By Eric, BlinkBench Founder · Last reviewed: October 2026
When you look at a half-lit grid for three seconds, are you remembering a picture or reciting a description? That question decides what your Visual Memory Test score is a score of, and the test cannot answer it for you.
What the test asks of you
The Visual Memory Test shows a matrix with exactly half its cells filled. The pattern stays up for 3000 ms and then clears, and you mark the filled cells on an empty grid in any order. The ladder starts at 2 filled cells in a 2×2 matrix. Each pattern you reproduce exactly adds one filled cell, and two misses at the same size end the test. Your score is the number of filled cells in the largest pattern you got exactly right, and the ladder stops at 15 filled cells in a 5×6 matrix.
The design comes from the Visual Patterns Test. Its authors describe it as "designed to measure short-term visual memory largely shorn of its spatio-sequential component" (Della Sala et al., 1999). That is why the order of your marks does not matter here, and why this test sits apart from the sequence memory test, which asks for blocks back in order.
"Largely shorn" is careful wording, and it is the right amount of care. Stripping out the sequence does not strip out everything that is not a picture.
Patterns can be put into words
A small pattern can be turned into a sentence: "top row full, then one down the left." Once you do that, you are no longer holding a picture. You are holding a short phrase, and the task has quietly become a verbal one.
This is a known problem, not a speculation about visitors. A 2006 study set out "to assess the availability of verbal coding" and its effect on performance in the Visual Patterns Test (Brown, Forbes & McConnell, 2006). In a pilot, participants were shown the patterns and asked to name the shapes within them. The authors report that "Availability of verbal codes was low overall; however, some patterns resulted in a higher mean number of labels than others."
They then built a modified version from the patterns that produced the fewest labels. In an experiment with 60 participants, the modified version "resulted in a lower visual working-memory span" than a version in which verbal coding was more available. Their conclusion: "verbal coding does influence visual matrix task performance."
Read that carefully. Two sets of patterns, both from the same test, gave different spans, and the difference tracked how easy the patterns were to name.
What this means for a random grid
The published test uses a fixed set of patterns. This site draws a new pattern at random every round, and it does not screen for describability. A random layout can land on a tidy shape that is easy to say aloud, or on a scatter that resists any description. The same person can therefore meet an easier or harder round for reasons that have nothing to do with their memory.
The grid's shape also changes with size. Seven, eleven and thirteen filled cells need 14, 22 and 26 cells, which only divide as 7×2, 11×2 and 13×2, so those levels appear as a strip. The site shows the strip rather than break the half-filled rule, and the test page flags it as a seam in the instrument.
What you can and cannot take from your score
Your score is the largest pattern you reproduced exactly, and that is all it is. It does not say whether you reached it by seeing or by saying. The test records only the cells you marked. It cannot see a finger tracing the screen, a sketch on paper or a half-whispered description, and it says so on the page.
The guess odds, by contrast, are exact. The answer is always "choose half the cells", so the number of possible answers at k filled cells is C(2k, k). Here is the ladder, with the matrix shape each level uses:
| Filled cells | Matrix | Possible answers (guess chance is 1 in this) |
|---|---|---|
| 2 | 2×2 | 6 |
| 4 | 4×2 | 70 |
| 6 | 4×3 | 924 |
| 7 | 7×2 | 3,432 |
| 8 | 4×4 | 12,870 |
| 10 | 5×4 | 184,756 |
| 15 | 6×5 | 155,117,520 |
Describing a pattern is not guessing, so these figures say nothing about it. They tell you how much of a score could be luck. They cannot tell you which kind of memory produced it.
Why there is no percentile to settle it
Normative data for the Visual Patterns Test exists, but it was published in a printed manual that this site has not read, so it quotes no range. Even a verified range would describe fixed printed patterns shown in person, not random ones on your screen. And a figure computed from this site's own visitors would be self-selected and unvetted, so none is shown.
So treat your result as a number from one pass through one random sequence of patterns, on your screen at your distance. If you retake the test, expect different patterns, and expect some of the difference between runs to come from how describable they happened to be. This test is not a cognitive assessment, and a result here reflects your equipment and screen size as well as you.
Sources
- Della Sala, S., Gray, C., Baddeley, A., Allamano, N., & Wilson, L. (1999). Pattern span: a tool for unwelding visuo-spatial memory. Neuropsychologia, 37(10), 1189-1199.
- Brown, L. A., Forbes, D., & McConnell, J. (2006). Limiting the use of verbal coding in the Visual Patterns Test. The Quarterly Journal of Experimental Psychology, 59(7), 1169-1176.