Social Cognition
Minimal Group Task
Also called: minimal group paradigm, Tajfel matrices, intergroup allocation task, in-group bias task
Measure in-group favoritism in groups made from nothing: dot guesses set the group, then twelve allocation matrices give pull scores.
5-7 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
- 96
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures in-group favoritism in groups made from nothing: guesses of dot numbers place a person in one of two groups, then twelve choices share points between a member of each group. Scores are pull scores of the Tajfel matrices.
Constructs
- In-group favoritism
- Social categorization
- Intergroup bias
- Fairness
Research uses
- Social psychology and intergroup relations research
- Studies of prejudice and group bias
- Social decision-making research in clinical groups
- Within-study group comparison
How the paradigm works
The minimal group task asks whether placing people in groups that mean almost nothing is enough for them to favor their own group (Tajfel, Billig, Bundy and Flament, 1971). In the classic procedure people do a trivial task, such as estimating the number of dots shown briefly on a screen, and are then told they belong to one of two groups (Pinter and Greenwald, 2011). Here the participant guesses how many dots are in ten pictures, each shown for one second. The guesses decide the group: a participant whose guesses were too high on average is in the group of high guessers, and one whose guesses were too low is in the group of low guessers. The groups are named by color, and the screen says the grouping has nothing to do with guessing well. Classic studies assigned the groups at random under a cover story (Blank, 2003); here the group follows the participant's own guesses, so nothing on the screen is untrue. Only guesses that are exactly right on average leave the group to a random draw, and the screen says so.
More about the paradigm
The participant then makes twelve choices. Each shares points between two other people, one from each group and never the participant, by choosing one of seven columns in two rows of numbers. The other people are imagined: by default the instructions say so, and that the points are for a research measure that no one receives. A study that pays for the points or pairs people can use a neutral wording. Three kinds of allocation matrices are used, with numbers written for this test: favoritism against parity, favoritism against joint profit, and maximum difference against in-group and joint profit (Bourhis, Sachdev and Gagnon, 1994). Each kind appears in two forms, one where every strategy points to the same end and one where two strategies point to opposite ends, and each form twice, the second time turned around. A check of the participant's group and one practice matrix with feedback come first. There is no time limit.
The scores are pull scores: how far one strategy draws choices away from another (Blank, 2003). They are computed on the seven-column matrices and doubled to the usual thirteen-column scale, from -12 to 12, as Blank did with his seven-column matrices. The main score is in-group favoritism, the pull of favoritism on parity: positive values mean points tilted toward the member of one's own group where equal points were possible. In minimal group studies people usually allocate more to members of their own group, while many choose the fair middle of the matrices (Blank, 2003), and in-group favoritism in allocations appears with several ways of forming the groups, random assignment included (Pinter and Greenwald, 2011). Scores support research and do not diagnose.
Scores
- In-Group Favoritism (FAV on P)score
- Pull of favoritism on parity, -12 to 12 on the thirteen-column scale: how far choices moved from equal points toward more for the in-group member.
- Positive favors the in-group, 0 is equal points, negative favors the out-group
- Parity on Favoritism (P on FAV)score
- Pull of parity on favoritism, -12 to 12.
- Higher means equal points held choices back from favoring either group
- Favoritism on Joint Profit (FAV on MJP)score
- Pull of favoritism on the most points for both people, -12 to 12.
- Positive means giving up points for both to favor the in-group
- Joint Profit on Favoritism (MJP on FAV)score
- Pull of the most points for both people on favoritism, -12 to 12.
- Higher means choosing the most points for both over favoring a group
- Maximum Difference on Profit (MD on MIP+MJP)score
- Pull of the largest difference in the in-group's favor on the most points for the in-group member and for both, -12 to 12.
- Positive means giving up in-group points to beat the other group
- Profit on Maximum Difference (MIP+MJP on MD)score
- Pull of the most points for the in-group member and for both on the largest difference, -12 to 12.
- Higher means more points over a larger lead
- High-Guess Group0/1
- 1 when the guesses were too high on average, so the participant was in the group of high guessers; 0 when too low.
- Informational
- Guess Ratioscore
- Geometric mean of each guess over the true number of dots.
- Above 1 means guesses ran high
- Group Check Correctflag
- 1 when the participant named their own group on the check, 0 when not.
- 1 supports interpretation
- Practice Correctflag
- 1 when the practice matrix was answered as asked, 0 when not.
- 1 supports interpretation
- Same Column Every Time0/1
- 1 when every matrix was answered in the same column on screen, other than the middle one.
- 1 is flagged for review
- Median Decision Timems
- Median time from a matrix's screen to Confirm.
- Informational
- Matrices Answereditems
- Scored matrices, out of twelve.
- All twelve are needed for every pull score
Use and evidence
96 papers from 2021 to 2025 name this task in their title or abstract, up 55% compared with 2016 to 2020. That ranks it 192 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 29 |
| 2024 | 28 |
| 2023 | 14 |
| 2022 | 13 |
| 2021 | 12 |
| 2020 | 18 |
| 2019 | 7 |
| 2018 | 13 |
| 2017 | 10 |
| 2016 | 14 |
| 2015 | 16 |
| 2014 | 11 |
| 2013 | 10 |
| 2012 | 9 |
| 2011 | 6 |
| 2010 | 8 |
| 2009 | 9 |
| 2008 | 4 |
| 2007 | 12 |
| 2006 | 5 |
| 2005 | 4 |
| 2004 | 3 |
| 2003 | 7 |
| 2002 | 5 |
| 2001 | 10 |
| 2000 | 5 |
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
- 0
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.
- Groups
- 10 pictures of 17 to 58 dots, each for 1,000 ms after a 500 ms blank, with a typed guess after each. Guesses too high on average place the participant in the group of high guessers, and too low in the group of low guessers; the groups are named blue and orange. Only guesses exactly right on average leave the group to a draw from the session's seed, and the screen says so.
- Matrices
- 12, each two rows of 7 numbers written for this test: favoritism against parity, favoritism against joint profit, and maximum difference against in-group and joint profit, each in its convergent and divergent form, and each form upright and turned around.
- Choice
- Each row belongs to an imagined person, one from each group and never the participant. Click or tap a column, then confirm. No column is selected at the start. No time limit.
- Practice
- A check of the participant's group and one practice matrix, both with feedback.
- Scores
- Pull scores on the 7-column matrices, doubled to the 13-column scale: -12 to 12. The main score is the pull of favoritism on parity.
- Other people
- The standard instructions say the other people are imagined, the points are for a research measure and no one receives them. A study that pays for the points or pairs people can use a neutral wording.
- Checks
- A wrong group check, a missed practice matrix and every matrix answered in one column other than the middle are flagged and never exclude.
Tajfel and colleagues (1971), and most studies since, assigned the groups at random under a cover story and used 13-column matrices with their own numbers. This test forms the groups from the participant's own guesses, so nothing it says is untrue, and keeps the three matrix types on 7 columns with numbers of its own, as Blank (2003) used 7 columns. Pull scores from other matrices are not directly comparable. The scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: the scored run's 12 matrices. All must be answered, or the session is invalid. No column is correct, so accuracy is not judged against chance, and repeating a column does not invalidate. A wrong answer on the group check, a missed practice matrix and every matrix answered in the same column (other than the middle) are flagged for review and never exclude. Confirming in under 200 ms on at least half the matrices invalidates. The other people are imagined, as the instructions say.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Each pull score compares a matrix type's two forms, each shown twice, so a missing matrix leaves a score undefined. The group check shows whether the participant took in the group. Half the matrices are shown turned around, so one column every time follows no single way of sharing, but it can still be a real choice, so these checks flag and never exclude. 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.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Who the other people are | Imagined people, points for a research measure (standard) | Imagined people, points for a research measure (standard), The study information explains the points |
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.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Minimal Group Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Tajfel H, Billig MG, Bundy RP, Flament C, 1971; Bourhis RY, Sachdev I, Gagnon A, 1994), with the standard settings (who the other people are: Imagined people, points for a research measure). The main score was in-group favoritism (FAV on P; pull of favoritism on parity, -12 to 12 on the thirteen-column scale: how far choices moved from equal points toward more for the in-group member).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Tajfel, H., Billig, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behaviour. European Journal of Social Psychology, 1(2), 149-178. https://doi.org/10.1002/ejsp.2420010202
- Bourhis, R. Y., Sachdev, I., & Gagnon, A. (1994). Intergroup research with the Tajfel matrices: Methodological notes. In M. P. Zanna & J. M. Olson (Eds.), The psychology of prejudice: The Ontario Symposium (Vol. 7, pp. 209-232). Erlbaum. The matrix types and pull scores.
- Blank, H. (2003). Identifying dominant allocation strategies of individuals in the minimal group paradigm. Current Research in Social Psychology, 9(6), 75-95. Pull scores for each participant and matrices of 7 columns.
- Pinter, B., & Greenwald, A. G. (2011). A comparison of minimal group induction procedures. Group Processes & Intergroup Relations, 14(1), 81-98. https://doi.org/10.1177/1368430210375251Ways of forming minimal groups compared.
Source papers
- Tajfel, Billig, Bundy and Flament (1971): the minimal group procedure. Bourhis, Sachdev and Gagnon (1994) and Blank (2003): the matrix types and pull scores. Pinter and Greenwald (2011): ways of forming minimal groups compared.Paradigm source, Adults
Questions
How long does the Minimal Group Task take?
About 5 to 7 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?
In-Group Favoritism (FAV on P), Parity on Favoritism (P on FAV), Favoritism on Joint Profit (FAV on MJP), Joint Profit on Favoritism (MJP on FAV), Maximum Difference on Profit (MD on MIP+MJP), Profit on Maximum Difference (MIP+MJP on MD), High-Guess Group, Guess Ratio, Group Check Correct, Practice Correct, Same Column Every Time, Median Decision Time and Matrices Answered. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Minimal Group Task used in research?
96 papers from 2021 to 2025 name it in their title or abstract, up 55% compared with 2016 to 2020. That ranks it 192 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 Minimal Group 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).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Who the other people are (Imagined people, points for a research measure (standard), The study information explains the points). 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.