Skip to main content
SearchLoginLogin or Signup

‘The Roles of Featural and Configural Face Processing in Snap Judgements of Sexual Orientation’ by Joshua A. Tabak & Vivian Zayas (2012)

Published onApr 03, 2020
‘The Roles of Featural and Configural Face Processing in Snap Judgements of Sexual Orientation’ by Joshua A. Tabak & Vivian Zayas (2012)

Tabak J. A., & Zayas, V. (2012). The Roles of Featural and Configural Face Processing in Snap Judgments of Sexual Orientation. PLoS ONE 7(5): e36671.



Research has shown that people are able to judge sexual orientation from faces with above-chance accuracy, but little is known about how these judgments are formed. Here, we investigated the importance of well-established face processing mechanisms in such judgments: featural processing (e.g., an eye) and configural processing (e.g., spatial distance between eyes). Participants judged sexual orientation from faces presented for 50 milliseconds either upright, which recruits both configural and featural processing, or upside-down, when configural processing is strongly impaired and featural processing remains relatively intact. Although participants judged women’s and men’s sexual orientation with above-chance accuracy for upright faces and for upside-down faces, accuracy for upside-down faces was significantly reduced. The reduced judgment accuracy for upside-down faces indicates that configural face processing significantly contributes to accurate snap judgments of sexual orientation.

Critical Annotation 

Identifying the Dataset 

The goal of this experiment is twofold. The first is to test the role of configural (relationships between features i.e. distance) and featural face (individual features)  processing in judging others’ sexual orientation by their face. The second goal is determining if there is a statistical difference between studying male and female faces in correctly predicting whether someone is gay or straight. The data included tens of images of gay men, straight men, gay women, and straight women. The images were collected from the Facebook profiles of those who identified as gay/straight (or were identified by their friends) and live in one of 11 major cities in the United States (US). Photos of minors were excluded. 


This study aims to understand how sexual orientation is socially constructed based on facial features that contribute most significantly to others’ ability to correctly label individuals as gay or straght. In the first experiment, 24 University of Washington students, the majority of which were women, participated in the experiment for extra credit. The first experiment tests whether or not there is a difference in guessing men and women’s sexual orientation. The luminescence of the photos was adjusted to minimise the difference between photos of males and females other than facial features. The experiment consisted of the following: 1000 ms of a fixation cross, a target face stimulus for 50 ms, and a masked face for 50 ms. The images were randomly ordered and mixed in with control trials. 

The second experiment followed the same steps with 129 students participating (primarily female participants). For the second experiment, participants were presented with alternating upright and upside down faces. 

Key Assumptions Stated by Authors 

The authors hypothesise that upside down faces will have a lower hit or accuracy rate than upright pictures. Additionally, while no direction is taken on how this difference would manifest itself, the authors assume that there will be a difference in the accuracy between male and female photographs. 

Inferred Assumptions 

The authors make several claims that should be classified as assumptions. Firstly, the authors state that sexual orientation is less obvious, as opposed to race and gender. Secondly, based on the accuracy of the experimental results, the authors conclude that if sexual orientation is easily distinguishable by facial features, then its concealment and existence should not necessarily be considered as stigmatised (because sexual orientation is apparent via facial features). 

The above claims falsely represent gender, race, and sexuality by assuming their staticity. Like supposed identifiability of gender and race, this experiment assumes that sexual orientation is inherent to one’s appearance. That any of these features have the potential to be identifiable is based on arbitrary societal constructions and conditioning. The ground-truth interpretation of media representations to evaluate the validity of their results demonstrates an exaggerated reliance on stereotypes and mainstream characteristics. 

Moreover, the authors grossly misconceived the concept of stigma. Stigma does not arise whether or not sexual orientation is visible or concealed - it is born out of the cultural, structural, and legal injustices that non-heterosexual folks face on a daily basis. No amount of guessing sexual orientation from facial features will resolve these structural issues.

Lastly,  in the same way that categories of gender are reduced to male and female, sexual orientation has also been defined as a binary; gay and straight. This neglects the spectrum of sexual orientations that are included in the LGBTQI+ movement. 

Evaluating Results 

Hit rate (proportion of correctly guessed gay faces) and false alarm rate (ratio of incorrectly identifying straight faces as gay) are used to evaluate the success of the experiment. In the first experiment, women had a higher false alarm rate than men, which according to the authors, is slightly unexpected due to the widespread representation of lesbians in media content. In the following experiment, the ANOVA data analysis feature suggests that upright faces were more often guessed correctly and that women’s faces were judged more accurately. 

Given the reliance on student labour (which was not compensated beyond extra credit), the small dataset, and the assumptions that people can and do wear their sexuality on their face (albeit a binary sexuality), these experiments and results are unlikely to be applicable, relevant,  or reproducible elsewhere. 

No comments here
Why not start the discussion?