How do we recognize a face? SMCS helps analyze the mechanisms employed by the human brain
pftplus |
Running into a colleague on the street without recognizing them, or hesitating for several minutes when trying to identify a neighbor seen from an unfamiliar angle... These everyday situations are more common than we might think. And they are not simply a matter of poor memory: recognizing a face relies on highly complex brain mechanisms that engage both visual perception and memory.
Viewed from the front, in profile, in shadow, or after a change in hairstyle, the same face can present highly variable visual information that the brain must interpret and connect to previously stored memories. Yet, despite these variations, it can determine in a fraction of a second that it is the same person. But what cues does the brain actually rely on to achieve this?
Our experience teaches us to retain certain visual cues
A study conducted at UCLouvain by Valérie Goffaux, an FRS-FNRS Research Associate at the Psychological Sciences Research Institute (IPSY) and Alexia Roux-Sibilon (Laboratoire de-Psychologie Sociale et Cognitive - Université de Clermont-Auvergne) reveals the brain’s secret: it does not actually retain every detail. Through experience, it learns to prioritize certain visual cues from a face, particularly those that remain stable over time.
Remarkably, these cues can be identified objectively using mathematical tools. As demonstrated by the study published in eLife, the brain relies primarily on horizontally oriented visual information, such as the eyes, eyebrows, and mouth, to recognize facial identity. This horizontal structure provides not only the richest but also the most stable information about a person’s identity.
In other words, it enables the brain to link together different appearances of the same individual, even when they turn their head or change their facial expression.
A methodology at the crossroads of mathematics, psychology, and neuroscience
To identify precisely which visual cues the brain uses, the team combined experimental psychology, mathematical modeling, and advanced statistical analyses, in collaboration with UCLouvain’s Statistical Methodology and Computing Service (SMCS) platform.
Participants were presented with faces viewed from different angles, whose images had been filtered using a Fourier transform, a mathematical tool that makes it possible to isolate specific visual orientations. This approach allowed the researchers to objectively test the role of horizontal facial structures in identity recognition. The resulting behavioral responses were then analyzed using Bayesian statistical models.
The study exemplifies a truly transdisciplinary approach, bringing together psychological sciences, neuroscience, and applied mathematics.