Freeman, J. B., Johnson, K. L., Ambady, N., & Rule, N. O. (2010). Sexual orientation perception involves gendered facial cues. Personality and Social Psychology Bulletin, 36(10), 1318-1331.
Perceivers can accurately judge a face’s sexual orientation, but the perceptual mechanisms mediating this remain obscure. The authors hypothesized that stereotypes casting gays and lesbians as gender “inverts,” in cultural circulation for a century and a half, lead perceivers to use gendered facial cues to infer sexual orientation. Using computer-generated faces, Study 1 showed that as two facial dimensions (shape and texture) became more gender inverted, targets were more likely to be judged as gay or lesbian. Study 2 showed that real faces appearing more gender inverted were more likely to be judged as gay or lesbian. Furthermore, the stereotypic use of gendered cues influenced the accurate judgment of sexual orientation. Although using gendered cues increased the accuracy of sexual orientation judgments overall, Study 3 showed that judgments were reliably mistaken for targets that countered stereotypes. Together, the findings demonstrate that perceivers utilize gendered facial cues to glean another’s sexual orientation, and this influences the accuracy or error of judgments.
Note: This paper conducts and discusses three separate studies. Based on the structure of the paper, some subsections of the critical annotation below will discuss Study 1, Study 2, and Study 3, separately.
Identifying the Dataset
Study 1: The goal of this study is to examine the relationship between what is known as “gender inversion” (masculinity inverted to femininity and vice versa) and how well individuals are able to judge sexual orientation. This relationship is tested through computer generated faces. All aspects of the faces remain controlled except for shape and texture. The authors predict that as faces become more gender inverted, it is more likely that participants will judge them to be homosexual.
Study 2: The goal of Study 2 mirrors that of Study 1, but comprises real faces. Photographs of heterosexual and homosexual males and femals were scrapped from public domain advertisement websites. The faces were of individuals aged 18-25 and were free of glasses, jewellery, and facial hair. All the faces were placed against a white background for the experiment. There were a total of 30 straight females, 30 straight males, 30 homosexual males, and 30 heterseoxual female faces. All the images were standardised according to colour scales and size. Each of the photos were pretested based on gender.
Study 3: Here, the authors are testing a blatant assumption that was being made in Study 1 and Study 2. The previous experiments suggest that sexual orientation predictions would likely be highly skewed if the faces in the dataset do not prescribe to stereotypical representations. If this is the case, then atypical homosexual and heterosexual faces, according to the results of the other studies, are likely to be judged much less accurately. This dataset was obtained in the same manner as Study 2, however the dataset is larger - 80 faces for each category.
Study 1: The 34 participants in this study were undergraduate university students who were compensated with extra credit or $10. The dataset was produced by FaceGen Modeler and manipulated by the researchers resulting in 25 different types of faces ranging from extremely masculine to extremely feminine. Each participant was presented with 100 faces in a randomized order and asked for a sex judgement, gender judgement, and sexual orientation judgement for each face.
Study 2: Like study 1, there are 27 undergraduate student participants in this study (compensated with partial course credit). Each participant was tasked with categorising the sexual orientation of 60 randomised male faces and 60 randomised female faces. The faces were presented one at a time on the participants’ computer screen.
Study 3: 24 undergraduate students participated in this study in exchange for course credit or $10. Each target was presented to the participants in random order and were tasked with judging the sexual orientation of each face (all faces were pretested in a similar fashion to Study 1 and 2).
Key Assumptions Stated by Authors
One of the key assumptions is tested in Study 3 - the accuracy of judging sexual orientation based on faces changes drastically with non-stereotypical representations of homesexual and heterosexual individuals. Despite the extensive research presented by the authors on work done to understand how sexual orientation can or cannot be predicted, the authors still come to the conclusion that many of the factors remain obscure when understanding the factors involved in accurately guessing sexual orientation.
It is particularly telling that the paper begins with explaining the consequences of Presdent Eisenhower’s decree to remove all homosexual folks from government staff positions. While the authors do acknowledge the danger and misjudgement that this provoked, they don’t necessarily discuss the structural issues at the heart of this. In contexts around the world, folks from the LGBTQ+ community have consistently been othered. Continuing to study instances of whether or not other people can identify folks who are not heterosexual is a continuation of othering. These kinds of studies are not necessarily to strengthen group identity, as the authors mentioned. Moreover, to assume gender inversion as a variable to study boxes in the LGBTQ+ identity; this is self-contradictory as any aspect of any individual’s identity is intersectional and is unlikely to be a compilation of generic factors.
Study 1: Based on the experiment conducted, the authors conclude that perceived men who appeared feminine were more likely to be judged as homesexual - concluding that gender-inverted facial cues are used to infer sexual orientation. The authors state that because these faces were computer generated, they were much easier to manipulate and therefore inspire confidence in terms of the exact facial features that changed participants' judgments. Based on the results and the experiment, this study was primarily used to determine the extent to which texture and face shape affect sexual orientation judgements in a highly controlled setting. Refer to the assumptions sections for more on these results; however, as the authors mention, controlling faces and sexual orientation judgements with computer generated faces does not guarantee applicability to human faces.
Study 2: On the heels of study 1, the authors conclude that participants leveraged gender inversion cues (similar cues used by participants in the first experiment) to determine an individual’s sexual orientation. Furthermore, the authors argue that these cues yield fairly accurate judgements of sexual orientation, whether the faces are computer generated or real. Although, because this experiment is less controlled, the authors sense that other factors come into play and are unable to pinpoint them within the scope of this study.
Study 3: As hypothesised, when asked to judge atypical or non-stereotypical faces, participants hit very low accuracy levels and were less likely to judge correctly. Response biases from previous studies were exaggerated in study 3.