What a span test asks
How long a list can you hold in mind and repeat straight away? A span test finds out by making the lists longer until you start to fail. Repeating a list backwards is harder: you have to keep the items and reverse their order before you answer. This page runs one procedure with three materials. Digits and letters are shown one at a time. For the blocks, nine squares in fixed places light up one after another, a screen version of the block-tapping task described by P. M. Corsi in 1972, traditionally nine cubical blocks on a board that an examiner taps.
How a run works
Digits and letters
The procedure is the one Woods and colleagues used in 2011. A run has 14 lists. Forwards the first list has three items, backwards two. After a correct list the next one is one item longer; after two wrong lists in a row at the same length it is one item shorter. Woods called this a 1:2 staircase. It keeps most of the 14 lists near the length where you start to make mistakes.
Items come at one per second: each is on the screen for 0.7 s, then the box is empty for 0.3 s. In the study the digits were spoken through headphones at the same rate, so the 0.7 / 0.3 split is this page’s choice. One second after the last item the empty answer slots appear, the same delay as the cue in the study. Digit lists use the digits 1 to 9: up to nine digits no digit repeats, and longer lists reuse digits, as in the study. Neighbours such as 4 and 5, or 3 and 5, never stand next to each other. Letter lists use twelve consonants, B F H J K L M N P R S T, without repeats and without two letters that follow each other in the alphabet, such as J and K. Digits forwards can also run with an extra 0.6 s pause after every third digit (also this page’s choice), to see what grouping does.
Blocks
The block version follows eCorsi, the tablet task of Brunetti, Del Gatto and Delogu (2014). Each block stays lit for 0.5 s and a new one lights up every second. The first sequences are two blocks long, and every length gets two sequences. If at least one of the two is right, the next length is one block longer; when both are wrong, or after nine blocks, the run ends. The layout of the nine places is this page’s own, not the board of the study. When the last block lights up, numbers appear on the blocks so you can answer with the keyboard too.
Practice
The first run of each task starts with practice that tells you after every list whether you were right: two lists of three items for digits and letters (the study gives no number), three sequences of three blocks for the blocks (as in eCorsi). Practice never counts, and later you can switch it off in Settings.
How the score is computed
For digits and letters the main score is the mean span (MS) of Woods and colleagues: an estimate of the list length at which you would get half of the lists right. It starts half an item below the shortest length you were given and adds, for every length, the share of lists you got right there:
The worked example in the study (its Table 1, a forward run from three digits) had every list of 3 to 6 digits right, 1 of 4 lists of 7, 1 of 3 lists of 8 and none of the 2 lists of 9: 2.5 + 1 + 1 + 1 + 1 + 0.25 + 0.33 + 0 = 7.08. The page also shows two other measures from the study. The longest list (ML) is the longest list you got right in any of the 14. The two-error span (TE-ML) is the longest list you got right before your first two misses in a row at one length, the number a test that stops there would report. In the example ML is 8 and TE-ML is 6. For the blocks the page gives the eCorsi span, the last length at which at least one of the two sequences was right, and the number of correct sequences.
One run is a small sample. In a simulation of 2,000 runs of 14 forward lists (29 October 2026), a simulated person whose true 50% point was 6.5 digits got an MS between 5.8 and 7.7 in 95% of runs, and two of its runs differed by up to 1.3 digits in 95% of pairs. A difference of a whole digit between two of your runs can therefore still be chance.
Published means
The chart marks these means with their standard deviations. They describe particular groups tested in a particular way: landmarks, not a score to reach, and the page gives no percentile. Letters and grouped digits have no published values here and are compared only with your own runs.
| Task | Study and people | How they were tested | Mean ± SD |
|---|---|---|---|
| Digits forward | Woods et al. 2011, Experiment 1: 30 adults aged 18–46 | Spoken digits, 14 lists from 3, averaged over three sessions | MS 7.41 ± 0.94 |
| Digits backward | the same | 14 lists from 2 | MS 6.00 ± 1.32 |
| Digits forward | Woods et al. 2011, Experiment 2: 763 adults aged 18–65 | Spoken digits, 10 lists from 5 | MS 6.52 ± 1.00 |
| Digits backward | the same | 10 lists from 4 | MS 4.91 ± 1.06 |
| Blocks forward | Brunetti et al. 2014: 73 young adults, mean age 21.6 | Tablet, their own board | Span 6.11 ± 0.80 |
| Blocks backward | the same | the same | Span 5.29 ± 1.20 |
| Blocks forward | Brunetti et al. 2014: 34 adults over 50 | the same | Span 4.76 ± 1.19 |
| Blocks backward | the same | the same | Span 4.35 ± 1.30 |
In the digit study you would have heard the digits and said them aloud while an examiner typed; here you read them and type them yourself. Woods and colleagues point out that spans differ between heard and seen digits, most of all backwards, and that answering on a keypad brings its own demands. The block board and the screen differ from the tablet of eCorsi, so the block means are only roughly comparable too.
What to look for
- In both studies the backward span was lower than the forward one. In the larger digit study the backward span also fell with age while the forward span did not, which fits the authors’ view that the two directions measure partly different mental operations.
- Cowan (2010) argues that a central store of working memory holds about 3 to 5 meaningful items, or chunks, in young adults, and that people repeat longer lists when they can use strategies such as grouping or silent rehearsal. If your grouped run is longer, the pauses may have helped you hold three digits as one chunk.
- Woods and colleagues found that spans rose slightly but significantly over repeated sessions, which they put down to learning the procedure. Compare runs from different days, and expect your first runs to be a little lower.
Things to try
- Run digits forwards and then backwards, and compare the gap with the gap between the two block directions.
- On the same day, run digits forwards once without and once with grouping, then compare them in the chart.
- Compare letters with digits forwards: the letter lists draw on twelve consonants, the digit lists on nine digits.
- Say a word such as “the” aloud over and over while the digits appear. That blocks silent rehearsal, one of the conditions under which Cowan describes the smaller central limit.
Limits
- A typo counts as a wrong list, and so does a block tapped by mistake; use Backspace before you confirm.
- Tiredness, noise and interruptions lower the score. A pause or a hidden tab drops the list in progress, and a new list of the same length replaces it.
- On a small phone the blocks are about the size of a fingertip, so mistaken taps are more likely than on a tablet or a computer.
- Each run is saved with your input and the refresh rate of your screen, and by default it is compared only with runs on the same input and a similar refresh rate (within 10%, or both below 50 Hz).
- This is not a clinical memory assessment and cannot show memory problems or dementia.
Controls
- Keyboard: 1–9 for digits and for the numbered blocks, the letter keys for letters, Backspace removes the last item, Enter confirms a list, Esc pauses.
- Touch and mouse: the keypad under the test area (then OK) and the blocks themselves.
- Buttons: Start, Pause, Resume, End run (stops without saving), Run again, and Continue after each practice list.
- Settings and History: settings are locked during a run and remembered; Clear history asks before it deletes.
Frequently asked questions
Why does the test go on after I miss two lists?
Classic tests stop when both lists of one length are missed, and Woods and colleagues showed that this can underestimate the span: some people miss two lists early and later get longer ones right. The mean span uses all 14 lists, and in their first experiment its test-retest correlation for digits forward was 0.67, higher than that of the two-error span.
Sources
- Woods DL, Kishiyama MM, Yund EW, Herron TJ, Edwards B, Poliva O, Hink RF, Reed B (2011). Improving digit span assessment of short-term verbal memory. Journal of Clinical and Experimental Neuropsychology 33(1), 101–111. The staircase, digit rules, mean span, means and learning effect.
- Brunetti R, Del Gatto C, Delogu F (2014). eCorsi: implementation and testing of the Corsi block-tapping task for digital tablets. Frontiers in Psychology 5, 939. Timing, span rule, practice and the means by age group.
- Cowan N (2010). The magical mystery four: how is working memory capacity limited, and why? Current Directions in Psychological Science 19(1), 51–57. The central limit of 3 to 5 chunks, grouping and rehearsal.