Research into first impressions is a well-established area. Hundreds of studies have been published with the goal of understanding how the subtleties of facial features cue assumptions about those we meet. Often, the stimuli used are tightly controlled, with some sets using faces whose features are digitally manipulated to be larger or smaller by tiny degrees; the effect of miniscule alterations to the wideness of eyes, for example, can be isolated and analysed without changing any other aspect of the face. By eliminating as many extraneous variables as possible, research teams hope to get a reading of exactly what specific features contribute to the first impressions we form.
While on the surface this sounds like a reasonable and solid scientific approach, it does tend to create one pressing problem in particular. If you’ve ever participated in or run a study using face stimuli, it’s likely that the faces used were exclusively White.
Authors typically give no explicit reasoning for this choice within their published work. Richard Cook of Birkbeck, University of London and Harriet Over of the University of York believe that there are four broad reasons for this common decision. In their recent paper in Royal Society Open Science, the two deconstruct the assumptions behind possible reasonings, and examine the limitations imposed on the field by avoiding non-White face stimuli.
A Confounding Factor?
One of the concerns researchers may have about including non-White faces is that participants will infer traits of face stimuli based on just on ethnicity, regardless of the actual features of the face presented.
While it’s understandable to seek to limit the effects of racial bias when using face stimuli, the authors argue that this particular concern doesn’t stand up to scrutiny. In the lab, slight alterations in facial structure of Black face stimuli have been shown to affect the responses of White participants, suggesting that White viewers can form first impressions of Black faces based on features. Many studies also illustrate that participants do not apply blanket judgements to faces belonging to a particular group. For example, in a policing and judicial context, Black people with facial features that look more “stereotypically Black” (a term used centrally in the original 2006 paper) have been found to receive harsher sentencing decisions than those who do not.
The Other-Race Effect
Another concern for many researchers is the fallout of the Other Race Effect, in which some people have issues discerning between faces of those from races different to their own. It is often assumed that participants may lack the necessary perceptual expertise to process these faces, so datasets containing “other race” faces may not produce meaningful results.
However, the authors point out that the Other Race Effect produces relatively mild face processing deficits compared to face processing disorders such as prosopagnosia; even then, people with prosopagnosia nevertheless make judgements of facial traits typical of those without the condition. This concern also fails to consider the diverse environments in which people are raised. In many cities, locals are likely to have grown up with exposure to individuals from a large number of races. And of course, many people have family members of a different race. This leads the authors to believe that the Other Race Effect poses little practical concern, which is a view that seems to be supported by inconsistencies in racial exclusion criteria between many first impression studies — some studies allow non-White participants to assess White face sets, whereas others exclude non-White participants.
In addition, many face stimuli sets are computer generated — at times obviously so. This leads the authors of this paper to pose the question: “Given that synthetic faces with which participants have little or no perceptual experience can be valuable stimuli, then why not faces of colour?”
Overshadowing Subtle Effects
When inserting non-White faces into sets of White faces, it is possible that race would become a very salient factor, and so researchers may be concerned that participants will attend to race above all else, ignoring subtle changes in features.
The authors of this paper, however, argue that this is not necessarily something to shy away from. To the contrary, they assert that this is a key part of what the field of first impressions seeks to understand. Perceived ethnicity is an influential social cue — if we accept that first impressions based on ethnicity is quantitatively similar to other types of first impression (eg. based on sex or age), then it makes sense that this is a variable that we should seek to investigate, rather than exclude. Many studies already use interleaved male and female faces, and if ethnicity exerts a comparable level of effect, the authors emphasise there should also be few issues with interleaving diverse faces in research studies.
If studies require the ethnicity of face stimuli to be controlled, the authors suggest that this can also be achieved while including non-White faces. For example, all Black or all Asian face stimuli sets would be as effective at removing the influence of race as an all-White set.
The Logistics of Using Diverse Face Stimuli
In years gone by there were relatively few diverse face stimuli sets for researchers to choose from. Due to the effort involved in creating new stimuli sets, researchers looking to include non-White faces were faced with the prospect of spending hours perfecting their own set. For many, the impracticality was too much, and the readily available all-White sets were often chosen.
This, however, is changing. These days, there are several readily available diverse face sets, and the authors also note that computer generated sets allow for non-White faces to be generated. Even so, it appears that this option is rarely used within research — likely for the reasons discussed above. With these concerns addressed, however, hopefully that feature will see more use.
In recent years, psychologists have increasingly recognised the problems of studying only people from WEIRD (Western, educated, industrialized, rich, and democratic) backgrounds. Broadening stimuli to more resemble the real world, including widening investigations to assess first impressions across all races, is an important piece of this puzzle. Beyond being just a tick-box exercise, there is the potential to unearth findings that have substantial real-world impact; for example, understanding the types of faces that receive the harshest first impressions could help combat bias in everything from university applications to court judgements. And, though this paper focuses on research into first impressions, this isn’t the only topic that uses face stimuli. There are many other areas of psychology that make use of face stimuli whose validity may be hindered by not using stimuli sets that reflect the diversity of the real world. Without diverse data, we can’t know for sure.
Thankfully, we are beginning to see progress on this issue, and there have been some promising steps in the right direction in recent years. For instance, there have been replication attempts of studies focusing on dominance and trustworthiness in facial features, re-run with stimuli comprising equal numbers of Black, White, Asian, and Latin faces. Going forward, the authors believe that this amount of diversity should be routinely used. And, where researchers do not include diverse stimuli, explicit statements of reasoning behind this would be useful.
Perhaps most centrally to this argument, it is important that research into faces move past this so as not to imply that White is the default. In generations before us, this was very much the commonly held view. But in 2021, we are moving beyond that, and our stimuli need to make that journey too.
Emma L. Barratt (@E_Barratt) is a staff writer at BPS Research Digest