Visual Memory

About visual memory

Visual memory is the cognitive system that briefly stores the position, shape and configuration of objects you have just seen. The grid-recall paradigm used on this page is descended from Phillip Corsi's 1972 block-tapping task, which is still the standard neuropsychological measure of visuospatial span. Where Corsi tests sequential order, this tool tests simultaneous-pattern recall: you see the lit squares all at once, hold them for roughly 1 to 1.8 seconds while the grid is shown, and then click them back from memory. That difference matters because parallel-pattern recall recruits the right posterior parietal cortex more heavily than sequential recall, which leans on the dorsolateral prefrontal loop.

The game scales difficulty along two axes: the count of lit squares (level + 2) and the grid size (3x3 grows to 7x7 every three levels). That double ramp prevents the early plateau most simple memory tests hit around 7 items, the magic-number-seven span George Miller reported in 1956. With three lives, the natural game terminates somewhere between level 9 and level 15 for most adult players. The score reported is the last completed level before two errors emptied your lives.

How the test works

Lit squares at level L  = min(grid_squares - 1, L + 2)
Grid size at level L    = min(7, start_size + floor(L / 3))
Flash duration (ms)     = max(800, 1800 - L * 80)
Cycle                   = flash, hide, click, score, level + 1
Lives                   = 3 (one per wrong click; correct order does not matter)

• Level 1: 3 lit on 3x3, 1800 ms flash. Trivial encoding load.
• Level 5: 7 lit on 4x4, 1400 ms flash. First real challenge for new players.
• Level 10: 12 lit on 5x5, 1000 ms flash. Above Miller's seven; requires chunking.
• Level 15: 17 lit on 6x6, 800 ms flash floor. Elite range.

Worked example: chunking a 5x5 pattern

At level 10 the screen shows 12 lit squares on a 5x5 grid for one second. Trying to memorise each square independently overflows working memory. Chunking strategy:

1. Scan corners first (250 ms). Note any of the four corners lit; record as a single chunk.
2. Scan central cross (300 ms). The middle row and column carry strong perceptual weight; encode lit cells as a sub-pattern.
3. Scan diagonals (250 ms). Lit diagonal pairs collapse into "NW-SE" or "NE-SW" tags.
4. Encode remainder (200 ms). Whatever survives turns into a memorised L-shape, T-shape, or named cluster.
5. Click in chunk order, fastest first to free up rehearsal capacity.

Percentile reference table

Based on aggregated scores from public visual-memory leaderboards (Human Benchmark, MyBrainTest), here is roughly where each level sits in the adult population.

Common pitfalls

• Sequential clicking habit. The game does NOT require recall in shown order. Click the fastest-to-find cells first to free up rehearsal capacity for the harder ones.
• Eye-darting during the flash. Saccades cost roughly 200 to 300 ms of usable encoding time each. Fix gaze on the centre of the grid and use peripheral vision to take in the whole pattern at once.
• Verbal labelling. Naming each lit square ("top-left, middle, bottom-right...") routes through the phonological loop, which is the wrong system. Stay visual: keep the after-image in the visuospatial sketchpad.
• Skipping warm-up levels. Brain takes 30 to 60 seconds to engage the right encoding network. Restarting after a game-over is often a higher-scoring attempt than the first run.
• Trying again immediately after a fail. The interference from the wrong recall pollutes the next trial. A 10 to 15 second gap clears working memory.
• Comparing scores across days. Sleep, caffeine and stress move scores by 1 to 2 levels easily. Track a 5-attempt median, not a single best.

Related tools on 3Tej

Frequently asked questions