Researchers show that super-recognizers — about 1–2% of people — excel not by gathering more visual data but by selecting higher-quality, more diagnostic information when they look. An eye-tracking study of 37 super-recognizers and 68 average observers fed gaze patterns into nine neural networks; AI face-matching improved significantly using the super-recognizers’ gaze data, even with equal visual input or partial occlusion. The study, published in Proceedings of the Royal Society B, indicates the advantage comes from early visual encoding rather than later brain processing and calls for tests in dynamic settings like video.
How Super-Recognizers Spot Faces the Rest of Us Miss
Researchers show that super-recognizers — about 1–2% of people — excel not by gathering more visual data but by selecting higher-quality, more diagnostic information when they look. An eye-tracking study of 37 super-recognizers and 68 average observers fed gaze patterns into nine neural networks; AI face-matching improved significantly using the super-recognizers’ gaze data, even with equal visual input or partial occlusion. The study, published in Proceedings of the Royal Society B, indicates the advantage comes from early visual encoding rather than later brain processing and calls for tests in dynamic settings like video.
09:52, 06.11.2025Science
How Super-Recognizers Spot Faces the Rest of Us Miss
Some people barely ever forget a face. A brief encounter at a party can leave an impression so lasting that these individuals can pick the same person out years later in a crowd. Known as super-recognizers, they make up roughly 1–2% of the population, and evidence suggests the ability is more likely inherited than learned.
Why the skill matters
The ability has practical value: super-recognizers outperform many current AI systems on certain face-recognition tasks, and people who score highly on standardized face-recognition tests can receive professional certification to work with law enforcement.
What recent research found
New research clarifies why super-recognizers excel. Rather than simply gathering more visual data, they selectively extract higher-quality, more diagnostic information from faces. The study, published in Proceedings of the Royal Society B: Biological Sciences, combined eye-tracking data with artificial neural networks to test which gaze patterns were most informative.
“Super-recognizers don’t just look harder, they look smarter — they select the most useful facial information,” said James Dunn, research fellow at the University of New South Wales and lead author of the study.
How the experiment worked
Researchers recorded eye movements from 37 identified super-recognizers as they viewed still photos of faces, logging where and how long they looked. They ran the same eye-tracking procedure on 68 people with average face-recognition skills. The team converted each group's gaze maps into inputs for nine different neural networks trained to judge whether pairs of faces belonged to the same person.
Key results
When the neural networks received the gaze-derived information from super-recognizers, the AIs matched faces more accurately than when given inputs based on average observers’ gaze. This advantage persisted even when the overall amount of visual information was controlled and when parts of faces were obscured.
The findings suggest the edge of super-recognizers lies in what their eyes sample early in visual intake (the retinal/input stage) rather than in later, higher-level brain processing. In practice, they sample broadly across a face to find the most diagnostic features for that particular person, rather than relying on any single “critical” feature.
Limitations and next steps
The study used static images. The authors note future work should test dynamic situations such as video, live interactions, and varying lighting or viewpoint conditions to see whether the same gaze advantages hold.
Either way, it appears harder to slip by a super-recognizer than you might think.
Want to try?
Try published tests if you’re curious about your own abilities: the UNSW Face Test (University of New South Wales) and the Cambridge Face Memory Test (Birkbeck, University of London).
Lead image credit: Fractal Pictures / Shutterstock. This story originally appeared on Nautilus.