Free Face Memory Test

How many faces can you actually hold in your head?

You meet a person, exchange thirty seconds of conversation, and half an hour later you see them again across the room. Do you know them? Most of the time the answer arrives instantly and without effort, which makes face memory feel like it costs nothing. It costs a great deal. Recognising a face means storing a configuration — the distances between eyes, the width of a jaw, the set of a mouth — and matching it later against every other configuration you have ever stored. This test measures how many of those configurations you can keep straight at once before they start collapsing into each other.

How the test works

Each round shows you exactly one face and asks a single question: have you seen this face before, in this session? Some rounds show a face you have already met. Some rounds show a face generated a moment ago, that you have never seen. Answer correctly and the level goes up by one. Answer wrongly and you lose one of three lives. The test ends when all three are gone, and your score is the level you reached.

This is called a continuous recognition task, and it is deliberately harsher than the study-then-test format used by most memory quizzes. There is no separate learning phase. Every face you answer about immediately becomes part of the pool you must remember, so the memory load grows with every single round rather than staying fixed. By level thirty you are holding dozens of faces, and by level sixty the pool is large enough that the repeats stop feeling familiar and start feeling merely plausible.

Why the faces are drawn, not photographed

Photographs are terrible test material. A photo of a person carries dozens of cues that have nothing to do with the face — a distinctive collar, unusual lighting, a slight head tilt, an out-of-focus background. People remember those cues, not the face, and the score ends up measuring picture memory instead. Every face here is generated as vector artwork from a small parameter set: head shape, hairstyle, hair colour, skin tone, eyes, eyebrows, nose and mouth. Two faces can differ by a single feature. That makes cheap shortcuts useless, and it also means no real person's likeness is ever stored or shown.

Reading your score

Treat any single run as a rough estimate. Recognition tests are noisy: a lucky sequence of obvious repeats early on can add ten levels, and one careless click at level five can end a run that would otherwise have gone far. Three or four attempts give a much more honest picture than one.

Getting a better score

The single biggest improvement comes from refusing to look at one feature. Beginners latch onto hair colour or a distinctive mouth, which works for about fifteen levels and then fails catastrophically once several faces share that feature. Take in the whole face as a shape — the relationship between the parts, not the parts themselves. That is how your brain recognises people you actually know.

The second trick is verbal. Give each new face a two-word label as it appears: "angry baker", "sleepy student", "square banker". Labels are cheap to store and easy to retrieve, and they give recognition a second route when the visual trace has faded. The third is simply patience. You have as long as you want on each face, and there is no time bonus, so a full extra second of looking at a new face is always a good trade.

How this differs from other memory tests

Face recognition is not general memory. There is strong evidence for partly specialised neural machinery involved in processing faces, which is why the ability dissociates so cleanly from other kinds of recall. People who score in the top few percent on digit span are frequently mediocre here, and people who cannot hold a phone number for ten seconds are sometimes outstanding. If you want to compare, the sequence memory and number memory tests on this site load the same working-memory system in a completely different way, and the gap between your scores is usually more informative than any one of them alone.

One honest caveat: this is entertainment and self-assessment, not diagnosis. Genuine difficulty recognising the faces of family and colleagues in daily life — prosopagnosia — is a real and well-documented condition, but a score on a browser game is not evidence for or against it. If real-life face blindness is affecting you, the right next step is a clinician, not a retake.