photographs, by 73 independent raters. Regression models showed that static cues were generally important predictors of overall perceptions, albeit less so for perceptions of trustworthiness. Dynamic cues only contributed to perceptions of male trustworthiness and female dominance. Movement can be a significant component some perceptions, perhaps because it indicates personality and behavioural disposition. Judgements of attractiveness might be more linked to biological quality which is indicated by static cues such as shape and colour, rather than dynamic cues.
Research Article
Movement and facial perception
https://doi.org/10.21203/rs.3.rs-3209351/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
Automatic judgements are made on the basis of people’s facial appearance. Although these judgements are potentially limitless, they can be summarised as falling across two primary dimensions: warmth/trustworthiness and competence/dominance (Fiske, Cuddy & Glick, 2007; Todorov, Said, Engell, & Oosterhof, 2008). These automaticity and consistency of these judgements may well reflect evolutionary mechanisms to quickly distinguish potential threats and allies from those who can be ignored. A third dimension—attractiveness—may identify suitable reproductive partners.
The main methods used to study face-trait judgements are ratings of photographs or computer-graphics techniques to experimentally morph faces. A neglected aspect of such studies is facial movement as most studies rely on static facial images. In real life, however, faces are dynamic and expressive. Dynamic cues are also important because they can indicate an individual’s disposition, which cannot be conveyed through static cues as it may change from moment to moment. Disposition is important as it provides a here-and-now signal of potential threat, or friendliness, or interest in mating. To assess whether static images accurately index judgements from videos, a recent meta-analysis compared attractiveness ratings between photographs and videos and showed good consistency, with an average correlation of r = 0.7 (Kaschel & Hildbrandt, 2023). Nevertheless, heterogeneity between included studies was high, and such a correlation still leaves about half of variance in ratings unexplained.
To truly assess the importance of movement in facial attractiveness, it is necessary to isolate dynamic cues. This is widely done in research on the body. Initially, Johansson (1973) showed how a point-light display of a body is meaningless when static, but looks strikingly human when in motion. More recently, motion capture has been used to show that gait conveys large amounts of social information, even about surprising things such as sexual orientation (Ambady, Hallahan, and Conner, 1999) and a woman’s position in her menstrual cycle (Provost, Quinsey, & Troje 2008).
Motion-capture has been very difficult to apply to the face because of practical limitations in the size of retroreflective markers needed, and the positioning and spatial resolution needed to accurately map out the face. For example, while a body can be represented with as few as 12 or 15 markers on major joints, a face may require many more to capture the more subtle movement of the eyes and mouth, as well as the general outline of the head. This proved a barrier to research until facial tracking methods were developed that could isolate facial movement directly from a video without the need for markers (e.g. Gralewski, Campbell, Morrison, & Penton-Voak, 2006). This technique has been used to show that sex-typical movement is attractive in faces (Morrison, Gralewski, Campbell, & Penton-Voak, 2007), and that women’s preferences for male facial movement vary across the menstrual cycle (Morrison, Gralewski, Campbell, & Penton-Voak, 2009).
A point-light face contains dynamic cues but not static ones, and a photograph contains static cues but not dynamic ones. Therefore, comparing the ratings made on these stimuli relative to that of the original video (containing both static and dynamic cues) will estimate the relative importance of dynamic versus static cues in overall trait judgments. We aimed to address this question by filming speed-daters interacting with each other, and motion-tracking their faces for playback to raters afterwards. Speed-dating is a series of pre-arranged dates where multiple dyads interact for a few minutes, before switching to a new partner and so on until everyone has met everyone else. It is a useful research tool because it involves real-life social interactions in a mating context, and people have limited time to make a good impression, motivating them to modulate their appearance, if that is possible.
Filming
48 participants (mean age 21.4 years, SD 2.7) were recruited for one of three speed-dating events. Participants were recruited by posters on the university campus and local cafes, and by adverts on the university intranet for staff and students. Participants were paid £10 for taking part. During the speed-dating event, men and women were separated until filming began to ensure that their interactions represented first impressions.
During the speed-dating itself, one man and one woman sat on opposite sides of a table and interacted freely for 3 minutes. Once the time was up, a buzzer sounded and one sex then rotated to the next table to meet a new partner. Men and women rotated alternately, resulting in each participant talking with every member of the opposite sex. For two of the tables, cameras (Canon Legria HFM 32) were placed behind the shoulders of the participants and recorded their interactions, focusing on the face. Facial photographs were also taken using a Nikon D90 camera. The study was approved by the ethics committee of the first author’s university.
After the speed-dating was completed, participants were thanked, debriefed and paid. We made no attempt to match participants if they liked each other, but we did inform them that they were free to make their own arrangements if they so desired. We hoped that this would motivate them to behave as they would in a real dating scenario.
Stimulus preparation
A 30s segments was taken from each video to give ample time for facial movement to be seen, typically involving the person talking and listening to his or her partner, while not being so long as to be boring to watch or practically difficult to track and analyse. Each of these segments was motion tracked using PsychoMorph software. PsychoMorph is a facial analysis and transformation program widely used for morphing facial photographs (Tiddeman & Perrett, 2002). It also has a tracker function, which allows it to track the position of landmark points of a face through each frame of a video sequence. This was done as follows.
First, the video was split into frames using VirtualDubMod (version 1.5.10.0), and the resulting JPEGs loaded into PsychoMorph’s tracker window. The first frame of the sequence was then delineated by hand to create a template consisting of 160 landmark points connected together to capture the position of the face (Fig. 1).
Two further frames were delineated by hand. Using these as a guide, the tracker then attempted to automatically delineate the remaining frames. This typically resulted in a moderate fit of the template to the underlying image. Two or three frames were chosen and corrected by hand, then the tracker run again, resulting in a better fit for all frames. For more details of this semi-automatically delineation process, see Morrison et al. (2010). This process was repeated, typically three of four times, until the fit was almost perfect. The template files were then saved as JPEGs using a PsychoMorph plugin, and these image files were reformed into a video using Photolapse (version 2.0.0.0). The end result was a “point-light” face that accurately represented the movement of the face without displaying and colour or texture information, and only minimal shape information. The filming or facial-tracking process failed for 6 of the videos, leaving 42 for subsequent rating.
Ratings
All raters were students at the first author’s university who received course credit upon completion of the study. 78 participants (mean age = 19.3 years) took part with 75 completing the study due to a dropout rate of 6.4%. Two participants answered the same score for all stimuli, therefore, their data were excluded from the study. Participants were randomly assigned to rate picture (n = 27), point-light faces (n = 22), or videos (n = 24).
Ratings were conducted using Gorilla. Raters viewed a participant information sheet and informed consent form, then were randomly assigned to the picture, point-light, or video conditions, where they viewed the stimuli in random order. Participants were shown pictures or watched 30 second clips, then rated the attractiveness, trustworthiness and dominance of each stimulus on a 10-point Likert scale, with 1 being “not at all” and 10 as “very”. Cronbach Alpha’s were conducted for all scales. For pictures, internal consistency was high for attractiveness (α = 0.97), trustworthiness (α = 0.94) and dominance (α = 0.93). For point-light faces, internal consistency was high for attractiveness (α = 0.94), trustworthiness (α = 0.94) and dominance (α = 0.88). For videos, internal consistency was high for attractiveness (α = 0.97), trustworthiness (α = 0.95) and dominance (α = 0.97). Therefore, these ratings were averaged to give an attractiveness score for each stimulus, which were used as the unit of analysis subsequently.
Descriptive statistics and sex differences
The average ratings made on the male and female stimuli were compared to see if they were rated differently (table 1).
Table 1: Descriptive statistics and sex differences. Data are mean (SD).
The results showed that female pictures were rated more attractive and trustworthy than male, male point-light faces were rated more dominant than female, and female videos were rated more trustworthy than male.
Zero-order correlations
To explore the relationships between the variables, we ran a series of correlations. First, we correlated the three different ratings separately for photographs, point-light faces, and videos. Separate correlations were run for male and female faces as the results were strikingly different (table 2).
Ratings of attractiveness and trustworthiness were generally positive related for all female stimuli, suggesting that women who were attractive were also judged as trustworthy, whether from static or dynamic cues, or both together. For male stimuli, this was only true for point-light faces. Ratings of attractiveness and dominance were positively related for male stimuli but only in videos for women. By contrast, dominance was not consistently correlated with trustworthiness judgements. Indeed, the relationship was negative or null in women, and negative in male photographs while positive in male point-light faces.
Second, we correlated each rating across the three classes of stimuli, again separately for each sex (table 3).
All ratings made on photographs were significantly correlated with ratings made on videos, suggesting that static cues give a good approximation of someone’s appearance even when moving. By contrast, ratings of attractiveness made on point-light faces were not significantly correlated with ratings of video attractiveness, but were so for ratings of dominance, suggesting that static cues might be important for dominance perception. For male, but not female faces, perceptions of trustworthiness were correlated from point-light faces and videos, suggesting that male trustworthiness might be influenced by dynamic cues. Interestingly, some perceptions showed correlations between pictures and point-light faces, suggesting that men who looked attractive also moved in an attractive way, and men and women who looked dominant also moved in a dominant way.
Regression models
To estimate the relative contributions of static and dynamic cues in overall ratings, regression models were run with ratings of the videos as the dependent variable, and of the point-light faces and photographs as independent variables along with sex of stimulus. Interaction terms between the variables were included initially but were removed from the final models as they were non-significant. Separate models were run for each sex as the correlations suggested different patterns of association for male and female stimuli (table 4).
For male attractiveness, the overall model was significant, F(2,19) = 60.3, p < .001. The same was true for female attractiveness, F(2,17) = 16.8, p < .001. In both cases, however, only static cues significantly predicted overall attractiveness, with dynamic cues non-significant. The interaction between static and dynamic cues was not significant. This suggests that static and dynamic cues act independently in attractiveness.
For male trustworthiness, the overall model was significant, F(2,19) = 5.6, p = .012. Only dynamic cues predicted overall trustworthiness. For female trustworthiness the model was not significant, F(2,17) = 2.1, p = .146, with neither static nor dynamic cues predicting overall perceptions. Again, the interaction term was non-significant, suggesting that static and dynamic cues act independently in trustworthiness.
For male dominance, the overall model was significant, F(2,19) = 33.1, p < .001. Only static cues predicted overall dominance. For female dominance the overall model was significant, F(2,17) = 22.3, p < .001, with both static and dynamic cues predicting overall perceptions. Again, neither interaction term was significant, suggesting that static and dynamic cues act independently in dominance.
Our findings suggest that facial movement plays a role in facial perception. Ratings made on photographs were related to those made on videos, suggesting that a face is not perceived dramatically differently when it is dynamic. Using motion-tracking to isolate dynamic cues suggests they were important in perceptions of trustworthiness and dominance, but not important in overall perceptions of attractiveness. This latter finding contrasts with Morrison et al. (2018), who showed that dynamic cues were as important as static cues in perceptions of attractiveness in female bodies.
Our findings support those of other studies that report positive correlations between the attractiveness of facial videos and pictures, even though we did not allow raters to see the same individual more than once, eliminating carry-over effects. The strength of our correlations is similar to previous studies of attractiveness (Kaschel & Hildbrandt, 2023). Furthermore, we extend this result for perceptions of trustworthiness and dominance, with a noticeably smaller but still positive correlation for trustworthiness.
This latter finding could be related to the finding that emotional expressions influence perceptions of dominance more than trustworthiness, although this study used static expressions rather than dynamic cues (Olzsanowski, Parzuchowski, & Szymków, 2019). Hehman, Flake and Freeman (2015) also considered facial expressions as dynamic and found that static cues yielded consistent perceptions, whereas dynamic cues yielded more variable perceptions. They found that perceived ability in the form of physical strength was not much affected by dynamic cues.
We find evidence for sex differences in facial perceptions, with female faces being rated more attractive and trustworthy, in line with previous research (Cross & Cross, 1971; Maret, 1983; McLellan & McKelvie, 1993; Morse, Gruzen & Reus, 1976; Wernick & Manaster, 1984). Male photographs and videos were not judged as more dominant, however, although their movement was. This suggests that isolating facial movement can reveal different insights into facial perception compared with photographs or videos. This idea is further corroborated by our findings that facial movement contributes to perceptions of male trustworthiness, and female dominance, but not vice versa. This replicates Hehman, Flake and Freeman’s (2015) result that perceived ability of male faces, in the form of physical strength, was not much affected by dynamic cues. Our results add that dynamic cues do affect perceptions of female dominance, and male trustworthiness. A speculative explanation is that static cues might drive judgements of male dominance as they indicate actual strength (Sell et al., 2009), whereas for female strength such cues are not as strong and so there is a role for dynamic cues. Likewise, static cues don’t seem so important in judgements of male trustworthiness, perhaps because there is less of an obvious biological link, leaving room for dynamic cues to be important.
Overall, static and dynamic cues explained much less variance in trustworthiness than in the other ratings. This could be because trustworthiness is less tied to underlying biology such as physical size, hormones, genes, or developmental history. Evolutionarily minded researchers have tended to theorise that attractive facial cues indicate biological quality such as good genes or a healthy early developmental history. However, there are at least two further factors relevant to mate choice: condition and disposition. Condition might be somewhat changeable e.g. due to temporary infection or current nutritional status, and might be cued not by facial shape but by skin texture and colouration (e.g. Stephen, Coetzee, & Perrett, 2011). Disposition, is even more changeable and can vary from moment to moment depending on the social context. Trustworthiness might therefore by more of a disposition compared with attractiveness or dominance, and subtler and harder to link to specific cues. Future research should consider physical attractiveness and facial perception generally as indicating any combination of biological quality, condition, or disposition.
Including dynamic cues increases our understanding of facial perception, yet it also introduces a new issue. The way that people move their faces depends on context. We induced a mate choice context as attractiveness was one of our main perceptions of interest. But people may well move differently in other contexts – consider a teacher in front of a class, a comedian entertaining a crowd, or a politician giving a speech. Dynamic cues in such contexts may by more or less important for different perceptions, depending on how people’s faces move. We encourage researchers in facial perceptions to explore such different contexts to see where dynamic cues matter more or less.
The study of facial movement using motion-capture or motion-tracking has great potential for many psychological studies of facial perception, in areas such as facial recognition, emotion production and emotion perception. Modern technology means that the dynamics cues involved in facial movement can be isolated and understood. Just as computer-graphics technology was instrumental in improving our understanding of static facial cues, we hope that techniques like those we used will open up new avenues of research in facial perception, meaning dynamic cues are no longer neglected.
Acknowledgments
This research was supported by ESRC grant RES-061-25-0487.
Author contributions
EM designed the study, assisted with collecting and creating the stimuli, secured the funding, analysed the data, supervised the other authors, and wrote the manuscript. ZO’D collected the rating data. NJG collected and created the stimuli.
Data availability statement
Data are available at the Open Science Foundation, including numerical data used in the analyses, and an example video. All videos are not available as this would breach privacy of the other individuals who have not all consented to have their data made public.
Morrison, E. (2023, July 26). Movement and facial perception. Retrieved from osf.io/zbkjm
Competing Interests Statement
The authors declare no competing interests.
- Ambady, N., Hallahan, M., & Conner, B. (1999). Accuracy of judgments of sexual orientation from thin slices of behavior. Journal of Personality and Social Psychology, 77, 538-547.
- Cross, J. F., & Cross, J. (1971). Age, sex, race, and the perception of facial beauty. Developmental Psychology, 5(3), 433–439. https://doi.org/10.1037/h0031591
- Fiske, S. T., Cuddy, A. J., & Glick, P. (2007). Universal dimensions of social cognition: Warmth and competence. Trends in Cognitive Sciences, 11(2), 77-83.
- Gralewski, L., Campbell, N., Morrison, E., Penton-Voak, I. (2006) Analysis of facial dynamics using a tensor framework. Journal of Multimedia; 6, 10–21.
- Hehman, E., Flake, J. K., & Freeman, J. B. (2015). Static and dynamic facial cues differentially affect the consistency of social evaluations. Personality and Social Psychology Bulletin, 41(8), 1123-1134.
- Johansson, G. (1973). Visual perception of biological motion and a model for its analysis. Perception and Psychophysics, 14, 201-211.
- Kaschel, P., & Hildebrandt, L. (2023). Can Beauty be Measured with Photos? A Systematic Review and Meta-Analysis on Static and Dynamic Physical Attractiveness Ratings. International Review of Social Psychology, 36(1): 5, 1–19. DOI: https://doi.org/10.5334/ irsp.758
- Maret, S. M. (1983). Attractiveness ratings of photographs of Blacks by Cruzans and Americans. The Journal of Psychology, 115(1), 113-116.
- McLellan, B., & McKelvie, S. J. (1993). Effects of age and gender on perceived facial attractiveness. Canadian Journal of Behavioural Science/Revue canadienne des sciences du comportement, 25(1), 135.
- Morrison, E. R., Gralewski, L., Campbell, N., & Penton-Voak, I. S. (2007). Facial movement varies by sex and is related to attractiveness. Evolution and Human Behaviour, 28, 186–192.
- Morrison, E. R., Clark, A. P., Gralewski, L., Campbell, N., & Penton-Voak, I. S. (2009). Women’s probability of conception is associated with their preference for flirtatious but not masculine facial movement. Archives of Sexual Behaviour, 39, 1297-1304. DOI 10.1007/s10508-009-9527-1.
- Morrison ER, Clark AP, Tiddeman BP, Penton-Voak IS (2010). Manipulating shape cues in dynamic human faces: sexual dimorphism is preferred in female but not male faces. Ethology, 116, 1243-1243.
- Morrison, E. R., Bain, H., Pattison, L., & Whyte-Smith, H. (2018). Something in the way she moves: biological motion, body shape, and attractiveness in women. Visual Cognition, 26(6), 405-411.
- Morse, S. J., Gruzen, J., & Reus, H. (1976). The “eye of the beholder”: A neglected variable in the study of physical attractiveness? 1. Journal of Personality, 44(2), 209-225.
- Olszanowski, M., Parzuchowski, M., & Szymków, A. (2019). When the smile is not enough: The interactive role of smiling and facial characteristics in forming judgments about trustworthiness and dominance. Roczniki Psychologiczne, 22(1), 35-52.
- Provost, M.P., Quinsey, V.L., and Troje, N.F. (2008). Differences in gait across the menstrual cycle and their attractiveness to men. Archives of Sexual Behavior, 37, 598-604.
- Sell, A., Cosmides, L., Tooby, J., Sznycer, D., Von Rueden, C., & Gurven, M. (2009). Human adaptations for the visual assessment of strength and fighting ability from the body and face. Proceedings of the Royal Society B: Biological Sciences, 276(1656), 575-584.
- Stephen, ID, Coetzee, V, & Perrett, DI (2011). Carotenoid and melanin pigment coloration affect perceived human health. Evolution and Human Behavior, 32, 216-227.
- Tiddeman, B, & Perrett, D. (2002). Transformation of Dynamic Facial Image Sequences Using Static 2D Prototypes, The Visual Computer, 18, 218-225.
- Todorov, A., Said, C. P., Engell, A. D., & Oosterhof, N. N. (2008). Understanding evaluation of faces on social dimensions. Trends in Cognitive Sciences, 12(12), 455-460.
- Wernick, M., & Manaster, G. J. (1984). Age and the perception of age and attractiveness. The Gerontologist, 24(4), 408-414.
No competing interests reported.