Participants in Xu and Adolphs' studies were shown examples of faces like those above and asked to group them based on personality traits.
Image Credit: Adolphs lab/Caltech
Scientific Frontline: Extended "At a Glance" Summary: Trait Judgments of Faces Without Word Labels
The Core Concept: A novel experimental framework demonstrating that humans can consistently categorize and judge facial features based on perceived personality traits without relying on language or predefined trait words.
Key Distinction/Mechanism: Traditional facial bias studies force participants to rate faces using experimenter-provided adjectives (e.g., "trustworthy," "competent"). This study eliminated language entirely, asking participants to spatially arrange faces on a computer screen based on perceived similarity in personality. Computational analysis of these spatial arrangements revealed the underlying dimensions people use to judge faces.
Major Frameworks/Components:
- Spatial Arrangement Method: A language-free experimental design where proximity indicates perceived personality similarity.
- Five Key Trait Dimensions: Computational modeling identified five distinct dimensions driving facial judgments: approachability, warmth, competence, morality, and youthfulness.
- Cross-Racial Generalization: A secondary experiment incorporating Black, Asian, and white faces confirmed that these five dimensions apply across races, while also highlighting the prominent role of racial stereotypes in snap judgments.
- Historical Concordance: The language-free spatial data strongly matched historical data obtained using traditional word-based rating scales.
Branch of Science: Psychology, Neuroscience, Behavioral Science.
Future Application: The language-free methodology enables facial bias research in non-English speakers, individuals with aphasia, pre-verbal infants, and potentially nonhuman primates. Future studies aim to track single-neuron activity in human subjects to identify where and when the brain processes these language-free facial judgments.
Why It Matters: It proves that facial bias and snap judgments are fundamental cognitive processes independent of language or specific cultural vocabularies, providing a more universally applicable model for understanding how first impressions influence human behavior and decision-making.
Imagine sitting on a city bus full of strangers. Look around at the people riding with you. If you needed help, how would you choose whom to trust? By scanning people's faces, your brain is already interpreting their personalities by judging certain facial features.
For decades, neuroscientists have studied how facial bias ultimately influences behavior, but experiments have been limited by the words used to describe people. Studies typically ask participants to rate faces on fixed rating scales provided by the experimenter using words like "competent" or "trustworthy." Now, a team of researchers at Caltech has developed a way to study the trait judgment of faces without using any word labels.
"We asked people to interpret the similarity in personality between faces directly," explains Yue Xu (PhD '26), who led the study and is the first author of a paper published online October 19 in Nature Human Behavior. "It's a more natural way of thinking about people's personalities without being cued to think about specific trait words."
In the past, studies aiming to link face perception with personality traits typically involved subjects looking at faces on a computer monitor and making quick judgments by answering questions like, "How trustworthy does this person look to you on a scale from 1 to 5?" The results of prior work have linked such snap opinions to real-life decisions regarding politicians, hiring, and more. However, they also raise questions about what happens if participants speak a different language, are given a particular set of words concocted by the experimenter, or are unable to use language at all.
The new study is the latest effort by Ralph Adolphs (PhD '93), Bren Professor of Psychology, Neuroscience, and Biology at Caltech and senior author of the paper, to reduce bias in face perception studies. Instead of rating faces on trait words, Adolphs, Xu, and the team asked more than 3,000 study participants across two experiments to look at faces on a computer screen without any rating scale. Instead, participants were instructed to arrange the faces based on perceived personality similarities; similar faces were placed close to one another, and different faces were placed farther apart.
Participants in Xu and Adolphs's studies were shown examples of faces like those above and asked to group them based on personality traits. Credit: Adolphs lab/Caltech
The first experiment focused on a set of predominantly white faces, allowing the researchers to make a direct comparison with a previous study using the same faces. The second experiment featured Black and Asian faces, in addition to white faces, to examine how the first experiment's findings extend to individuals of different races.
Using computational methods to assess how participants arranged the faces on their screens, the researchers identified five key trait dimensions present across both experiments: approachability, warmth, competence, morality, and youthfulness. As expected, the second experiment also showed prominent effects of stereotypes associated with race.
"It is important to note that what we are studying is people's bias and stereotypes because they actually do not know about these people outside of this task," says Xu, who was a Chen Graduate Fellow during her time at Caltech. "Our findings do not say anything about the true personality of the actual individuals represented by these faces but tell us what people's first impressions are when they look at these faces."
Next, the team compared their five key dimensions with historical data from prior experiments that used word ratings, such as "sociable" and "caring," and found they matched up well.
"When I took this task myself, it felt extremely difficult to do without language ratings, as is usually done," says Adolphs, who is also an affiliated faculty member of the Tianqiao and Chrissy Chen Institute for Neuroscience. "So, I was amazed that it worked at all. People gave consistent results that they could not have produced if they weren't doing the experiment properly."
Not only does the new method help reduce study bias by removing trait words, but it also makes language unnecessary, opening future experiments to people from different cultures and backgrounds, or those who are nonverbal.
"One obvious application would be that you could do the task with participants who cannot use the English language, perhaps because they are aphasic or do not speak English," Adolphs says. "I could imagine extending it also to infants or nonhuman primates. It opens up a whole new field of research since we now have a model of how we judge people's faces that does not depend on language."
Xu is continuing her research as a postdoctoral fellow at UC Riverside, and the Adolphs lab plans to study brain activity to help answer questions regarding where these faces are represented in the brain and at what point language enters the judgment process.
In one new project, Adolphs aims to record single-neuron activity in patients with epilepsy who have electrodes temporarily implanted in their brains to monitor seizures. This work will be performed in collaboration with electrophysiologist and neuroscientist Ueli Rutishauser (PhD '08), a faculty associate in biology and biological engineering at Caltech, an affiliated faculty member of the Tianqiao and Chrissy Chen Institute for Neuroscience, and a faculty member at Cedars-Sinai, where he directs the Center for Neural Science and Medicine. Such recordings allow for detailed investigations of these questions as subjects look at faces—providing more precise timing than is possible with standard neuroimaging methods like functional magnetic resonance imaging.
Funding: The work was supported by funding from the National Institute of Mental Health, the Simons Foundation Collaboration on the Global Brain, and the Japan Science and Technology Agency.
Published in journal: Nature Human Behaviour
Title: Trait judgements of faces using a task without word labels
Authors: Yue Xu, Umit Keles, Chujun Lin, and Ralph Adolphs
Source/Credit: California Institute of Technology | Katie Neith
Edited by: Scientific Frontline
Reference Number: psy102026_01
