Abstract
Studies that attempt to define facial attractiveness often do so in terms of structural features of the face (e.g., symmetry, averageness). However, these studies typically use static images of faces that may not be analogous to dynamic faces that are frequently used in other areas of attractiveness research, such as research investigating the impact of attractiveness on social interaction. The current studies investigated similarities and differences in how dynamic and static faces are perceived and evaluated. Study 1 demonstrated that dynamic and static faces are judged by different evaluative standards. Study 2 demonstrated that perceived emotion may be more salient in judging the attractiveness of dynamic faces than in judging the attractiveness of static faces. These findings illustrate the need to more fully explore the differences between dynamic and static faces to facilitate a better understanding of the characteristics underlying perceived attractiveness.
How does physical attractiveness affect our lives? This question has been of interest to philosophers for centuries and the focus of research by psychologists since the early 1900s. This research, much of it focusing on facial attractiveness, has provided robust findings regarding the perception of attractiveness and the impact it has on our social lives.
One consequence of the vast extant literature on attractiveness has been the desire to summarize the findings in the field. A few researchers have approached the daunting task of meta-analyzing this literature. Such reviews have focused on the attributions people make on the basis of physical appearance (e.g., Eagly, Ashmore, Makhijani, & Longo, 1991), how treatment directed toward individuals is related to their attractiveness, and how people's own behavior is related to their attractiveness (e.g., Feingold, 1992; Langlois et al., 2000). Numerous commonalities emerge from these reviews. People readily attribute positive characteristics to attractive people and negative characteristics to unattractive people (e.g., Heilman & Stopeck, 1985). People are judged and treated differently as a function of their attractiveness (e.g., Marlowe, Schneider, & Nelson, 1996), and actual behavioral differences are associated with facial attractiveness (e.g., Hamermesh & Biddle, 1994). These meta-analyses present robust findings that provide a valuable structure to investigations of attractiveness. However, one aspect of these reviews worth noting is their high heterogeneity. All of these meta-analyses found inconsistencies in the magnitude of effect in the reviewed studies. A potential explanation for this variability may be the studies' methodologies. Specifically, the inconsistencies may arise from variability in the manner in which faces were presented for evaluation.
In their meta-analysis of the reliability of attractiveness ratings, Langlois et al. (2000) indicated that different studies used different kinds of stimuli, including static photographic images, dynamic video images, and in situ encounters. These modes of stimulus presentation are very different perceptually and may convey different types and amounts of information (Lander, Christie, & Bruce, 1999). Information about facial characteristics, emotional expression, and other evaluative dimensions may be differentially salient in these formats. Thus, studies that use different types of facial stimuli to investigate ostensibly the same issue may in fact be investigating different issues.
Understanding the similarities and differences of the various kinds of stimuli used in attractiveness research is essential to fully understanding the concept of attractiveness. The two studies presented here were conducted to evaluate two different stimulus types: dynamic faces and static faces.
STUDY 1
Study 1 investigated the relationship between attractiveness ratings of static and dynamic faces. Most studies investigating the impact of facial attractiveness provide measures of interrater reliability. However, they measure reliability within a single stimulus format (e.g., static or dynamic). The current study presented the same faces in different formats so that evaluations in the two formats could be compared.
Method
Participants
Seventy undergraduate students (39 male, 31 female) participated. All students received course credit for participation and were recruited from introductory psychology classes. Fifty-four participants were Caucasian, 6 were African American, 3 were Asian, and 7 reported mixed ethnicity.
Stimuli
Forty-eight female, Caucasian women served as models for the stimuli. The models were filmed with a Sony digital camcorder (DCR-VX1000) for 30 s while reading text from a cue card. They were requested to remove all jewelry prior to filming, had their clothing covered by a neutral-colored drape, and were requested to keep a neutral expression on their faces while reading. The resulting videos were captured on a Macintosh G3 computer using Media 100 1x (version 4.5). The videos were standardized for size, contrast, brightness, and color balance using MediaCleanerPro 3.0. Dynamic stimuli were created by editing the videos to 10-s clips and saving them in Quicktime format at 30 frames per second. A pilot group of 20 undergraduates rated the emotional expressions in these clips using a 5-point scale (1 =very negative, 5 =very positive) and verified that the expressions were neutral (M= 2.90, SD= 0.21). Static stimuli were created by taking from each video a frame in which the face maintained a neutral expression with the mouth closed and eyes open, looking forward. Thus, dynamic and static versions of each face were created as stimuli.
Procedure
Participants were seated in a classroom facing a projection screen. All seats had unobstructed views of the screen and were close enough to allow for detailed examination of the stimuli. Stimuli were presented in separate dynamic and static blocks, with block order counterbalanced between groups. Each face was presented only in a single format (dynamic or static), to prevent familiarity effects. Thus, each participant saw 24 faces in a dynamic format and the other 24 faces in a static format. Participants judged the facial attractiveness of each face on a 5-point scale (1 =very unattractive, 5 =very attractive). Stimuli were presented for 10 s each, with intervals of 5 s between images.
Results and Discussion
The reliability of the attractiveness ratings was calculated using coefficient alpha. Agreement for both the dynamic and the static stimuli was very high (α= .94 and .90, respectively). The average attractiveness ratings for the two kinds of stimuli were similar, with a mean rating of 2.94 (SD= 0.92) for the dynamic stimuli and 2.87 (SD= 0.95) for the static stimuli. A paired-sample t test revealed no significant difference between the attractiveness ratings of the two kinds of stimuli, t(34) = 0.303, p= .767.
Calculating the correlation between the attractiveness ratings of each face in the static and dynamic formats yielded a noteworthy result. Across the 48 faces, this correlation was relatively low (r= .19, p= .26), indicating that the rating of a face in the dynamic format did not necessarily match the rating of the same face in the static format.
The absence of significant differences between the descriptive statistics for the dynamic and static stimuli is not surprising. Many studies have used static or dynamic faces and found high reliability among attractiveness raters and similar means and ranges for these different types of faces. One of the cited meta-analyses of the attractiveness literature emphasizes this point. Langlois et al. (2000) calculated the reliability of attractiveness ratings for 67 studies using adults from the same culture. Average reliability for this sample was high, r= .90, and stimulus form was not a significant moderator variable. Regardless of the stimulus type, attractiveness ratings tend to have similar mean values and variability.
The finding of note from this study is the low correspondence between the ratings of the same face in the two different formats. A face that received a high attractiveness rating when presented in a dynamic form did not necessarily receive a high rating when presented in a static form. Thus, the results of Study 1 lead to the following conclusions: First, raters show high reliability when rating the attractiveness of dynamic faces; this result indicates the use of a common evaluative standard. Second, raters show high reliability when rating the attractiveness of static faces; again, this result indicates the use of a common evaluative standard. Third, ratings of the same face in dynamic and static forms do not correspond; this result indicates that different evaluative standards are used to judge the attractiveness of dynamic and static faces. This pattern of results raises the following question: What standards of attractiveness are used to judge faces in these different stimulus forms? Study 2 was designed to begin an investigation into this issue.
STUDY 2
What defines an attractive face has recently been the subject of much study and debate. Leading theories equate attractiveness with exaggerated features in the human face (e.g., Perrett et al., 1998), high symmetry (e.g., Jones et al., 2004), or a mathematically average configuration (e.g., Rubenstein, Kalakanis, & Langlois, 1999). Although these are competing explanations, two commonalities link these theoretical views. First, all of these explanations lead researchers to investigate attractiveness evaluations through a structural analysis of the human face. Second, all of these explanations lead researchers to utilize static images in their investigations. The perception of static images may in fact be related to a pure structural analysis of facial characteristics (O'Toole, Wenger, & Townsend, 2001), but this structural approach may not be sufficient to define attractiveness in dynamic faces. As noted earlier, static faces may convey different amounts of information than dynamic faces (Lander et al., 1999). Characteristics such as emotional expression, age, and facial adiposity may be more salient in dynamic than static presentations and thus may be more relevant to attractiveness evaluations of dynamic than static faces.
Although the stimuli from Study 1 were pilot-tested for emotional neutrality, the possibility remains that an attribution of emotion can influence perceptions, and such an effect may have led to the difference between ratings of static and dynamic faces in Study 1. Montepare and Dobish (2003) indicated that emotion can be perceived in faces not intentionally displaying emotion. Such perceptions may arise from certain facial characteristics related to structure or motion, as well as from traces of emotions that are present. The social context of an attractiveness evaluation and the perception of other facial cues could allow for an attribution of emotion even though none is being displayed. Study 2 began an investigation of this issue by focusing on the relationship between perceived emotion and attractiveness in faces.
Method
Participants
Eighty undergraduate students (46 male, 34 female) participated. All students received course credit for participation and were recruited from introductory psychology classes. Sixty-two participants were Caucasian, 10 were African American, 4 were Asian, and 4 reported mixed ethnicity.
Stimuli
Stimuli were the 48 dynamic faces and 48 static faces created for use in Study 1.
Procedure
Participants were seated in a classroom facing a projection screen. All seats had unobstructed views of the screen and were close enough to allow for detailed examination of the stimuli. Half the participants saw the dynamic faces, and half saw the static faces. Stimuli were presented in separate attractiveness-rating and emotion-rating blocks. Participants saw all 48 faces in the attractiveness-rating block and then the same faces in the emotion-rating block. Block order was held constant to allow emotion to be rated in the context of attractiveness evaluations. Participants judged the attractiveness of each face on a 5-point scale (1 =very unattractive, 5 =very attractive) and the emotional expression of each face on a 5-point scale (1 =very negative, 5 =very positive). Stimuli were presented for 10 s each, with intervals of 5 s between images.
Results and Discussion
Coefficient alphas were calculated to assess the reliability of the attractiveness and emotion ratings. Both sets of ratings had relatively high reliability for both the dynamic faces (α= .95 and .89, respectively) and the static faces (α= .95 and .96, respectively).
The attractiveness ratings for the faces were very similar to those obtained in Study 1, with a mean rating of 2.90 (SD= 0.97) for the dynamic faces (correlation with Study 1: r= .94, p < .001) and a mean rating of 2.85 (SD= 0.91) for the static faces (correlation with Study 1: r= .90, p < .001). As in Study 1, the correspondence between attractiveness ratings for the dynamic faces and attractiveness ratings for the static faces was relatively low (r= .21, p= .19). As expected, the emotion ratings of the dynamic faces (M= 3.05, SD= 0.60) were more variable than the emotion ratings of the static faces (M= 2.96, SD= 0.14). Correlational analyses were conducted to determine any association between the attractiveness ratings and perceived emotion in the faces. A strong relationship was found for the dynamic stimuli (r= .48, p < .001), indicating a trend for positive emotion to be perceived in highly attractive faces and negative emotion to be perceived in unattractive faces. No such relationship was found for the static stimuli (r= .11, p= .33); the emotion in the static faces was generally perceived as neutral.
Two findings are of note. First, emotion is a relevant cue in the perception of attractiveness in dynamic faces. In the context of an attractiveness-rating task, emotion may be perceived as the result of this social judgment. Given the relatively low variability of the emotion ratings in the pilot study for Study 1, it appears that emotion became more relevant in the context of an attractiveness-rating task. Second, perception of emotion in static faces does not appear to be influenced by the context of an attractiveness-rating task. This could be due to the fact that the standard used to evaluate static images is largely rooted in a structural analysis of the face (e.g., Grammer, Fink, Juette, Ronzal, & Thornhill, 2002; Rubenstein, Langlois, & Roggman, 2002) and is therefore resistant to an attribution of emotion. These results highlight the differences in the perception of dynamic and static faces. Further, it appears that structural cues are not the only basis for evaluating attractiveness; although static faces may be evaluated primarily on the basis of their structure, at least one socially relevant cue, emotion, influences evaluation of the attractiveness of dynamic faces.
GENERAL DISCUSSION
A vast body of literature has demonstrated that raters show high agreement regarding what is and is not attractive in the human face. The studies reported here confirm this long-standing pattern of agreement but introduce an important qualification: Different types of facial stimuli may be evaluated according to different standards. Although mean attractiveness ratings and ranges are similar across studies involving dynamic and static faces, a closer investigation reveals that these two kinds of stimuli are perceived differently. Studies investigating the characteristics that make a face attractive typically utilize static images and focus on structural components of the face (O'Toole et al., 2001). Studies involving dynamic faces may make certain structural features more salient (e.g., facial adiposity) or introduce nonstructural features (e.g., emotion).
The differences between dynamic and static faces have strong implications for research investigating characteristics that make a face attractive. Summarizing reliability across studies utilizing different kinds of stimuli may be misleading in that raters tend to use different standards of attractiveness for different kinds of stimuli. Furthermore, conclusions reached in studies of static faces may be only one part of the attractiveness equation for dynamic faces. It is obvious that people use a common evaluative standard to judge dynamic faces. Perceived emotion may be part of that standard. However, it should be noted that the goal of the present studies was not to define the standard used to evaluate dynamic faces, and such a goal is beyond the scope of this investigation. The findings from this research indicate only that emotion becomes more salient in the evaluation of dynamic faces than in the evaluation of static faces. A host of other characteristics may be equally salient or more salient. Thus, a critical goal of attractiveness research will be to unravel the aspects of dynamic faces that are salient to observers and how they affect perceptions of attractiveness.
Footnotes
Acknowledgements
I would like to thank Meghan Nolan for her help with creating stimuli and Laura Levit for her help with data collection. I would also like to thank J. Dixon for his comments on early versions of this manuscript.