The Face Memory Test is a five-round visual-recall game built entirely from fifteen original SVG illustrations of fictional people named Alder through Opal, never real photographs, never uploaded portraits, and never biometric samples. Each round briefly reveals one target profile during a two-tick countdown, hides the target, and then asks you to pick the same profile from three illustrated options. A clean match is worth 200 points, and matching all five profiles before a second distinct miss ends the run with exactly 1,000 points. The tool runs entirely in your browser, so no camera request, no file upload, no account, no machine learning model, and no network call happens while you play, and only the shared best completed score can remain in this browser. Because every profile is code-drawn, the experience is fully deterministic: the round order, the five targets, the three options per round, and the score are all fixed.
In academic settings, a face memory test usually means a psychology study such as the Cambridge Face Memory Test, the female Cambridge Face Memory Test long form, the EGEFACE static and dynamic face memory test, or super-recogniser research. Those benchmarks use real unfamiliar faces, are designed to probe recognition ability, and are administered under controlled conditions for research. The Face Memory Test described here is a short browser activity, not a research instrument. It does not store identity, does not compare uploads, does not estimate recognition ability, and does not report a score outside the browser. The shared similarity is the task shape: study a face, lose the face, pick the right one from a small set. The shared difference is the data: real photographs on one side, original illustrations on the other.
What the Face Memory Test Actually Shows
The Face Memory Test assembles every profile from a small authored vocabulary of geometric primitives. Each head outline is one of three shapes: round, oval, or square. Each profile draws from five hair shapes, three brow shapes, three eye shapes, three nose shapes, four mouth shapes, and optional cheek marks or freckles. The combinations are produced in code, so the same fifteen profiles play in the same order every time, and the correct option appears in the same labeled position each round.
A complete profile is an abstract SVG drawing, not a measurement of a person. The named profiles do not represent a demographic, a diagnosis, an emotion label, or a biometric template. The tool is not a facial recognition system, and it does not match an uploaded image to anything. It only asks which original illustration was shown a moment earlier.
The illustrations use semantic foreground, surface, accent, and border color variables, so the same artwork stays visible under the site's light and dark themes without downloading an image. The view boxes are stable, which means the layout does not jump when a profile mounts or unmounts. The answer buttons include text labels and visible focus, and a status line announces the current round, matches, and errors without relying only on color.
How to Play the Face Memory Test
The run is deterministic and short. A complete game is five rounds, and each round follows the same two-state flow: a Reveal phase that shows the target, and a Recall phase that shows three options. The two states are isolated. During Reveal, the target illustration is mounted and no answer buttons are visible. After two one-second ticks, or after you press Enter, Space, or the Recall now control, the target branch unmounts and three option buttons mount.
- Study the original illustrated target during the two-tick Reveal, or choose Recall now when ready.
- After the target unmounts, select the matching fictional profile with a large button or keys 1 through 3.
- Match all five profiles before a second distinct miss to finish with exactly 1,000 points.
The hidden target is not left behind with reduced opacity, clipping, alternate text, a data attribute, or a concealed option list in the active Reveal document. Once Recall is active, you make a choice and the round resolves. Unknown identifiers, options from another round, cursor moves outside the three-button tray, and actions attempted in the wrong phase are atomic no-ops, meaning they are ignored without changing the game state.
Scoring Rules and the 1,000-Point Finish
Every correct profile adds 200 points and opens the next target Reveal. Five correct answers therefore equal exactly 1,000 points. The result sequence is deterministic, so there is no shuffled answer position, no random portrait, no timer bonus, no adaptive model, and no hidden score adjustment. The score of 5 matches × 200 points = 1,000 points is fixed, and there is no other way to reach the displayed total.
The error model is forgiving but strict. The first time you pick a wrong profile in a round, the error is recorded and you stay in the round. If you pick the same wrong profile again in the same round, the duplicate is deduplicated and the error count does not change. A second distinct miss closes the run and disables further answers. The five fixed correct option positions are 2, 1, 3, 2, and 1 when counted as the labeled buttons shown to the player.
| Round | Correct Label | Result on Match | Running Score |
|---|---|---|---|
| 1 | 2 | +200 points, advance | 200 |
| 2 | 1 | +200 points, advance | 400 |
| 3 | 3 | +200 points, advance | 600 |
| 4 | 2 | +200 points, advance | 800 |
| 5 | 1 | +200 points, run ends | 1,000 |
The fixed label positions matter for keyboard play. When you press 1, 2, or 3, the labeled button in that slot is chosen, regardless of where the visual focus indicator is. If you use Left and Right to walk a selection, the focus indicator moves across the three buttons, and Enter or Space confirms the highlighted one. Up and Down deliberately do nothing because the option tray is a single row.
Keyboard and Mouse Controls
The pointer path uses three large illustrated buttons. Mouse or touch users simply tap the option that matches the target they remember. The answer buttons are wide enough to be hit accurately on a phone, and each one carries a textual label so the choice is also available to screen readers.
Keyboard users have two equivalent routes. Enter or Space during Reveal hides the target. After the three options appear, pressing 1, 2, or 3 chooses directly by label. Left and Right move the visible focus between adjacent buttons, and Enter or Space confirms the highlighted option. Up and Down intentionally do nothing because the option tray is a single row. Out-of-phase keypresses, unknown identifiers, and option clicks aimed at a tray that is not the current round are atomic no-ops, so a stray tap never derails the run.
The shared Restart control returns you to the first target and restores the two-tick countdown. If you need to hide the screen quickly, double-press Escape to open the simulated spreadsheet boss screen, and double-press Escape again to resume the same round. Both keyboard and mouse paths resolve the same fixed answer positions, so the controls are interchangeable rather than separate modes.
How This Differs From Academic Face Memory Tests
Searches for "face memory test" often lead to academic instruments such as the Cambridge Face Memory Test, the female Cambridge Face Memory Test long form, the EGEFACE static and dynamic face memory test, and super-recogniser studies. Those benchmarks use real unfamiliar faces, are designed to probe recognition ability, and are administered under controlled laboratory conditions for research.
The Face Memory Test is a short browser activity, not a research instrument. It does not store identity, does not compare uploads, does not estimate recognition ability, and does not claim to measure memory ability. It simply asks which original illustration was displayed a moment earlier. The fictional profiles Alder through Opal are fixture identifiers. They do not represent a person, demographic group, diagnosis, emotion label, or biometric template.
Because the artwork is code-drawn and the answer positions are fixed, the tool is useful for practicing the study-and-recall task shape with a transparent, repeatable stimulus. If you want a similar short visual recall task with different stimuli, try the Visual Memory Test for grid patterns, the Pattern Memory Game for 3×3 layouts, the Position Memory Game for shapes on a board, the Shadow Matching Game for silhouettes, or the Paired Associates Memory Game for cue-to-scene links. For a longer analogue-style experience, the Change Blindness Test and the Multiple Object Tracking Test also use original abstract scenes with no camera or upload.
What Runs, and What Doesn't
Everything in the Face Memory Test runs locally. There is no camera request, no file upload, no account, no API, no machine learning model, and no cookie-based identity. The illustrations are code-drawn SVG, so the game does not download a single image. The only value that persists between sessions is the shared best completed score, and that score lives in this browser only.
If you want to make the challenge harder, use Recall now to shorten the Reveal window, or rely on Left and Right instead of the direct number keys so you must confirm with Enter. If you want to make it easier, let both ticks finish before selecting, and click the buttons in the order they appear instead of relying on the keyboard. Either way, the round order, the answer positions, and the 1,000-point finish are unchanged, so the tool is a stable, auditable visual challenge rather than a randomized quiz.