Social Cognition
Who to Believe Task
Also called: selective trust task, selective social learning, trust in testimony, informant reliability task, accurate and inaccurate informants
Measure selective trust: two drawn friends name familiar objects, one rightly and one wrongly, then give different made-up names for new objects. Participants choose whom to ask and whose name is right, scored as the share of choices for the accurate friend.
5-7 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
- 126
- Evidence grade
- C
- Registered clinical studies
- 0
What it measures
Measures selective trust: whether people learn new words from someone who named familiar things correctly. Two drawn friends name four familiar objects, one rightly and one wrongly, then give different made-up names for new objects. Participants choose whom to ask and whose name is right. The main score is the percentage of those choices that go to the accurate friend. Built for children on touch screens, with an adult reading the questions, and for adults.
Constructs
- Selective trust
- Learning from testimony
- Epistemic vigilance
- Word learning
Research uses
- Research on how children learn from other people
- Research on trust in accurate and inaccurate informants across development
- Research on word learning from speakers
How the paradigm works
Selective trust is the tendency to learn from people who have been right before. In Koenig and Harris (2005), preschoolers watched two speakers name familiar objects, one correctly and one wrongly, then heard them give different names for new objects; four-year-olds asked and believed the accurate speaker. Pasquini, Corriveau, Koenig and Harris (2007) used four familiar objects, kept the speakers in place and varied how often each was right: 100% against 0%, 75% against 0% and 100% against 25%. Hermes, Rakoczy and Behne (2019) moved the task to a touch screen with drawn informants for children under four, and adults also follow a reliable informant more (Blakey and colleagues, 2025).
More about the paradigm
This version shows two drawn friends, one in blue and one in orange, who keep their sides for the whole session. Each friend says "That's a ..." in the speech kit voice of the session's language, and the sentence appears in a speech bubble. After two practice questions (tap the object named), the friends name four familiar objects: one names each rightly, the other calls it by another object's name, as "That's a shoe" for a ball, and the participant taps the friend whose name is right. The participant then says which friend was not very good at naming things. For four to six new objects, the participant first chooses whom to ask, then hears both friends give different made-up names and taps the name they think is right; the question about who was not very good comes again at the end. Which friend is accurate, the sides, the objects and the made-up words are drawn from the session's seed.
Scores
- Choices for the accurate friend%
- Percent of the choices about new objects, whom to ask and whose name is right, that went to the friend who named the familiar objects right, or right more often.
- Higher means more selective trust; 50% is no preference.
- Asked the accurate friend%
- Percent of new objects for which the participant chose to ask the accurate friend.
- Informational
- Believed the accurate friend%
- Percent of new objects for which the accurate friend's made-up name was chosen as right.
- Informational
- Inaccurate friend namedcount
- Of the two questions about who was not very good at naming things, the number answered with the less accurate friend.
- Informational
- Familiar names judged right%
- Percent of familiar objects named rightly by only one friend for which the participant chose that friend's name.
- Quality check
- Practice correctcount
- Practice questions answered right, of two.
- Quality check
- Median response timems
- Median time from the moment answers open to a choice about a new object, excluding answers under 100 ms.
- Informational
- Choices answeredcount
- Choices made about the new objects, whom to ask and whose name is right.
- Quality check
- Planned choicescount
- Two per new object: 8, 10 or 12.
- Informational
- Anticipationscount
- Choices about new objects made within 100 ms of answers opening.
- Quality check
Use and evidence
126 papers from 2021 to 2025 name this task in their title or abstract, up 38% compared with 2016 to 2020. That ranks it 242 of 711 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 44 |
| 2024 | 25 |
| 2023 | 18 |
| 2022 | 16 |
| 2021 | 23 |
| 2020 | 20 |
| 2019 | 22 |
| 2018 | 22 |
| 2017 | 11 |
| 2016 | 16 |
| 2015 | 11 |
| 2014 | 7 |
| 2013 | 17 |
| 2012 | 8 |
| 2011 | 6 |
| 2010 | 6 |
| 2009 | 3 |
| 2008 | 3 |
| 2007 | 1 |
| 2006 | 1 |
| 2005 | 1 |
| 2004 | 1 |
| 2003 | 1 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade C
100 to 299 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
- 0
- Papers on reliability or validity
- 59
- Registered clinical studies using it as an outcome
- 0
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.
Note on the counts: includes studies of trust in media, experts and AI.
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 13 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Friends
- Two drawn friends, one in blue and one in orange, also told apart by shape. They keep their sides for the whole session; which friend is accurate, and on which side, is drawn from the session's seed.
- Familiar objects
- 4 of 9 familiar objects our code draws: ball, shoe, hat, spoon, apple, key, book, chair and fish. Each friend says That's a ... in turn; the inaccurate friend uses another object's name, as That's a shoe for a ball. After each object, the participant taps the name they think is right.
- Accuracy
- One friend always right and one always wrong in the standard form. Two settings follow Pasquini and colleagues (2007): one friend right 3 of 4 times against one never right, or one always right against one right 1 of 4 times.
- New objects
- 4 in the standard form, up to 6: shapes our code draws, with no name a person knows. For each, the participant chooses whom to ask, then hears the two friends give different made-up words and taps the name they think is right.
- Questions
- Two practice questions come first. Who was not very good at naming things is asked after the familiar objects and again at the end.
- Speech
- Each sentence is spoken by the speech kit voice of the session's language and shown in a speech bubble. The 12 made-up words are written the same in every language, in katakana in Japanese. Answers are taps or clicks with no time limit.
- Scoring
- The share of choices about the new objects, whom to ask and whose name is right, that go to the accurate friend; 50% is no preference. Asking and believing are also scored apart, with the familiar names as a check.
Built for children on touch screens, with an adult to read the questions, and for adults. Sessions are short and make no developmental claim. Norms are off. 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: who_to_believe_ask and who_to_believe_endorse from the most complete run. Choosing either friend is an answer, not an error. Fewer than 75% of planned choices answered invalidates; other incomplete runs, and fewer than half of the familiar names judged right, are flagged. Answers under 200 ms on at least half of 4 or more answers invalidate. A hidden tab is flagged.
Why these rules
Selective trust is a preference between two friends, so there is no chance level to judge choices against. The familiar names check that the participant saw which friend was wrong, as Pasquini and colleagues (2007) checked it. Completion and fast-answer limits are ConductScience choices; response times include listening and looking.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Naming accuracy | One always right, one always wrong (standard) | One always right, one always wrong (standard), One right 3 of 4 times, one always wrong, One always right, one right 1 of 4 times |
| New objects | 4 | 4 to 6 |
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 Who to Believe Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Koenig and Harris, 2005), with the standard settings (naming accuracy: One always right, one always wrong; new objects: 4). The main score was choices for the accurate friend (percent of the choices about new objects, whom to ask and whose name is right, that went to the friend who named the familiar objects right, or right more often).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Koenig, M. A., & Harris, P. L. (2005). Preschoolers mistrust ignorant and inaccurate speakers. Child Development, 76(6), 1261-1277. https://doi.org/10.1111/j.1467-8624.2005.00849.xFamiliar objects named rightly and wrongly, then asking and believing names for new objects.
- Pasquini, E. S., Corriveau, K. H., Koenig, M., & Harris, P. L. (2007). Preschoolers monitor the relative accuracy of informants. Developmental Psychology, 43(5), 1216-1226. https://doi.org/10.1037/0012-1649.43.5.1216Four familiar objects, fixed positions, the question about who was not very good, and the 100%, 75% and 25% conditions.
- Koenig, M. A., Clément, F., & Harris, P. L. (2004). Trust in testimony: Children's use of true and false statements. Psychological Science, 15(10), 694-698. https://doi.org/10.1111/j.0956-7976.2004.00742.xAccurate and inaccurate labelers and learning new names from them.
- Hermes, J., Rakoczy, H., & Behne, T. (2019). Making sense of conflicting information: A touchscreen paradigm to measure young children's selective trust. Infant and Child Development, 28(1), e2119. https://doi.org/10.1002/icd.2119Drawn informants on a touch screen for young children.
Source papers
Questions
How long does the Who to Believe 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?
Choices for the accurate friend, Asked the accurate friend, Believed the accurate friend, Inaccurate friend named, Familiar names judged right, Practice correct, Median response time, Choices answered, Planned choices and Anticipations. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Who to Believe Task used in research?
126 papers from 2021 to 2025 name it in their title or abstract, up 38% compared with 2016 to 2020. That ranks it 242 of 711 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.58.0, counted 13 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Who to Believe 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 C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Naming accuracy (One always right, one always wrong (standard), One right 3 of 4 times, one always wrong, One always right, one right 1 of 4 times) and new objects (4 to 6). 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.