Processing Speed
Dot Arithmetic Task
Also called: approximate arithmetic task, nonsymbolic arithmetic task, non-symbolic addition task, approximate addition task, nonsymbolic addition, dot addition task
Measure approximate arithmetic: how well people add two brief clouds of dots in their head and judge the total against a third cloud, without counting.
7-9 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. Try it here first, with no account.
- Papers, 2021 to 2025
- 32
- Evidence grade
- D
- Registered clinical studies
- 1
What it measures
A measure of approximate arithmetic: adding numbers of things without counting. Two clouds of dots go into a box one after the other, then a new cloud flashes beside it, each for 750 ms, and participants press the key for the box or the cloud, whichever had more. The main score is accuracy, reported at four fixed ratios between the total and the new cloud.
Constructs
- Approximate arithmetic
- Approximate number system
- Numerical magnitude
Research uses
- Math learning and dyscalculia research
- Numerical cognition research
- Development and education research
How the paradigm works
The Dot Arithmetic Task measures approximate arithmetic: adding quantities without counting, with the same sense of number that tells which of two groups has more. Five-year-olds who have been taught no arithmetic solve it well above chance, and so do adults, with accuracy that falls as the two quantities grow closer in ratio, the signature of approximate number (Barth and colleagues, 2005 and 2006). Children's accuracy on it has been linked to their mathematics achievement in their first school year (Gilmore, McCarthy and Spelke, 2010).
More about the paradigm
On each trial a cloud of dots goes into a gray box in the middle of a dark panel, then a second cloud goes in after it, and then a new cloud shows in a circle on the right. Each cloud shows for 750 ms, too briefly to count, and moves to the box in steps rather than smoothly, as in Barth and colleagues (2006). Participants then press F if the box holds more dots in all or J if the circle had more, with no time limit. The total in the box and the new cloud differ by one of four fixed ratios, 4:7, 2:3, 3:4 and 5:6, with 9 to 30 dots on each side. Each cloud that goes into the box has fewer dots than the new cloud, so neither one alone gives the answer, and every pair of numbers is shown once with the box ahead and once with the circle ahead. On half the trials every dot has the same mean size, so the box's dots also cover more area when they are more; on the other half the dots in the box and in the circle cover the same total area, so the larger number has the smaller dots. Each round has 32 trials, every problem under each control; the standard session has 3 rounds (2 to 4), 96 trials in a seeded order, after 4 practice trials with feedback, about 8 minutes. Every display time and dot layout is fixed before the session. Displays switch on screen frames, the measured times are saved on every trial, and trials whose dots ran more than one frame off plan leave the score. Desktop and laptop only.
The main score is accuracy, the share of right answers. Accuracy at each ratio, the ratio effect, a Weber fraction fitted to the total and the new cloud with the Dot Comparison Task's model, accuracy when the box or the circle had more, the share of answers for the box and accuracy under each control are also reported, with the size, area and spread of every cloud on every trial. Scores support research and do not diagnose.
Scores
- Accuracy%
- Share of scored trials answered with the side that had more dots: the box, with both clouds that went into it, or the new cloud in the circle.
- Higher is better
- Accuracy by Ratio%
- Share right at each ratio of the smaller number to the larger, keyed 4:7, 2:3, 3:4 and 5:6.
- Falls as the ratio nears 1
- Ratio Effect%
- Accuracy at 4:7, the easiest ratio, minus accuracy at 5:6, the hardest.
- Positive values mean answers followed the ratio, as approximate number does
- Weber Fractionscore
- The spread of the judgment as a share of the number, fitted with a guess rate by maximum likelihood to the total and the new cloud at the four ratios, with the Dot Comparison Task's model (Halberda and Feigenson 2008, equation 1), searched from 0.02 to 2. It covers the whole judgment: seeing both clouds, adding them and comparing.
- Lower is better; finer judgments of number
- Weber Fraction at Lowest Valueflag
- 1 when the fit sits at 0.02, the bottom of its range: answers were right at nearly every ratio, so the score is an upper bound.
- Informational
- Weber Fraction at Highest Valueflag
- 1 when the fit sits at 2, the top of its range: answers barely followed the number of dots.
- 1 is flagged for review
- Box-Ahead Accuracy%
- Share right when the box held more dots than the new cloud.
- Higher is better
- Circle-Ahead Accuracy%
- Share right when the new cloud had more dots than the box.
- Higher is better
- Box Answers%
- Share of scored answers naming the box; the box had more dots on half the trials.
- 0.5 means no lean toward either answer
- Same-Size Accuracy%
- Share right when every dot had the same mean size, so the side with more dots also covered more area.
- Higher is better
- Same-Area Accuracy%
- Share right when the dots in the box and in the circle covered the same total area, so the larger number had the smaller dots.
- Higher is better
- Control Effect%
- Same-size accuracy minus same-area accuracy.
- Positive values mean answers leaned on the area or size of the dots
- Median Answer Timems
- Median time from the question to the answer.
- Informational; the answer has no time limit
- Anticipationscount
- Answers under 100 ms from the question, which are not scored.
- Lower is better
- Early Key Pressescount
- F or J pressed before the question appeared, which are not taken as answers.
- Informational
- Frame Intervalms
- Median time between screen frames during the trials.
- About 16.7 on a 60 Hz screen
- Refresh RateHz
- Screen refresh rate from the frame interval.
- Informational
- Displays on Time%
- Share of answered trials whose three clouds each showed within one frame of 750 ms; only these are scored.
- Under 0.9 is flagged for review
- Completed Trialstrials
- Trials answered in the scored run.
- Fewer than planned is flagged for review
- Planned Trialstrials
- Trials the run set out to show.
- Informational
Use and evidence
32 papers from 2021 to 2025 name this task in their title or abstract, down 18% compared with 2016 to 2020. That ranks it 546 of 711 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 9 |
| 2024 | 4 |
| 2023 | 6 |
| 2022 | 4 |
| 2021 | 9 |
| 2020 | 6 |
| 2019 | 4 |
| 2018 | 10 |
| 2017 | 6 |
| 2016 | 13 |
| 2015 | 11 |
| 2014 | 11 |
| 2013 | 6 |
| 2012 | 6 |
| 2011 | 7 |
| 2010 | 7 |
| 2009 | 6 |
| 2008 | 5 |
| 2007 | 4 |
| 2006 | 1 |
| 2005 | 5 |
| 2004 | 5 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 1 |
| 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
- 5
- Registered clinical studies using it as an outcome
- 1
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
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 0.58.0, counted 12 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Display
- On a dark panel, a gray box between two circles 96 px in radius. A cloud of white dots shows in the left circle for 750 ms, moves to the edge of the box for 100 ms and goes behind it; a second cloud follows the same way; then a new cloud shows in the right circle for 750 ms, and the question.
- Answer
- F when the box holds more dots in all and J when the circle had more, with no time limit. Keys pressed while the dots show are counted and not taken as answers.
- Ratios
- 4:7, 2:3, 3:4 or 5:6 between the smaller and the larger number, with 2 pairs at each and 9 to 30 dots on each side; each pair once with the box ahead and once with the circle ahead.
- Clouds
- The total is split into two clouds of unequal size, 4 to 16 dots each, and each has fewer dots than the new cloud, so neither alone gives the answer.
- Controls
- On half the trials every dot has the same mean size, so the side with more dots covers more area; on the other half the box's dots and the circle's cover the same total area, so the larger number has smaller dots. Each dot's diameter varies by up to 20% around its cloud's mean.
- Trials
- 3 rounds of 32 (standard; 2 to 4): 96 trials, every problem under each control once per round, in a seeded order.
- Practice
- 4 trials at ratios of 0.55 to 0.59, with problems not in the test and feedback for 1,000 ms.
- Fixed times
- Every display time and dot layout is set before the session, and no answer changes it: there is no staircase.
- Score
- Accuracy, the share of right answers, with accuracy at each ratio and a Weber fraction fitted with the Dot Comparison Task's model. Trials whose clouds ran more than one frame off plan, and answers under 100 ms, leave the score.
The clouds are drawn by our code with the Dot Comparison Task's controls, and each cloud's mean dot size, total area, convex hull and total circumference are saved. Across the 16 problems, no rule on one count alone, the new cloud's or either cloud's in the box, gets more than 11 right.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: dot_arithmetic_test from the scored run, F (the box) or J (the circle) after the dots, with no time limit (chance 50%; the box has more dots on half the trials), graded from the saved counts. Below chance invalidates and not above chance is flagged; one key on 90% of answers invalidates when accuracy is not above chance. Answer times run from the question. Responses under 100 ms follow the timed-test rule. A run short of its planned trials, a fit at the top of its range, or fewer than 90% of trials with every cloud on time is flagged for review.
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 100 ms from the question was pressed before the question could be read. The display times are part of the design, so trials whose dots ran late leave the score.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Rounds of the 32 trials | 3 | 2 to 4 |
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 Dot Arithmetic Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Barth and colleagues, 2005, 2006; Gilmore, McCarthy and Spelke, 2007, 2010), with the standard settings (rounds of the 32 trials: 3). The main score was accuracy (share of scored trials answered with the side that had more dots: the box, with both clouds that went into it, or the new cloud in the circle).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Barth, H., La Mont, K., Lipton, J., & Spelke, E. S. (2005). Abstract number and arithmetic in preschool children. Proceedings of the National Academy of Sciences, 102(39), 14116-14121. https://doi.org/10.1073/pnas.0505512102Two clouds added behind a screen and compared with a third, each smaller than the third, at ratios of 4:7, 2:3 and 4:5.
- Barth, H., La Mont, K., Lipton, J., Dehaene, S., Kanwisher, N., & Spelke, E. (2006). Non-symbolic arithmetic in adults and young children. Cognition, 98(3), 199-222. https://doi.org/10.1016/j.cognition.2004.09.011Adults' approximate addition at ratios of 0.75 to 0.86, with the second array moving behind the screen in steps.
- Gilmore, C. K., McCarthy, S. E., & Spelke, E. S. (2007). Symbolic arithmetic knowledge without instruction. Nature, 447(7144), 589-591. https://doi.org/10.1038/nature05850
- Gilmore, C. K., McCarthy, S. E., & Spelke, E. S. (2010). Non-symbolic arithmetic abilities and mathematics achievement in the first year of formal schooling. Cognition, 115(3), 394-406. https://doi.org/10.1016/j.cognition.2010.02.002Accuracy on non-symbolic addition at 4:7, 2:3 and 4:5, and its link to mathematics achievement.
- Halberda, J., & Feigenson, L. (2008). Developmental change in the acuity of the "number sense": The approximate number system in 3-, 4-, 5-, and 6-year-olds and adults. Developmental Psychology, 44(5), 1457-1465. https://doi.org/10.1037/a0012682The Weber fraction model with a guess rate.
Source papers
- Barth H, La Mont K, Lipton J, Dehaene S, Kanwisher N, Spelke E (2006). Non-symbolic arithmetic in adults and young children. Cognition, 98(3):199-222.Paradigm source, Adults and children of 5 years
Questions
How long does the Dot Arithmetic Task take?
About 7 to 9 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?
Accuracy, Accuracy by Ratio, Ratio Effect, Weber Fraction, Weber Fraction at Lowest Value, Weber Fraction at Highest Value, Box-Ahead Accuracy, Circle-Ahead Accuracy, Box Answers, Same-Size Accuracy, Same-Area Accuracy, Control Effect, Median Answer Time, Anticipations, Early Key Presses, Frame Interval, Refresh Rate, Displays on Time, Completed Trials and Planned Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Dot Arithmetic Task used in research?
32 papers from 2021 to 2025 name it in their title or abstract, down 18% compared with 2016 to 2020. That ranks it 546 of 711 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.58.0, counted 12 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Dot Arithmetic Task have?
In the published research, 0 papers report its test-retest reliability and 1 registered clinical study uses it as an outcome (evidence grade D in the ConductCognition test census).
Which settings can a researcher change?
Rounds of the 32 trials (2 to 4). 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.