Working Memory
Delayed Estimation Task
Also called: continuous report task, color wheel task, delayed color estimation, color recall task
Measure how precisely colors are held in visual working memory: participants click a remembered color on a color wheel, with 1 to 6 squares to hold.
8-10 minTimed: desktop or laptop with a mouse or trackpad
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
- 80
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how precisely colors are held in visual working memory. One, two, four or six colored squares appear for 100 ms; after a 900 ms blank, one place is marked and participants click its color on a color wheel. The main score is the mean error in degrees around the wheel, with the guess rate and precision of a mixture model at each number of squares.
Constructs
- Visual working memory
- Precision of memory
- Working-memory capacity
Research uses
- Working memory research
- Aging and development studies
- Within-study group comparison
How the paradigm works
The delayed estimation task measures how precisely colors are held in visual working memory, a method also called continuous report. One, two, four or six colored squares appear for 100 ms at places on a circle around a cross. After a 900 ms blank, a color wheel appears with an outline at each square's place, one of them thick, and participants click the color that the square in the thick outline had, with no time limit. The wheel holds 360 colors of equal lightness and is turned at random on every trial, so a color's place on the screen gives no clue.
More about the paradigm
Wilken and Ma (2004) introduced reports of a remembered color on a continuous wheel, and Zhang and Luck (2008) used this design to separate how many items are held from how precisely each one is held. The standard session is 25 trials with each number of squares, 100 in all, intermixed in a seeded order, after 6 practice trials that show the right color. Colors in one array are at least 30 degrees apart on the wheel. Each display is timed to screen frames, and its shown length is saved with every trial.
The error of an answer is the shortest distance around the wheel from the color chosen to the right one, from 0 to 180 degrees; a guess averages 90. The main score is the mean error over all trials. For each number of squares, the mean error and Zhang and Luck's mixture model are reported: a von Mises distribution centered on the right color for answers from memory and an even spread for guesses, fitted by maximum likelihood, which gives the guess rate and the standard deviation of the remembered answers (precision). Model estimates from 25 trials are noisy, so the mean error is the main score. Presses under 100 ms after the wheel appears are counted as anticipations. Scores support research and do not diagnose.
Scores
- Mean Errordegrees
- Mean shortest distance around the wheel from the color chosen to the right one, over all test trials.
- Lower is better; a guess averages 90
- Mean Error, 1 Squaredegrees
- Mean error on trials with one square.
- Lower is better
- Mean Error, 2 Squaresdegrees
- Mean error on trials with two squares.
- Lower is better
- Mean Error, 4 Squaresdegrees
- Mean error on trials with four squares.
- Lower is better
- Mean Error, 6 Squaresdegrees
- Mean error on trials with six squares.
- Lower is better
- Guess Rate, 1 Square%
- Mixture model share of answers that were guesses, on trials with one square. Null with fewer than 10 answers.
- Lower is better
- Guess Rate, 2 Squares%
- Mixture model share of answers that were guesses, on trials with two squares.
- Lower is better
- Guess Rate, 4 Squares%
- Mixture model share of answers that were guesses, on trials with four squares.
- Lower is better
- Guess Rate, 6 Squares%
- Mixture model share of answers that were guesses, on trials with six squares.
- Lower is better
- Memory SD, 1 Squaredegrees
- Mixture model standard deviation of the remembered answers (precision), on trials with one square. Null with fewer than 10 answers; the fit reports up to about 73 degrees.
- Lower is more precise
- Memory SD, 2 Squaresdegrees
- Mixture model standard deviation of the remembered answers, on trials with two squares.
- Lower is more precise
- Memory SD, 4 Squaresdegrees
- Mixture model standard deviation of the remembered answers, on trials with four squares.
- Lower is more precise
- Memory SD, 6 Squaresdegrees
- Mixture model standard deviation of the remembered answers, on trials with six squares.
- Lower is more precise
- Median Response Timems
- Median time from the wheel's onset to the answer.
- Informational; there is no time limit
- Anticipationscount
- Presses before the wheel appeared or under 100 ms after it. They are not taken as answers.
- Lower is better
- Frame Intervalms
- Median time between screen frames on the participant's device.
- Informational
- Displays on Time%
- Share of answered trials whose squares and delay ran within one frame of plan.
- Below 0.9 is flagged for review
- Planned Trialstrials
- Test trials the session planned to show.
- Informational
- Total Trialstrials
- Trials answered in the scored run.
- Fewer than planned is flagged for review
Use and evidence
80 papers from 2021 to 2025 name this task in their title or abstract, up 31% compared with 2016 to 2020. That ranks it 205 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 19 |
| 2024 | 16 |
| 2023 | 14 |
| 2022 | 19 |
| 2021 | 12 |
| 2020 | 16 |
| 2019 | 14 |
| 2018 | 10 |
| 2017 | 10 |
| 2016 | 11 |
| 2015 | 7 |
| 2014 | 3 |
| 2013 | 1 |
| 2012 | 2 |
| 2011 | 1 |
| 2010 | 1 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 6
- Registered clinical studies using it as an outcome
- 0
Note on the counts: includes general uses of the term.
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 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Display
- 1, 2, 4 or 6 colored squares 48 pixels wide at places on a circle 108 pixels around a cross, on dark gray. The wheel runs from 170 to 223 pixels from the center.
- Colors
- 360 colors of equal lightness on a circle in CIE L*a*b* (L* 55, center a* 26 and b* 6, radius 54), inside the colors an ordinary screen shows. Colors in one array are at least 30 degrees apart on the wheel.
- Timing
- Cross 500 ms, squares 100 ms, blank 900 ms, then the wheel until an answer, with no time limit. Displays switch on screen frames, and their shown length is saved.
- Answer
- A click on the wheel at the color that the square in the thick outline had. The wheel turns at random on every trial. Clicks before the wheel or within 100 ms of it are counted and not taken.
- Length
- 25 trials with each number of squares (20 to 50), 100 in all, intermixed in a seeded order, with a rest after every 50.
- Practice
- 6 trials, with 1, 1, 2, 2, 4 and 6 squares, each showing the right color for 1,500 ms.
- Scores
- The mean error over all trials, then at each number of squares the mean error and Zhang and Luck's mixture model fitted by maximum likelihood: the guess rate and the standard deviation of the remembered answers.
Zhang and Luck (2008) used 180 colors at L* 70 on a calibrated monitor, 1, 2, 3 and 6 squares, and 150 trials per condition. This version keeps their timing and layout, uses a color circle that ordinary screens can show, 1, 2, 4 and 6 squares, and fewer trials, so the model estimates are noisier and the mean error is the main score.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: delayed_estimation_test from the scored run, one answer per trial on a color wheel with no time limit. An answer less than 90 degrees from the right color counts as near; a guess is near on 179 of 360 colors (chance 49.7%). Fewer near answers than chance invalidates and not more than chance is flagged. Presses in one eighth of the wheel's circle on 90% of answers invalidate when near answers are not above chance; the wheel turns at random on every trial, so the right color is in each eighth equally often. Answers under 200 ms on at least half the trials invalidate. A run short of its planned trials, or under 90% of trials with the squares and the delay within one frame of plan, is flagged and its percentile withheld.
Why these rules
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. One option on at least 90% of answered trials, when the correct options are balanced and accuracy is not above chance, means the answers did not depend on the items (the task counterpart of long-string responding in surveys; Curran 2016). An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast. The squares show for 100 ms and the delay lasts 900 ms, so trials whose displays ran late change what was measured.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Trials per set size | 25 | 20 to 50 |
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
- Timed test: participants need a desktop or laptop with a mouse or trackpad. 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 Delayed Estimation Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Wilken and Ma, 2004; Zhang and Luck, 2008), with the standard settings (trials per set size: 25). The main score was mean error (mean shortest distance around the wheel from the color chosen to the right one, over all test trials).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Zhang, W., & Luck, S. J. (2008). Discrete fixed-resolution representations in visual working memory. Nature, 453(7192), 233-235. https://doi.org/10.1038/nature06860
- Wilken, P., & Ma, W. J. (2004). A detection theory account of change detection. Journal of Vision, 4(12), 11. https://doi.org/10.1167/4.12.11Reports of a remembered color, orientation or spatial frequency on a continuous scale.
Source papers
- Zhang W, Luck SJ (2008). Discrete fixed-resolution representations in visual working memory. Nature, 453(7192):233-235.Paradigm source, Adults
- Wilken P, Ma WJ (2004). A detection theory account of change detection. Journal of Vision, 4(12):11.Paradigm source, Adults
Questions
How long does the Delayed Estimation Task take?
About 8 to 10 minutes, as listed in the test library.
What device does a participant need?
Timed test: participants need a desktop or laptop with a mouse or trackpad. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Mean Error, Mean Error, 1 Square, Mean Error, 2 Squares, Mean Error, 4 Squares, Mean Error, 6 Squares, Guess Rate, 1 Square, Guess Rate, 2 Squares, Guess Rate, 4 Squares, Guess Rate, 6 Squares, Memory SD, 1 Square, Memory SD, 2 Squares, Memory SD, 4 Squares, Memory SD, 6 Squares, Median Response Time, Anticipations, Frame Interval, Displays on Time, Planned Trials and Total Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Delayed Estimation Task used in research?
80 papers from 2021 to 2025 name it in their title or abstract, up 31% compared with 2016 to 2020. That ranks it 205 of 395 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Delayed Estimation Task have?
In the published research, 0 papers report its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade D in the ConductCognition test census).
Which settings can a researcher change?
Trials per set size (20 to 50). 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.