Face Memory Test

1. Let the portraits load. Every image is loaded before learning begins. Use a clear screen and your usual glasses if you need them.
2. Learn the faces. They appear one at a time and change automatically. Look at each person's features.
3. Find a familiar face. Tap the face you learned, or press its numbered key. Only one choice in each group belongs to the learning set. You can take your time choosing.
4. Try the new view. Later choices show the same identities from another angle. You do not see the learning pictures alongside the choices.
5. Read your result. Your answers are scored on our server. Leaving the tab during play ends the attempt so you can start a fresh one.

This is a visual task. Screen readers can announce the controls, but a verbal description cannot preserve the face-recognition challenge. Phone and desktop use the same images and rules; screen size and viewing conditions can still affect performance.

Accuracy is the number of correct choices divided by the number of choices. The two view scores show what happened in each part, with only a few answers in each. A single error can move the percentage noticeably. There is no passing mark.

Once enough eligible players have completed this version, the result shows a below, typical or above comparison band. Until then it shows the number of eligible runs collected. Comparisons use other players' first runs of the same version, with flagged runs excluded. Website visitors are a self-selected group, not a population norm. A missing comparison never changes your accuracy.

The face set is small. Retaking it makes the identities more familiar, even when their order and roles change. Shared links let a friend try the same set and order; the friend's displayed score is supplied by the link, not independently verified.

Duchaine and Nakayama's Cambridge Face Memory Test introduced a learn-then-recognize format with unfamiliar faces, including recognition across image changes [1]. Dunn and colleagues' UNSW Face Test combines challenging face memory and matching tasks, illustrating why a broader task battery matters when assessing individual differences [2].

Our game borrows the general learning-and-recognition idea. It is not either research test. Its synthetic faces are not the validated research stimuli; its length, image set and scoring are different. We do not import those tests' norms or use this score to assign an exceptional-ability label or diagnose a condition. The new-view images have been visually checked for identity consistency, but this bank has not been independently validated.