Working Memory
Memory Updating Task
Also called: numerical memory updating, memory updating numerical, numerical updating task, memory updating paradigm
Measure working memory updating with numbers: boxes start with digits, changes such as +3 appear in single boxes, and participants enter each box's final value.
7-9 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 30
- Evidence grade
- C
- Registered clinical studies
- 1
What it measures
Measures how well people update working memory. Boxes each start with a digit, then changes such as +3 or −2 appear in single boxes, one at a time. Participants keep each box's value in mind and enter the final values at the end.
Constructs
- Working memory updating
- Working memory capacity
- Mental arithmetic
Research uses
- Working memory research in aging and clinical groups
- Latent-variable studies of working memory capacity and reasoning
- Within-study comparison of memory load and number of changes
How the paradigm works
The Memory Updating Task measures working memory updating with numbers: holding a value in each of several places and replacing a value whenever it changes. Three to five boxes appear on the screen. A start digit appears in each box, one box at a time, for 1,000 ms. Then changes such as +3 or −2 appear in single boxes, one at a time, for 1,300 ms each, and the participant works out the box's new value and keeps it in place of the old one. After two to six changes, a question mark appears in one box at a time, in a random order, and the participant types the value that box holds.
More about the paradigm
The task follows the memory updating task of Salthouse, Babcock and Shaw (1991), as Oberauer and colleagues (2000) adapted it and Lewandowsky, Oberauer, Yang and Ecker (2010) gave it in their working memory battery: start digits for 1 s each, changes of 1 to 7 up or down for 1.3 s each with a 250 ms blank after each, values that stay between 1 and 9, each change in a different box from the one before, answers asked for box by box in a random order, and 15 rounds that cross 3, 4 and 5 boxes with 2 to 6 changes, after two practice rounds (2 boxes with 4 changes, then 3 boxes with 3). The time each change shows is a setting from 1,000 to 3,000 ms, and the most boxes a setting of 4, 5 or 6 (10, 15 or 20 rounds); both are fixed for the session, with no staircase.
Lewandowsky and colleagues gave every participant the same rounds; here each session draws its own from its seed, by the same rules. The participant starts each round, as in the source, and can take back a digit before submitting. Practice rounds give feedback; test rounds do not. In their three samples of university students (110 and 114 in Australia, 160 in Taiwan) the mean share of boxes right was .54, .61 and .72, with internal consistency of .85 to .91, and the task loaded on one working memory factor with the complex span tasks.
The main score is the share of a round's boxes entered right, averaged over rounds, as Lewandowsky and colleagues scored it. The share at each number of boxes, the share on boxes that changed and on boxes that kept their start digit, and the share of answers that were a value the changes had replaced are also given. Scores support research and do not diagnose.
Scores
- Boxes Right%
- Share of a round's boxes entered right, averaged over rounds.
- Higher is better
- Three Boxes%
- The same share on rounds with 3 boxes.
- Higher is better
- Four Boxes%
- The same share on rounds with 4 boxes.
- Higher is better
- Five Boxes%
- The same share on rounds with 5 boxes. Null when the most boxes is set to 4.
- Higher is better
- Six Boxes%
- The same share on rounds with 6 boxes. Null unless the most boxes is set to 6.
- Higher is better
- Changed Boxes%
- Share of boxes that changed at least once that were entered right.
- Higher is better
- Unchanged Boxes%
- Share of boxes that kept their start digit that were entered right.
- Higher is better
- Outdated Values%
- Share of answers on changed boxes that were a value the box held before its last change.
- Lower is better
- Rounds All Correct%
- Share of rounds with every box right.
- Higher is better
- Median Answer Timems
- Median time from the answer screen to Submit.
- Informational
- Rounds Completedtrials
- Test rounds recorded in the scored run.
- Under 90% of the planned rounds is invalid
- Planned Roundstrials
- Test rounds the session planned to give.
- Informational
Use and evidence
30 papers from 2021 to 2025 name this task in their title or abstract, down 19% compared with 2016 to 2020. That ranks it 302 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 7 |
| 2024 | 7 |
| 2023 | 6 |
| 2022 | 6 |
| 2021 | 4 |
| 2020 | 8 |
| 2019 | 13 |
| 2018 | 9 |
| 2017 | 4 |
| 2016 | 3 |
| 2015 | 4 |
| 2014 | 6 |
| 2013 | 3 |
| 2012 | 2 |
| 2011 | 3 |
| 2010 | 1 |
| 2009 | 0 |
| 2008 | 4 |
| 2007 | 2 |
| 2006 | 4 |
| 2005 | 2 |
| 2004 | 0 |
| 2003 | 1 |
| 2002 | 0 |
| 2001 | 3 |
| 2000 | 1 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 2
- Papers on reliability or validity
- 9
- Registered clinical studies using it as an outcome
- 1
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Boxes
- 3, 4 or 5 in one round (up to 4 or 6 as a setting), three to a row.
- Start digits
- 1 to 9, one box at a time, each for 1,000 ms, after the empty boxes for 1,000 ms.
- Changes
- +1 to +7 or −1 to −7 in one box at a time, each for 1,300 ms (standard; 1,000 to 3,000 ms as a setting, fixed for the session), every value staying between 1 and 9, each in a different box from the one before.
- Blanks
- 250 ms of empty boxes after each start digit and each change.
- Rounds
- Every number of boxes with 2, 3, 4, 5 and 6 changes, once each, in a seeded order: 15 rounds at the standard settings. Each round starts when the participant is ready.
- Answer
- A question mark in one box at a time, in a random order; the participant types the box's value, or taps it on a keypad of 1 to 9 on touch screens. Untimed; a digit can be taken back.
- Practice
- 2 boxes with 4 changes, then 3 boxes with 3, each followed by the right values, the values entered and the boxes right.
- Score
- The share of a round's boxes right, averaged over rounds, overall and at each number of boxes; changed and unchanged boxes and outdated values as well.
Lewandowsky and colleagues gave every participant the same rounds; here each session draws its own by the same rules. The time per change and the most boxes are set before the session and never change. No published norms are applied.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: test-phase answers of the scored run, one for each box on each round, a digit from 1 to 9 (chance 1 in 9). Accuracy significantly below chance invalidates; accuracy not above chance is flagged. Fewer than 90% of the planned rounds invalidates. A hidden tab withholds the percentile.
Why these rules
Keeping a box's start digit, the error this test measures, still leaves the boxes that did not change right, so a participant who stops updating stays above chance. Accuracy significantly below chance (one-sided exact test, p < .05) cannot come from guessing; it means the response mapping was reversed or the task was not followed (Slick, Sherman and Iverson 1999 treat below-chance forced-choice performance as definite response bias). Chance-level accuracy can be a real result in impaired participants, so it is only flagged. Answers are untimed and every box needs a digit, so nothing can be missed and no answer can come too fast. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Time each change shows (ms) | 1,300 | 1,000 to 3,000 |
| Most boxes | 5 | 4 to 6 |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Memory Updating Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Salthouse et al., 1991; Oberauer et al., 2000; Lewandowsky et al., 2010), with the standard settings (time each change shows (ms): 1,300; most boxes: 5). The main score was boxes right (share of a round's boxes entered right, averaged over rounds).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Salthouse, T. A., Babcock, R. L., & Shaw, R. J. (1991). Effects of adult age on structural and operational capacities in working memory. Psychology and Aging, 6(1), 118-127. https://doi.org/10.1037/0882-7974.6.1.118The method: reporting the latest value of variables changed by arithmetic, as the number of variables and operations grows.
- Oberauer, K., Süß, H.-M., Schulze, R., Wilhelm, O., & Wittmann, W. W. (2000). Working memory capacity: Facets of a cognitive ability construct. Personality and Individual Differences, 29(6), 1017-1045. https://doi.org/10.1016/S0191-8869(99)00251-2The psychometric version of the task, memory updating numerical.
- Lewandowsky, S., Oberauer, K., Yang, L.-X., & Ecker, U. K. H. (2010). A working memory test battery for MATLAB. Behavior Research Methods, 42(2), 571-585. https://doi.org/10.3758/BRM.42.2.571The procedure: 3 to 5 boxes, start digits for 1 s, changes of up to 7 for 1.3 s with 250 ms blanks, 15 rounds and partial-credit scoring.
Source papers
- Lewandowsky S, Oberauer K, Yang LX, Ecker UKH (2010). A working memory test battery for MATLAB. Behavior Research Methods, 42(2):571-585.Paradigm source, University students
- Oberauer K, Süß HM, Schulze R, Wilhelm O, Wittmann WW (2000). Working memory capacity: facets of a cognitive ability construct. Personality and Individual Differences, 29(6):1017-1045.Paradigm source, Adults
Questions
How long does the Memory Updating Task take?
About 7 to 9 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Boxes Right, Three Boxes, Four Boxes, Five Boxes, Six Boxes, Changed Boxes, Unchanged Boxes, Outdated Values, Rounds All Correct, Median Answer Time, Rounds Completed and Planned Rounds. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Memory Updating Task used in research?
30 papers from 2021 to 2025 name it in their title or abstract, down 19% compared with 2016 to 2020. That ranks it 302 of 395 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Memory Updating Task have?
In the published research, 2 papers report its test-retest reliability and 1 registered clinical study uses it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Time each change shows (ms) (1,000 to 3,000) and most boxes (4 to 6). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.
Related tests