Social Cognition
Illusory Correlation Task
Also called: distinctiveness-based illusory correlation task, Hamilton and Gifford paradigm, group assignment task, stereotype formation task, minority-group illusory correlation
Measure the illusory correlation: participants read statements about two made-up groups with the same mix of acts, then link acts to groups, scored as the phi between group and desirability.
9-12 minUntimed
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
- 41
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures the illusory correlation: linking a smaller group with rarer acts, though both groups show the same mix. Participants read statements about desirable and undesirable acts by members of two made-up groups, then rate each group, place each act in a group and estimate how many acts were undesirable. The main score is the phi between the group chosen and an act's desirability. The last screen explains the design.
Constructs
- Illusory correlation
- Stereotype formation
- Memory for group information
- Judgment of frequency
Research uses
- Research on stereotype formation and intergroup perception
- Research on memory and judgment of how often things occur
- Within-study comparison of judgments about groups
How the paradigm works
An illusory correlation is a link a person sees between two things that are unrelated. Hamilton and Gifford (1976) showed people statements about members of two groups: Group A had twice as many statements as Group B, and both groups did desirable things more than twice as often as undesirable ones, in the same ratio of 9 to 4. People still linked the smaller group with the rarer, undesirable acts: they placed more of the undesirable acts in Group B than its share, and rated Group B less favorably. The smaller group and the rarer kind of act are both less frequent, so their pairings stand out, and the effect has been offered as one way a stereotype of a minority group could form with no real difference between the groups. A meta-analysis found it moderate in frequency estimates and small to moderate in group assignments (Mullen and Johnson, 1990).
More about the paradigm
This test follows Hamilton and Gifford's Experiment 1 with everyday acts written for it in each of the eight languages: 39 statements, each "A member of Group A" or "Group B" and one act, 26 about the larger group (18 desirable, 8 undesirable) and 13 about the smaller one (9 and 4). Each stays on screen for 8 seconds (4 as a setting) with a 1-second blank, and nothing is answered. The participant then rates each group on four words (friendly, considerate, rude and selfish) on bars from not at all to very, places each of the 39 acts in Group A or Group B with the letter left blank, and last, told how many statements each group had, estimates how many described an undesirable act. The groups are only letters; which letter is the smaller group, which acts each group gets and both orders come from the session's seed. The last screen says that the groups and statements were made up, gives the true counts and explains the effect. About 9 to 12 minutes.
The main score is the phi coefficient between the group each act was placed in and whether the act was undesirable, as Hamilton and Gifford computed it for each person: positive when the smaller group gets more than its share of undesirable acts, and 0 for the statements as shown. Next come the shares of undesirable and of desirable acts placed in the smaller group (one in three in truth), the same phi from the frequency estimates with the two estimates, the difference in favorability between the larger and the smaller group, accuracy, the share of acts placed in the larger group and the assignment time. A session that places fewer than half of the acts in the larger group, Hamilton and Gifford's exclusion rule, is flagged for review. The groups are made up and no result describes a real group; the scores support research and do not diagnose.
Scores
- Assignment Phiscore
- Phi between the group each act was placed in, smaller or larger, and whether the act was undesirable.
- 0 matches the statements; positive links the smaller group with undesirable acts
- Smaller Group Share of Undesirable Acts%
- Share of the 12 undesirable acts placed in the smaller group; 4, one in three, were its own.
- Above 0.33 links the smaller group with undesirable acts
- Smaller Group Share of Desirable Acts%
- Share of the 27 desirable acts placed in the smaller group; 9, one in three, were its own.
- Informational
- Estimate Phiscore
- The same phi from the frequency estimates: each group's estimated undesirable statements and the rest.
- 0 matches the statements; positive links the smaller group with undesirable acts
- Smaller Group Undesirable Estimatecount
- Estimated undesirable statements among the smaller group's 13; there were 4.
- Informational
- Larger Group Undesirable Estimatecount
- Estimated undesirable statements among the larger group's 26; there were 8.
- Informational
- Rating Differencescore
- The larger group's favorability minus the smaller group's. Favorability is the mean rating of friendly and considerate minus that of rude and selfish, each rated 0 to 10.
- Positive: the smaller group was rated less favorably
- Smaller Group Favorabilityscore
- The smaller group's mean rating of friendly and considerate minus that of rude and selfish, from -10 to 10.
- Informational
- Larger Group Favorabilityscore
- The larger group's mean rating of friendly and considerate minus that of rude and selfish, from -10 to 10.
- Informational
- Assignment Accuracy%
- Share of acts placed in the group of the member who did them.
- Informational
- Larger Group Share%
- Share of the 39 acts placed in the larger group, which had two in three of the statements.
- Below 0.5 is flagged for review
- Median Assignment Timems
- Median time to place an act in a group.
- Informational
- Assignments Madecount
- Acts placed in a group in the scored run.
- Fewer than planned is flagged for review
- Planned Assignmentscount
- Acts to place: 39.
- Informational
- Earlier Sessionscount
- This participant's earlier sessions of the test, when scoring counted them.
- Above 0 is flagged for review
Use and evidence
41 papers from 2021 to 2025 name this task in their title or abstract, down 25% compared with 2016 to 2020. That ranks it 324 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 6 |
| 2024 | 9 |
| 2023 | 8 |
| 2022 | 5 |
| 2021 | 13 |
| 2020 | 11 |
| 2019 | 13 |
| 2018 | 11 |
| 2017 | 10 |
| 2016 | 10 |
| 2015 | 16 |
| 2014 | 11 |
| 2013 | 13 |
| 2012 | 14 |
| 2011 | 15 |
| 2010 | 5 |
| 2009 | 13 |
| 2008 | 5 |
| 2007 | 9 |
| 2006 | 7 |
| 2005 | 8 |
| 2004 | 6 |
| 2003 | 9 |
| 2002 | 9 |
| 2001 | 12 |
| 2000 | 18 |
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
- 29
- 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 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Groups and statements
- 39 statements written for this test in each of the eight languages, each "A member of Group A" or "Group B" and one everyday act: 26 about the larger group (18 desirable, 8 undesirable) and 13 about the smaller group (9 and 4), the same 9 to 4 ratio. The groups are the letters A and B; which one is smaller, and which acts each gets, is drawn from the session's seed.
- Statements
- One at a time, each for 8 seconds by default (4 as a setting), then a 1,000 ms blank, with nothing to answer. No more than 2 undesirable acts, and no more than 3 statements about the smaller group, in a row.
- Ratings
- Each group on friendly, rude, considerate and selfish, on a bar from not at all to very, scored 0 to 10, with no numbers shown.
- Assignment
- All 39 acts again in a new seeded order, as "A member of Group ___", each placed in Group A or Group B by click or tap, with no time limit.
- Estimates
- Last, told how many statements each group had, the participant splits them into desirable and undesirable acts on a bar.
- Last screen
- Says that the groups and statements were made up, gives the true counts and explains the effect.
- Scores
- The phi between the group an act was placed in and its desirability, the shares of undesirable and desirable acts placed in the smaller group, the same phi from the estimates, and the difference in favorability between the groups.
Hamilton and Gifford (1976) used first names, slides and 20 traits on 10-point scales; this test uses letters only, its own acts and four traits on a 0 to 10 bar, so its numbers are not directly comparable with theirs. As in their Experiment 2, nothing is said beforehand about the groups' sizes. The groups are made up, and no result describes a real group. 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: illusory_correlation_assignment from the scored run, one choice of Group A or Group B for each of the 39 acts, untimed, with the ratings and estimates of the same run. The score is how choices link the groups with kinds of act, not their accuracy, so accuracy is not judged against chance. One group for at least 95% of the acts, fewer than half of the acts placed in the larger group, a run short of its planned assignments, missing estimates and an earlier session of this test are flagged for review. Choices under 200 ms on at least half the acts invalidate. A hidden tab withholds the percentile.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Hamilton and Gifford (1976) left out participants who placed fewer than half of the acts in the larger group; their result held in the full sample in Experiment 1 and not in Experiment 2, so the rule flags rather than excludes. One group for nearly every act leaves the phi undefined. The last screen of a session says that both groups had the same mix of acts, so a second session is flagged. 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. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Seconds per statement | 8 seconds (standard) | 4 seconds (shorter), 8 seconds (standard) |
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 Illusory Correlation Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Hamilton and Gifford, 1976; Mullen and Johnson, 1990), with the standard settings (seconds per statement: 8 seconds). The main score was assignment phi (phi between the group each act was placed in, smaller or larger, and whether the act was undesirable).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Hamilton, D. L., & Gifford, R. K. (1976). Illusory correlation in interpersonal perception: A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 12(4), 392-407. https://doi.org/10.1016/S0022-1031(76)80006-6The design: 26 and 13 statements in a 9 to 4 ratio, each shown for 8 seconds; trait ratings, group assignment and, last, frequency estimates; a phi for each person.
- Mullen, B., & Johnson, C. (1990). Distinctiveness-based illusory correlations and stereotyping: A meta-analytic integration. British Journal of Social Psychology, 29(1), 11-28. https://doi.org/10.1111/j.2044-8309.1990.tb00883.xA meta-analysis: a moderate effect in frequency estimates and a small to moderate one in group assignments.
- Sherman, J. W., Kruschke, J. K., Sherman, S. J., Percy, E. J., Petrocelli, J. V., & Conrey, F. R. (2009). Attentional processes in stereotype formation: A common model for category accentuation and illusory correlation. Journal of Personality and Social Psychology, 96(2), 305-323. https://doi.org/10.1037/a0013778An account of the effect through how attention is shared between the groups as they are learned.
- Van Dessel, P., Ratliff, K., Brannon, S. M., Gawronski, B., & De Houwer, J. (2021). Illusory-correlation effects on implicit and explicit evaluation. Personality and Social Psychology Bulletin, 47(10), 1480-1494. https://doi.org/10.1177/0146167220977706An online version with statements shown for 4 seconds: the effect appeared in trait ratings and not in proportion estimates.
- Johnson, C. (2024). Letter labels and illusory correlation: Infrequent letters bias reactions to the group. The Journal of Social Psychology, 164(6), 978-994. https://doi.org/10.1080/00224545.2023.2224546Letter labels, A and B among them, can bias the effect, so the smaller group's letter is drawn for each session.
Source papers
- Hamilton DL, Gifford RK (1976). Illusory correlation in interpersonal perception: A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 12(4):392-407.N 33, University students
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Illusory Correlation Task take?
About 9 to 12 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?
Assignment Phi, Smaller Group Share of Undesirable Acts, Smaller Group Share of Desirable Acts, Estimate Phi, Smaller Group Undesirable Estimate, Larger Group Undesirable Estimate, Rating Difference, Smaller Group Favorability, Larger Group Favorability, Assignment Accuracy, Larger Group Share, Median Assignment Time, Assignments Made, Planned Assignments and Earlier Sessions. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Illusory Correlation Task used in research?
41 papers from 2021 to 2025 name it in their title or abstract, down 25% compared with 2016 to 2020. That ranks it 324 of 495 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Illusory Correlation 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?
Seconds per statement (4 seconds (shorter), 8 seconds (standard)). 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, with its words in the same language. 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.