How do humans recognize one another?

Meike Ramon shares her insights in our newly launched interview series "Aha! What do we know about…". In this interview series, we ask researchers four questions to help us understand the world a little better. Meike Ramon is a Full Professor of Applied Data Science at Bern University of Applied Sciences (BFH), where she directs the Applied Face Cognition Lab.

interview by Hannah Schoch, edited by Elisabeth Abs

Meike Ramon is a Full Professor of Applied Data Science at Bern University of Applied Sciences (BFH), where she directs the Applied Face Cognition Lab. Her research asks a deceptively simple question: how do humans recognize one another, and why are some people astonishingly good at it while others struggle? She has studied the full spectrum of face recognition ability — from “face blindness” to so-called Super-Recognizers — combining cognitive neuroscience, psychometrics, and data science. She is known internationally for translating this basic science into practice, advising police and security agencies on how to identify and deploy their most able personnel responsibly. She is a vocal advocate for rigorous, transparent, and independent research. When she isn’t measuring faces, she dances ballet and enjoys cooking.

What is one insight from your research that you think everyone should know? 

The one thing I wish everyone understood is that we are far worse, and far more different from one another, at recognizing faces than we assume. We treat face recognition as automatic and universal — as if it were something every healthy brain simply can achieve. Two findings complicate that picture.

First, ability varies enormously from person to person, and it’s remarkably stable across life. At one end are people with “face blindness,” who may fail to recognize close friends or even their own; at the other are “Super-Recognizers,” who can pick a stranger out of a crowd based on a single image, or recognize someone years after a fleeting encounter. Most of us sit in between — with almost no insight into where.

Second, most people are surprisingly poor at matching unfamiliar faces — deciding whether two photos of someone you don’t know show the same person. We’re excellent with familiar faces and quietly unreliable with unfamiliar ones, and training doesn’t seem to positively affect rapid face matching or recognition.

Why does this matter? Because we build systems — eyewitness lineups, border checks, courtroom testimony — that assume everyone is an equally reliable judge of identity. They aren’t. So who is doing the looking matters as much as the procedure, and a confident “I’m sure it was him” deserves real caution. The “Aha!” is this: recognizing faces feels effortless and certain – because in real life we recognize many people that we are personally familiar with – and that very feeling is the trap.

What is one myth, misunderstanding or half-truth that you would like to see dispelled?

The myth I’d most like to retire is that facial recognition technology has “solved” face identity processing — that algorithms are objective, infallible, and have made the fallible human eye obsolete. It’s a comforting story, but both assumptions are wrong.

First, algorithms are not neutral. Their accuracy depends heavily on image quality and/or large training sets that are often time missing. Many systems perform measurably worse for some demographic groups than others, simply because of the data they were trained on. Second, humans haven’t been replaced: the best real-world results tend to come from combining capable people with good tools, not from substituting one for the other. In high-stakes settings a human still makes the final call — and which human does so matters enormously.

The deeper misunderstanding is that “the computer says so” feels objective in a way “the witness says so” doesn’t. But a technology can encode error and bias just as a person can — only at scale, and with a veneer of authority that’s harder to question. When a system flags or identifies someone, that output isn’t truth or certainty; it’s a probabilistic suggestion that needs human judgment, transparency, and the possibility of being wrong. 

The honest position is that humans and machines are both imperfect, that they fail in different ways, and that good identification means understanding those failures in order to minimize them as best as possible.

What is one finding that has surprised or delighted you most in the course of your research? 

What has delighted me most is watching a finding from basic science walk out of the lab and change real lives. When my colleagues and I began studying Super-Recognizers seriously, it could have stayed an academic curiosity: a neat fact that a small percentage of people are exceptionally good at recognizing faces. 

Working with police agencies, I saw that identifying and deploying these individuals could genuinely change the outcome of investigations — helping to recognize a suspect from poor-quality footage, or, just as importantly, to recognize that a face does not match and that someone has been wrongly flagged.

A surprise was personal rather than scientific. Some Super-Recognizers themselves are often completely unaware that they are unusual. Many have spent their lives assuming everyone remembers faces the way they do — sometimes even hiding it, because recognizing people who don’t recognize you back can feel socially awkward. Giving someone a name for their ability or naturally occurring difference from most of us can be quietly moving.

Rather than dismissing individual differences as variance or noise, we should all take them as what they really can be: a genuine resource — if we take the trouble to measure them properly and use them wisely. The thrill of this work, for me, is exactly that bridge: a question that began as “why are people different?” turning into something useful, careful, and human.

What do you consider to be one of the most important research questions currently being addressed—or that should be addressed—in your field? 

To me, the most pressing question in my field is deceptively basic: how do we know how good someone — or something — really is at identifying faces, before we rely on them? It sounds obvious, but the stakes are high and the tools are often weaker than people assume. We are increasingly making consequential decisions — about security, policing, and access — based on human and automated identification. Yet few people look into the details of how we select “experts” or how systems are validated. There are reports of scientists “identifying” Super-Recognizers using inappropriate tests (e.g. one on which even faceblind people achieve “normal” scores). For automatic solutions, the only way to anticipate their limitations and errors is to deeply understand how they were developed. Getting both aspects (i.e. human selection and system deployment) wrong has two potential costs: we could overlook genuinely able people, and we could place undue trust in unreliable technology.

So the research question I care about is how to build rigorous, transparent, and fair ways of measuring human abilities — using proper psychometrics and brain imaging and other objective techniques — and how to validate them across languages, populations, and real-world conditions rather than in the lab alone.

This is as much a question of governance as of measurement. If societies are going to use these capabilities, we need independent, evidence-based standards for how people and tools are tested, deployed, and held accountable. Getting the measurement right is the prerequisite for using these powerful tools in a way that is both effective and just. In a new project we want to build an open-source, validated, secure, and scalable SR certification platform. We aim to provide the first international standard for the assessment of naturally occurring human cognitive abilities that can be used by law enforcement and the judicial system – and any other professional domain.

If you are interested in a deep dive, Meike Ramon has some suggestions for further reading:

Russell R, Duchaine B & Nakayama K (2009). Super-recognizers: People with extraordinary face recognition ability. Psychonomic Bulletin & Review 16, 252–257. https://doi.org/10.3758/PBR.16.2.252

Ramon M & Vowels M (in press). Demonstrating the Super-Recognizer Advantage for Law Enforcement. iScience.

Ramon M (in press). A Review of Super-Recognizers: From Scientific Discovery to Societal Deployment. In: Barton JJ, Duchaine B, Freiwald W, O’Toole A & Yovel G (Eds.) The Face Book: An Oxford Handbook on Face Perception.

Nador JD, Uittenhove K, Gordillo D & Ramon M (2025). Super-Recognizers, or Su-Perceivers? Insights from fast periodic visual stimulation (FPVS) EEG. Brain Topogr 38, 61. https://doi.org/10.1007/s10548-025-01136-9

Ficco L & Ramon M (2026). Neural Reliability as a Marker of Face Processing Ability: Insights from Berlin Police Super-Recognizers. OSF preprint, osf.io/bnj2c

Fröhlich A, Dellwo V, French JP & Ramon M (2026). Zürich Voice Super-Recognizer Test (ZVSRT): Normative Data from Law Enforcement Personnel on Voice Discrimination and Sorting. OSF preprint, osf.io/r7gcw_v1

Aha! What do we know about...

For many of the decisions taken in democracies, we often rely on our general knowledge without double checking how up-to-date our understanding actually is. The interview series “Aha! What do we know about…” is curated and led by Reatch members. It makes current knowledge and research findings accessible in a concise and straightforward way: We ask researchers four questions to help us understand the world a little better and make more informed decisions.