By Eric, BlinkBench Founder · Last reviewed: August 2026
How many times should you run the sequence memory test before trusting what you see? The answer depends on what's making your score move — and not all of it is you.
What one trial actually tells you
The sequence memory test works by showing you a lengthening chain of lit blocks and asking you to tap them back in order. Your score is the longest run you get right before making a mistake. That structure means a single error at a long length costs you far more than a single error at a short one. A momentary lapse at length nine drops you to eight; the same lapse at length five drops you to four. The score is sensitive to where the error lands, not just whether one occurs.
That sensitivity is a feature when you're trying to measure something real, but it also means a single trial carries a lot of noise. Your attention drifts for half a second, a notification appears at the edge of your screen, your tap registers a fraction late on a touchscreen — any of these can cut a trial short in a way that has nothing to do with your visuospatial memory. One number from one attempt is a thin basis for any conclusion.
The underlying task is a computerized version of the Corsi Block-Tapping Task, which has a published normative mean of 4.97 blocks (SD 1.03) in a sample of 340 healthy adults aged 21 to 89. But that reference comes from the physical version of the task — an examiner touching wooden cubes on a board — and research has found that moving the task to a screen changes what it measures. Forward reproduction was more accurate on the standard physical task than on a digital version; the authors concluded that computerization "has serious consequences for the cognitive concepts that the Corsi Task is assumed to assess." That's why this site doesn't place your score against that normative range. The comparison would be misleading, not informative.
What the test can do honestly is give you a stable estimate of your own performance on this particular computerized version — and for that, more than one trial helps.
Why a second trial is worth taking
Because the score is a whole-number count of correctly reproduced lengths, the gap between a good trial and a bad one can easily be two or three blocks. If your true range sits somewhere around six to eight, a single trial might land anywhere in that window depending on when your attention peaked and where the sequence happened to demand it. A second trial, taken after a short rest, gives you a second independent draw from that range. If the two are close, you have reasonable confidence in the region. If they're far apart, you know the first number was less informative than it looked.
This is the same logic that applies to other single-trial measures. The guide on within-person reaction time variability makes the point clearly for a different test: a single person's score can swing substantially across five attempts, and that swing often exceeds the gap between two different people. The sequence memory test has the same property. One trial tells you something. Two trials, spaced by a few minutes or more, tell you considerably more.
Why five trials in a row tells you less
Here's where the retake logic reverses. The sequence memory task asks you to hold and reproduce a specific pattern of spatial locations. When you run the task repeatedly in a short window, your brain doesn't just reset between trials — it starts building a representation of how the sequences tend to behave. The grid has a fixed number of blocks. The sequences follow a consistent timing. After several runs, you may begin anticipating structure rather than encoding each sequence fresh.
This matters because what you're trying to measure is your capacity to hold a novel visuospatial sequence in working memory. If you've seen the task enough times that it no longer feels novel, the score starts reflecting familiarity with the format as much as it reflects the underlying memory capacity. The number may go up. But it's measuring something different from what it measured on trial one.
There's also a fatigue effect running in the opposite direction. Sustained attention over multiple sequential trials is effortful. By trial four or five in a single session, your performance may drop not because your visuospatial memory has changed but because you're less alert than you were. Either way — familiarity inflating the score or fatigue deflating it — the later trials in a long session are harder to interpret.
The practical range: two or three, spaced by days
Two trials in a single session, with a few minutes between them, is a reasonable starting point. You get a second independent estimate without giving the task enough repetition to feel routine. If the two scores are within one block of each other, that region is probably a fair description of where you are today on this device and in this context.
If you want a more stable picture across time — one that smooths out day-to-day variation in alertness and attention — two or three trials spread across different days is more informative than five trials on a single afternoon. Memory performance is not fixed. It varies with sleep, with how long you've been awake, with how much you've had to concentrate on other things before sitting down. Sampling across days captures that variation honestly rather than averaging it away inside a single session.
What you're building, across a small number of spaced trials, is a range rather than a point. Your sequence memory score on this test is probably somewhere between X and Y, under these conditions, on this device. That's a more accurate statement than any single number, and it's the kind of statement two or three honest trials can support.
Frequently asked questions
Does a higher score on a later trial mean my memory got better?
Not necessarily. A higher score on a second or third trial could reflect reduced anxiety with the format, a moment of better attention, or the absence of whatever disrupted the earlier attempt. It could also reflect growing familiarity with the task structure, which changes what the score measures. Treat a higher later score as additional information about your range, not as evidence of improvement.
Why doesn't the test show me a percentile?
The published normative data for the Corsi task comes from the physical board version, administered by an examiner under standardized conditions. Research has found that the computerized version produces different results — forward reproduction was more accurate on the physical task than on a digital equivalent. Placing your score against norms collected on a different apparatus would give a misleading comparison, so the site declines to show one.
Does the device I use affect my score?
Yes. A touchscreen introduces different timing and spatial precision demands than a mouse on a desktop. Tap registration latency varies across devices. Your score reflects your visuospatial memory and your device's input characteristics together — the test cannot separate them. If you retake on a different device, treat the results as a separate series rather than a continuation of the first.
How long should I wait between trials in the same session?
A few minutes is enough to let the specific sequence from the previous trial fade. The goal is for each trial to feel like a fresh encoding task rather than a continuation of the last one. There's no precise interval that guarantees this, but two to five minutes of doing something else is a reasonable guide.
Is there a point at which more trials stop adding information?
Yes. Once you have two or three trials that cluster in a similar region, additional trials in the same session are more likely to introduce familiarity or fatigue effects than to narrow your estimate further. At that point, a trial on a different day — when your alertness and attention are in a different state — adds more than a fourth trial taken immediately.
Sources
- Facchin, A., Pegoraro, S., Rigoli, M., et al. (2024). Regression-based normative data for Corsi Span and Supraspan learning and recall among Italian adults. Neurological Sciences, 45(12), 5707-5718.
- Claessen, M. H. G., van der Ham, I. J. M., & van Zandvoort, M. J. E. (2015). Computerization of the standard Corsi Block-Tapping Task affects its underlying cognitive concepts: a pilot study. Applied Neuropsychology: Adult, 22(3), 180-188.