Executive Function
Random Number Generation
Also called: random number generation task, random generation task, RNG task, random digit generation
Measure how well participants hold back habits such as counting: they press a digit from 1 to 9 at each flash of a light, making the series as random as they can.
5-6 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
- 16
- Evidence grade
- C
- Registered clinical studies
- 4
What it measures
Measures how well people hold back habits such as counting while they produce a random series. A light flashes at a steady pace, and participants press a number key from 1 to 9 at each flash, making the order as random as they can.
Constructs
- Inhibition of habitual responses
- Working memory updating
- Response monitoring
Research uses
- Executive-function research
- Schizophrenia and Parkinson's disease research
- Aging research
How the paradigm works
Random number generation measures how well people hold back habits while they produce a series. A light in the middle of the screen flashes once per beat, and at each flash participants press one number key from 1 to 9, making the order of the digits as random as they can, as if drawing a numbered ball from a bag of nine and putting it back each time. People tend to count up or down, avoid repeating a digit and fall back on favorite pairs, and holding these habits back draws on executive control (Baddeley et al. 1998).
More about the paradigm
This is the task of Evans (1978) with a key press in place of a spoken digit and a light in place of a metronome, so it needs no sound. A part has 100 beats, Evans's series length, and the standard session has two parts at one beat per second, his pace, with a rest between them. The light is lit for the first 250 ms of each beat. The first key pressed in a beat is its digit, and later keys are counted but not used. Fifteen practice beats come first, with a note on the next beat after a missed flash or a key other than 1 to 9, and a second practice round when fewer than 12 beats got a digit. The pace can be slowed to 1.5 or 2 seconds, and a session can have one to three parts.
The main score is Evans's RNG index, which measures how unevenly pairs of neighbouring digits are used, with the last digit paired with the first: lower is closer to random. The other indices follow Towse and Neil (1998): redundancy (uneven use of the nine digits), adjacency (pairs one apart, the mark of counting), the turning point index, RNG2 (pairs two places apart), the coupon score, the gap between repeats of a digit, and immediate repeats. Each index is computed per part, leaving out missed beats and other keys, and averaged over parts with at least 50 digits. On 100 digits drawn at random by a computer from 1 to 9, the RNG index averages about 0.28, adjacency about 20% and the turning point index about 94 (ConductScience simulation). Evans reported about 0.30 for people and 0.24 for random number tables, with the digits 1 to 10. Presses under 100 ms after a flash count as anticipations and stay in the series.
Scores
- RNG Indexscore
- Evans's index of how unevenly pairs of neighbouring digits are used, from 0 to 1, averaged over parts; about 0.28 for 100 random digits.
- Lower is closer to random
- Redundancy (R)%
- How unevenly the nine digits are used: 0 when all are used equally often.
- Lower is closer to random
- Adjacency (A)%
- Pairs of digits one apart, up or down, the mark of counting; about 20% for a random series.
- Lower means less counting
- Turning Point Index%
- Peaks and troughs in the series against the 2(n - 2)/3 a random series of continuous values has; about 94 for random digits 1 to 9.
- Lower means long runs up or down
- RNG2 Indexscore
- The RNG index on digits two places apart.
- Lower is closer to random
- Coupon Scorecount
- Digits needed, on average, before all nine have appeared; about 25.5 for a random series.
- Lower means cycling through the digits
- Mean Repetition Gapcount
- Mean distance from a digit to its next use; about 8.2 for 100 random digits.
- Informational
- Median Repetition Gapcount
- Median distance from a digit to its next use; about 5.9 for 100 random digits.
- Informational
- Immediate Repeats%
- The same digit twice in a row, in percent of neighbouring pairs; about 11% for a random series.
- Lower means avoiding repeats
- Digits Pressedcount
- Beats answered with a key from 1 to 9.
- Informational
- Missed Beatscount
- Beats with no key pressed.
- Lower is better
- Missed Beats (%)%
- Missed beats in percent of all beats; over 10% withholds the percentile.
- Lower is better
- Other Keyscount
- Beats whose first key was not 1 to 9.
- Lower is better
- Extra Pressescount
- Keys pressed after the first in a beat, which are not used.
- Informational
- Anticipationscount
- Digits pressed within 100 ms of a flash; they stay in the series.
- Informational
- Median Response Timems
- Median time from a flash to the digit.
- Informational
- Parts Scoredcount
- Parts with at least 50 digits, over which the indices are averaged.
- Informational
- Pacems
- Time from one flash to the next.
- Informational
Use and evidence
16 papers from 2021 to 2025 name this task in their title or abstract, down 30% compared with 2016 to 2020. That ranks it 345 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 3 |
| 2024 | 2 |
| 2023 | 4 |
| 2022 | 3 |
| 2021 | 4 |
| 2020 | 2 |
| 2019 | 4 |
| 2018 | 7 |
| 2017 | 6 |
| 2016 | 4 |
| 2015 | 3 |
| 2014 | 7 |
| 2013 | 4 |
| 2012 | 3 |
| 2011 | 11 |
| 2010 | 5 |
| 2009 | 8 |
| 2008 | 3 |
| 2007 | 4 |
| 2006 | 5 |
| 2005 | 1 |
| 2004 | 5 |
| 2003 | 2 |
| 2002 | 3 |
| 2001 | 4 |
| 2000 | 4 |
Use grade E
Fewer than 30 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
- 8
- Registered clinical studies using it as an outcome
- 4
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
- A light in the middle of a dark panel, lit for the first 250 ms of each beat, with the pressed digit shown under it.
- Pace
- 1,000 ms per beat (standard; 1,500 or 2,000 ms).
- Response
- One number key from 1 to 9 per beat, on the top row or the number pad. The first key counts; later keys in the beat are counted but not used.
- Length
- 2 parts of 100 beats with a rest between them (standard; 1 to 3 parts).
- Practice
- 15 beats at the session's pace, with a note after a missed flash or a wrong key, and once more if fewer than 12 beats got a digit.
- Score
- Evans's RNG index for each part, with the last digit paired with the first, averaged over parts with at least 50 digits.
Evans (1978) had participants say digits from 1 to 10 in time with a metronome. This version uses the digits 1 to 9, one key each, and a light in place of the metronome, so it needs no sound. On a keyboard neighbouring digits are neighbouring keys, so adjacency is not directly comparable with spoken norms.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: random_generation_test beats from the scored run, one per flash of the light. A digit has no right answer, so accuracy is not judged against chance. No key on any beat invalidates, and an unbroken run of missed beats covering half the beats withholds the percentile. Missing more than 25% of the beats invalidates; more than 10% is flagged and withholds the percentile. One digit on 90% of presses invalidates. Keys other than 1 to 9 on more than 10% of presses are flagged. Presses under 100 ms are counted as anticipations and kept.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. The task asks for one digit per flash: a series that misses many flashes was not produced at the pace, and the indices depend on the length of the series (Towse and Neil 1998), so the limits of 10% and 25% are ConductScience choices. Pressing one digit nearly every time cannot follow an instruction to vary the digits. A press within 100 ms of a flash keeps the beat in a paced task, so it is counted but not held against the session.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Time per flash | 1,000 ms (standard) | 1,000 ms (standard), 1,500 ms, 2,000 ms |
| Parts | 2 | 1 to 3 |
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 Random Number Generation task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Evans, 1978; Towse & Neil, 1998), with the standard settings (time per flash: 1,000 ms; parts: 2). The main score was RNG index (Evans's index of how unevenly pairs of neighbouring digits are used, from 0 to 1, averaged over parts; about 0.28 for 100 random digits).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Evans, F. J. (1978). Monitoring attention deployment by random number generation: An index to measure subjective randomness. Bulletin of the Psychonomic Society, 12(1), 35-38. https://doi.org/10.3758/BF03329617
- Towse, J. N., & Neil, D. (1998). Analyzing human random generation behavior: A review of methods used and a computer program for describing performance. Behavior Research Methods, Instruments, & Computers, 30(4), 583-591. https://doi.org/10.3758/BF03209475The randomness indices and how they group.
- Baddeley, A., Emslie, H., Kolodny, J., & Duncan, J. (1998). Random generation and the executive control of working memory. The Quarterly Journal of Experimental Psychology Section A, 51(4), 819-852. https://doi.org/10.1080/713755788Random generation as a measure of executive control.
- Peters, M., Giesbrecht, T., Jelicic, M., & Merckelbach, H. (2007). The random number generation task: Psychometric properties and normative data of an executive function task in a mixed sample. Journal of the International Neuropsychological Society, 13(4), 626-634. https://doi.org/10.1017/S1355617707070786Factor structure and test-retest stability of the spoken task.
Source papers
- Evans FJ (1978). Monitoring attention deployment by random number generation: an index to measure subjective randomness. Bulletin of the Psychonomic Society, 12(1):35-38.Paradigm source, Adults
- Peters M, Giesbrecht T, Jelicic M, Merckelbach H (2007). The random number generation task: psychometric properties and normative data of an executive function task in a mixed sample. Journal of the International Neuropsychological Society, 13(4):626-634.Paradigm source, Adults
Questions
How long does the Random Number Generation take?
About 5 to 6 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?
RNG Index, Redundancy (R), Adjacency (A), Turning Point Index, RNG2 Index, Coupon Score, Mean Repetition Gap, Median Repetition Gap, Immediate Repeats, Digits Pressed, Missed Beats, Missed Beats (%), Other Keys, Extra Presses, Anticipations, Median Response Time, Parts Scored and Pace. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Random Number Generation used in research?
16 papers from 2021 to 2025 name it in their title or abstract, down 30% compared with 2016 to 2020. That ranks it 345 of 395 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Random Number Generation have?
In the published research, 2 papers report its test-retest reliability and 4 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Time per flash (1,000 ms (standard), 1,500 ms, 2,000 ms) and parts (1 to 3). 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.