Language
Moses Illusion Task
Also called: semantic illusion task, distorted question task
Measure how often people answer general-knowledge questions without noticing a false detail, such as a wrong name, among facts they know.
8-10 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
- 27
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how often people answer a question without noticing a false detail in it. Participants answer 32 English general-knowledge questions with buttons; half have one detail, such as a name or a place, swapped for a false one of the same kind, and the task is to mark those. A check of each fact comes last. The main score is the share of distorted questions answered among facts the person knew.
Constructs
- Semantic illusion
- Reading for meaning
- Error detection
- Memory retrieval
Research uses
- Language comprehension research
- Misinformation and knowledge research
- Within-study comparison of groups and conditions
How the paradigm works
The Moses Illusion Task measures how often people answer a question without noticing that one of its details is false. The illusion takes its name from the question "How many animals of each kind did Moses take on the Ark?", which most people answer "two" although they know it was Noah, not Moses (Erickson and Mattson 1981). Across Erickson and Mattson's four distorted questions, participants gave the distorted answer 52.3% of the time despite knowing the facts. Speckmann and Unkelbach (2021) found the illusion in a multiple-choice form that offered "This question can't be answered in this form" on every question: distorted questions got the answer they would have had without the false detail 49% of the time in their first experiment and 52.6% in their second, where an open-ended check at the end showed that people knew the critical term in 73.2% of cases, and 52% among the facts they knew.
More about the paradigm
Each session asks 32 general-knowledge questions about well-known stories, music, inventions, places, animals, sport and everyday things, with no time limit. Half of them, decided by one of two lists drawn from the session's seed, have one detail, a name, a place or a thing, swapped for a false one of the same kind, so across sessions every fact is asked as often with its false detail as without it. Each question shows three answers in an order drawn from the seed, then "The question has a false detail" and "I don't know". The instructions say that some questions have a false detail and should be marked, with an example, and two practice questions with feedback come first; the 32 questions have no feedback, and no more than 3 with a false detail, or without one, come in a row. A check then asks about each fact's detail itself, with the true detail, the false one and another of the same kind as choices, plus "I don't know". The last screen lists every question that had a false detail, with its false detail and the true one. The questions were written for this library, none from a published item set, and each fact was checked against a public source; none is about religion, health, politics or a group of people. They are English in every session; the instructions follow the session's language.
The main score is the illusion rate: of the distorted questions about facts the person knew at the check, the share answered rather than marked as having a false detail. "I don't know" counts as neither. The task also reports the rate over all distorted questions, the share of true questions about known facts marked as false, the share of them answered right, the number of facts known, "I don't know" answers and median answer times. Bottoms, Eslick and Marsh (2010) found that failing to notice a false detail made people more likely to use it to answer later general-knowledge questions, so a check after the questions may count a few known facts as unknown. Hannon and Daneman (2001) related resistance to the illusion to skill at reasoning with knowledge from long-term memory and to working memory capacity. Because the last screen names the false details, the test is meant to be taken once by each person, and a repeat session is flagged. Scores describe how often false details went unnoticed in this task, a research measure; they do not diagnose.
Scores
- Illusion Rate%
- Of the distorted questions about facts the person knew at the check, the share answered rather than marked as having a false detail. "I don't know" counts as neither.
- Higher means more false details went unnoticed
- Illusion Rate, All Facts%
- The same share over every distorted question, whether or not the person knew the fact.
- Context for the illusion rate
- Known Distorted Questionscount
- Distorted questions about facts the person knew at the check that were answered or marked, the illusion rate's denominator.
- Under 4 is flagged for review
- False Alarm Rate%
- Of the true questions about facts the person knew, the share marked as having a false detail.
- Lower means fewer true questions were marked
- True Question Accuracy%
- Of the true questions about facts the person knew, the share answered with the right answer.
- Higher is more right answers
- Facts Knowncount
- Facts whose check named the true detail, out of 32.
- Context for the illusion rate
- Checks Answeredcount
- Checks with any answer, "I don't know" included.
- 32 when every check was answered
- I Don't Know Answerscount
- Questions answered "I don't know".
- Context for the rates
- Median Answer Timems
- Median time of right answers to true questions about known facts, leaving out times under 100 ms.
- Context for the rates
- Median Illusion Answer Timems
- Median time of answers to distorted questions about known facts, leaving out times under 100 ms.
- Empty when every such question was marked
- Median Detection Timems
- Median time to mark a distorted question about a known fact, leaving out times under 100 ms.
- Empty when none was marked
- Questions Answeredcount
- Questions with any response in the scored run.
- 32 when every question was answered
- Session In Another Language0/1
- 1 when the session ran in a language other than English; the questions are English in every session.
- Flagged for review
- Earlier Sessionscount
- This participant's earlier sessions of the task, when scoring counted them.
- Above 0 is flagged for review
Use and evidence
27 papers from 2021 to 2025 name this task in their title or abstract, up 69% compared with 2016 to 2020. That ranks it 373 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 9 |
| 2024 | 5 |
| 2023 | 9 |
| 2022 | 4 |
| 2021 | 0 |
| 2020 | 3 |
| 2019 | 2 |
| 2018 | 3 |
| 2017 | 4 |
| 2016 | 4 |
| 2015 | 5 |
| 2014 | 9 |
| 2013 | 5 |
| 2012 | 6 |
| 2011 | 3 |
| 2010 | 2 |
| 2009 | 4 |
| 2008 | 2 |
| 2007 | 1 |
| 2006 | 1 |
| 2005 | 4 |
| 2004 | 4 |
| 2003 | 2 |
| 2002 | 1 |
| 2001 | 1 |
| 2000 | 3 |
Use grade E
Fewer than 30 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 6
- Registered clinical studies using it as an outcome
- 0
Note on the counts: semantic illusion also names other reading effects.
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.
- Questions
- 32 general-knowledge questions written for this test, about well-known stories, music, inventions, places, animals, sport and everyday things. Half have one detail, a name, a place or a thing, swapped for a false one of the same kind. They are English in every session.
- Lists
- Two lists drawn from the session's seed decide which 16 facts are asked with a false detail; each fact appears once in a session.
- Answers
- Three answers in an order drawn from the seed, then "The question has a false detail" and "I don't know", on buttons with no time limit. No more than 3 questions in a row have a false detail, or have none.
- Session
- Instructions with an example, 2 practice questions with feedback, then the 32 questions with none.
- Check
- After the questions, one question about each fact's detail, with the true detail, the false one and another of the same kind as choices, plus "I don't know". A fact is known when its check names the true detail.
- Debrief
- The last screen lists every question that had a false detail, with the false detail and the true one.
- Scoring
- The illusion rate is the share of distorted questions about known facts that were answered rather than marked; "I don't know" counts as neither. Fewer than 4 such questions are flagged for review.
The questions were written for this test, not taken from Erickson and Mattson's (1981) or later item sets, and none is about religion, health, politics or a group of people. The multiple-choice form and the check after the questions follow Speckmann and Unkelbach (2021), whose check was open-ended. Results describe how often false details went unnoticed, a research measure.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: moses_illusion_question from the scored run, one button per question, untimed: one of three answers, the false-detail button or "I don't know", with each fact's check from the same run. Answering a distorted question is the illusion the test measures, not an error, so accuracy is not judged against chance. The same button on 90% of 12 or more questions invalidates, and so do answers under 200 ms on at least half of 4 or more questions. Fewer than 4 distorted questions about known facts, the false-detail button on half or more of the true questions about known facts, no distorted question marked, an earlier session of this test and a session in a language other than English are flagged for review.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. The answers are shuffled for every question, so one button on nearly every question cannot depend on the questions; always marking a false detail cannot either. 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. The illusion rate is read among facts the person knew, so it needs a few such distorted questions; marking most true questions as false, or never marking a distorted one, suggests the instructions were not followed. Every false detail is named on the last screen, so an earlier session is flagged. The word lists are English. A session run in another language shows English words to a participant who may not read English fluently, so it is flagged for review. The limits after the shared ones are ConductScience choices.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
This test runs with fixed settings.
Need a different version of this test?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Instructions and screens: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. Words and sentences: English.
- 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 Moses Illusion Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Erickson & Mattson, 1981; Speckmann & Unkelbach, 2021). The main score was illusion rate (of the distorted questions about facts the person knew at the check, the share answered rather than marked as having a false detail. "I don't know" counts as neither).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Erickson, T. D., & Mattson, M. E. (1981). From words to meaning: A semantic illusion. Journal of Verbal Learning and Verbal Behavior, 20(5), 540-551. https://doi.org/10.1016/S0022-5371(81)90165-1
- Speckmann, F., & Unkelbach, C. (2021). Moses, money, and multiple-choice: The Moses illusion in a multiple-choice format with high incentives. Memory & Cognition, 49(4), 843-862. https://doi.org/10.3758/s13421-020-01128-zThe multiple-choice form with an answer for questions that cannot be answered, and a check of each fact after the questions: distorted questions got the undistorted answer about half the time.
- Reder, L. M., & Kusbit, G. W. (1991). Locus of the Moses illusion: Imperfect encoding, retrieval, or match? Journal of Memory and Language, 30(4), 385-406. https://doi.org/10.1016/0749-596X(91)90013-ALiteral answering with "can't say" for distorted questions, and the partial-match account of the illusion.
- Hannon, B., & Daneman, M. (2001). Susceptibility to semantic illusions: An individual-differences perspective. Memory & Cognition, 29(3), 449-461. https://doi.org/10.3758/BF03196396Individual differences in the illusion.
- Bottoms, H. C., Eslick, A. N., & Marsh, E. J. (2010). Memory and the Moses illusion: Failures to detect contradictions with stored knowledge yield negative memorial consequences. Memory, 18(6), 670-678. https://doi.org/10.1080/09658211.2010.501558False details caught more often when they were common, and unnoticed ones used in later answers.
Source papers
- Erickson TD, Mattson ME (1981). From words to meaning: a semantic illusion. Journal of Verbal Learning and Verbal Behavior, 20(5):540-551.Paradigm source, Adults
- Speckmann F, Unkelbach C (2021). Moses, money, and multiple-choice: the Moses illusion in a multiple-choice format with high incentives. Memory & Cognition, 49(4):843-862.Paradigm source, Adults, Germany and the United States
- Hannon B, Daneman M (2001). Susceptibility to semantic illusions: an individual-differences perspective. Memory & Cognition, 29(3):449-461.Paradigm source, Adults
- Bottoms HC, Eslick AN, Marsh EJ (2010). Memory and the Moses illusion: failures to detect contradictions with stored knowledge yield negative memorial consequences. Memory, 18(6):670-678.Paradigm source, Adults
Questions
How long does the Moses Illusion Task take?
About 8 to 10 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?
Illusion Rate, Illusion Rate, All Facts, Known Distorted Questions, False Alarm Rate, True Question Accuracy, Facts Known, Checks Answered, I Don't Know Answers, Median Answer Time, Median Illusion Answer Time, Median Detection Time, Questions Answered, Session In Another Language and Earlier Sessions. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Moses Illusion Task used in research?
27 papers from 2021 to 2025 name it in their title or abstract, up 69% compared with 2016 to 2020. That ranks it 373 of 495 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Moses Illusion 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 languages can participants use?
The instructions and screens are in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. The words and sentences stay in English, so sessions in other languages are flagged for review. 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.